Datasets:
J0nasW
/
paperswithcode

Dataset card Data Studio Files
xet
 Community
1
paperID
stringlengths
36
36
pwc_id
stringlengths
8
47
arxiv_id
stringlengths
6
16
⌀
nips_id
float64
url_abs
stringlengths
18
329
url_pdf
stringlengths
18
742
title
stringlengths
8
325
abstract
stringlengths
1
7.27k
⌀
authors
stringlengths
2
7.06k
published
stringlengths
10
10
⌀
conference
stringlengths
12
47
⌀
conference_url_abs
stringlengths
16
198
⌀
conference_url_pdf
stringlengths
27
199
⌀
proceeding
stringlengths
6
47
⌀
taskID
stringlengths
7
1.44k
areaID
stringclasses
688 values
embedding
stringlengths
9.26k
12.5k
umap_embedding
stringlengths
29
44
d530a53c-a87a-44e5-b829-874ad185ddef
chart-question-answering-state-of-the-art-and
2205.03966
null
https://arxiv.org/abs/2205.03966v2
https://arxiv.org/pdf/2205.03966v2.pdf
Chart Question Answering: State of the Art and Future Directions
Information visualizations such as bar charts and line charts are very common for analyzing data and discovering critical insights. Often people analyze charts to answer questions that they have in mind. Answering such questions can be challenging as they often require a significant amount of perceptual and cognitive effort. Chart Question Answering (CQA) systems typically take a chart and a natural language question as input and automatically generate the answer to facilitate visual data analysis. Over the last few years, there has been a growing body of literature on the task of CQA. In this survey, we systematically review the current state-of-the-art research focusing on the problem of chart question answering. We provide a taxonomy by identifying several important dimensions of the problem domain including possible inputs and outputs of the task and discuss the advantages and limitations of proposed solutions. We then summarize various evaluation techniques used in the surveyed papers. Finally, we outline the open challenges and future research opportunities related to chart question answering.
['Ahmed Masry', 'Parsa Kavehzadeh', 'Enamul Hoque']
2022-05-08
null
null
null
null
['chart-question-answering', 'chart-question-answering']
['computer-code', 'computer-vision']
[-4.31724405e-03 2.46931121e-01 1.46151215e-01 -3.69989783e-01 -7.52168179e-01 -1.01688409e+00 4.67935950e-01 9.68389094e-01 2.66236812e-01 3.40775400e-01 4.22300309e-01 -1.03934360e+00 -1.02273278e-01 -6.04343593e-01 -1.34157762e-01 -1.35697350e-01 -1.46337613e-01 2.34347105e-01 2.24805072e-01 -1.09463014e-01 8.92587900e-01 4.60115671e-01 -1.83310044e+00 7.18801975e-01 1.22352231e+00 1.00817168e+00 -8.05405434e-03 1.13730574e+00 -1.07395971e+00 1.42569578e+00 -1.34206951e+00 -4.69412237e-01 -1.27530411e-01 -6.73990905e-01 -1.14213264e+00 -5.59595122e-04 5.89226484e-01 3.74899693e-02 2.50382215e-01 1.01109040e+00 4.10276800e-01 2.59949297e-01 4.97503042e-01 -1.69600832e+00 -1.05685437e+00 -1.06142282e-01 -5.64727008e-01 5.11804104e-01 1.04987025e+00 1.62766591e-01 1.18055534e+00 -1.06200731e+00 6.56045973e-01 1.47862947e+00 1.17072113e-01 6.38934746e-02 -1.06947386e+00 -2.63863444e-01 5.21898746e-01 4.86743450e-01 -8.36152256e-01 -1.48392141e-01 9.26452160e-01 -6.62245452e-01 1.00990951e+00 1.00207984e+00 7.39941597e-01 1.23405017e-01 -1.62802134e-02 9.00537968e-01 1.05045450e+00 -6.57000542e-01 5.78262687e-01 1.70462206e-01 5.63542187e-01 6.64311469e-01 1.08058237e-01 -5.33639371e-01 -5.60342729e-01 -2.62867570e-01 6.36615455e-01 -2.12196723e-01 -7.15069696e-02 -5.74776530e-01 -1.03971875e+00 8.94992173e-01 6.26605690e-01 1.63846344e-01 -2.04186842e-01 -1.46978214e-01 6.09018743e-01 5.29006302e-01 2.04268351e-01 9.26827490e-01 2.07830429e-01 -1.13900192e-01 -6.37804091e-01 6.50849104e-01 9.43995416e-01 1.03814662e+00 6.42151773e-01 -8.55388641e-02 -7.27250814e-01 3.94488126e-01 2.03764856e-01 2.12863818e-01 -1.25321403e-01 -8.97831142e-01 6.81809604e-01 1.19904339e+00 5.18512487e-01 -1.25496769e+00 -5.36518395e-01 3.77922691e-02 -3.65793824e-01 6.58292711e-01 5.84809244e-01 -4.68148328e-02 -7.00829327e-01 7.89665878e-01 5.64820707e-01 -8.54364455e-01 -8.81814584e-02 9.12853539e-01 1.18962765e+00 7.16419876e-01 1.66056588e-01 -1.77207395e-01 1.92865407e+00 -1.08243906e+00 -1.36405623e+00 -1.61414117e-01 4.50464249e-01 -8.99928629e-01 1.85081148e+00 1.62852585e-01 -1.05708706e+00 -4.53589439e-01 -8.75065625e-01 -8.18543434e-01 -6.75066173e-01 1.54920354e-01 3.51569802e-01 5.93073368e-01 -9.73592103e-01 -1.67623356e-01 -3.97412628e-01 -5.47864735e-01 3.40966910e-01 -3.21671516e-01 3.16377848e-01 1.09586537e-01 -7.08918154e-01 7.64496028e-01 -1.77814923e-02 -3.11726984e-02 -3.18305343e-01 -8.85667980e-01 -6.69542789e-01 2.69127399e-01 6.67964756e-01 -6.04063511e-01 1.77383745e+00 -3.33797246e-01 -9.74799633e-01 6.57953262e-01 -6.11968696e-01 -1.13671154e-01 5.12669027e-01 -5.15808523e-01 -4.96769816e-01 5.22114933e-01 2.19668150e-01 3.71265501e-01 3.35443884e-01 -1.32088268e+00 -5.48727036e-01 -5.63930333e-01 4.82619911e-01 1.97845057e-01 3.61470468e-02 4.76449519e-01 -3.49057227e-01 -5.72773695e-01 -2.19028071e-01 -1.57447726e-01 -2.57880062e-01 3.64977211e-01 -5.93894064e-01 -5.33098161e-01 1.30777884e+00 -7.67090797e-01 1.99481487e+00 -1.92989671e+00 -4.24007416e-01 2.22055003e-01 4.88378674e-01 6.26279190e-02 2.12155387e-01 1.07339466e+00 -1.99176848e-01 3.14127594e-01 -2.59387463e-01 8.55169222e-02 4.40650545e-02 -1.90006331e-01 -6.09944642e-01 -1.20910980e-01 2.51430631e-01 1.13462603e+00 -9.93738949e-01 -7.15061724e-01 2.85425276e-01 7.67484680e-02 -1.68656960e-01 6.85245216e-01 -7.26691663e-01 4.11368400e-01 -4.76764351e-01 7.85563409e-01 4.83345002e-01 -6.02501631e-01 6.94954768e-02 1.52606919e-01 -6.65714622e-01 4.31970716e-01 -1.11849797e+00 1.29525650e+00 -9.28991958e-02 1.17606938e+00 -4.08415385e-02 -7.86174953e-01 1.38394260e+00 2.35758707e-01 -1.97466165e-02 -1.05309427e+00 -2.43674636e-01 -1.07384346e-01 -1.98318064e-01 -1.08306658e+00 7.78165877e-01 7.57093057e-02 -4.16369364e-03 5.68846464e-01 -5.32389402e-01 -2.59439558e-01 7.24001586e-01 4.17982697e-01 9.16838169e-01 -1.86794996e-01 8.40182006e-01 -1.68291226e-01 6.78410232e-01 6.59727275e-01 -2.19609067e-01 7.02950060e-01 -3.49177390e-01 3.94572735e-01 1.23041058e+00 -8.35488319e-01 -8.52977037e-01 -1.02010810e+00 3.52762431e-01 1.19981039e+00 -8.84598717e-02 -8.76736939e-01 -6.63416624e-01 -5.70301235e-01 -1.05779208e-01 9.97112691e-01 -9.24398899e-01 5.63225567e-01 -3.78817469e-01 -7.28402659e-02 7.24403635e-02 7.07740307e-01 3.14257592e-01 -1.46422231e+00 -1.38311231e+00 -2.26717487e-01 -6.48456216e-02 -6.65712655e-01 -6.63861185e-02 -2.90440887e-01 -1.05037749e+00 -1.46138060e+00 -6.08452082e-01 -7.93536186e-01 7.88734138e-01 5.94988227e-01 1.62944520e+00 2.47643828e-01 -3.66158724e-01 5.55911064e-01 -4.24492508e-01 -9.13543284e-01 -1.71815634e-01 -1.59248233e-01 -8.29383850e-01 -3.41019243e-01 5.55601358e-01 -9.07760635e-02 -7.91524708e-01 1.92786530e-01 -9.82729614e-01 1.33316934e-01 1.57868221e-01 4.14023213e-02 5.04938781e-01 -3.32402349e-01 7.11506367e-01 -1.24279630e+00 1.53626788e+00 -3.95563006e-01 -6.36473477e-01 7.38846302e-01 -5.49872160e-01 1.03474356e-01 6.92985594e-01 2.30283424e-01 -1.26168668e+00 -2.95455068e-01 1.85659289e-01 -4.58518267e-02 -2.37557590e-01 5.52908659e-01 3.01245023e-02 4.06479090e-01 9.44055080e-01 -4.15289521e-01 -3.31106633e-01 -6.66662276e-01 8.22866738e-01 3.09874326e-01 5.99652588e-01 -3.46415251e-01 6.62537873e-01 5.30593216e-01 -1.61942706e-01 -7.49918044e-01 -7.35471487e-01 -7.05553234e-01 -4.83698517e-01 -7.15498269e-01 1.04847753e+00 -1.70509696e-01 -1.17388201e+00 -3.69855881e-01 -1.40463769e+00 6.84095100e-02 -5.65411031e-01 -2.85185784e-01 -3.26593757e-01 1.97683826e-01 3.56736295e-02 -1.39548922e+00 -4.40427333e-01 -9.29790974e-01 8.02860260e-01 6.58675313e-01 -6.76208436e-01 -9.84443665e-01 2.74081498e-01 2.90266663e-01 4.49639320e-01 8.53665590e-01 1.43738925e+00 -3.53877306e-01 -6.91106796e-01 -3.11885029e-02 -5.82095921e-01 -5.77799916e-01 1.96550488e-01 2.19656646e-01 -1.05487132e+00 6.95410967e-02 -1.20590150e-01 -3.51891875e-01 5.09064913e-01 1.57981649e-01 1.36209905e+00 -3.82440805e-01 -1.02026701e-01 -1.14031099e-01 1.25954020e+00 5.64031422e-01 4.08051252e-01 2.34402135e-01 6.90038502e-01 1.16789317e+00 8.35602582e-01 4.33387607e-01 6.74085319e-01 1.13201641e-01 5.60512245e-01 -2.16084078e-01 1.04124166e-01 -5.27481616e-01 -2.36061394e-01 5.31409323e-01 7.07108453e-02 -1.44391313e-01 -1.29403186e+00 6.26312315e-01 -1.96129215e+00 -6.86125755e-01 -6.60590112e-01 1.99716914e+00 1.99165180e-01 -2.18907133e-01 5.16018212e-01 5.29133499e-01 5.79380810e-01 4.76442337e-01 -6.57751739e-01 -9.48985279e-01 2.27516636e-01 -9.67865288e-02 -3.95603418e-01 4.26027209e-01 -7.38558292e-01 3.30806345e-01 7.11027956e+00 7.72967795e-03 -6.85134411e-01 -4.55186933e-01 6.60694718e-01 2.20401525e-01 -5.69439828e-01 2.08648279e-01 4.86654043e-02 -2.09917113e-01 8.11591804e-01 -4.86351103e-01 7.40642771e-02 8.20895016e-01 4.03810501e-01 -5.60446262e-01 -1.25749731e+00 1.09023368e+00 2.89514512e-02 -1.51583016e+00 2.54575402e-01 -4.60757971e-01 4.43711847e-01 -8.51327896e-01 1.63092926e-01 -9.68687311e-02 2.45122120e-01 -1.13645494e+00 6.37630999e-01 6.46588266e-01 7.43834972e-01 -7.45697975e-01 1.58476606e-01 3.98105457e-02 -1.45894182e+00 -2.68935978e-01 -1.07836984e-01 -6.43321276e-01 1.34203956e-01 3.40792686e-01 -7.77603626e-01 6.18484497e-01 1.04521167e+00 2.50907034e-01 -1.12570012e+00 1.40906370e+00 -5.35180151e-01 4.72922295e-01 2.21707389e-01 -4.93250281e-01 3.33564095e-02 -1.65731072e-01 2.85666466e-01 1.09758794e+00 8.61538947e-02 6.95473924e-02 1.39143318e-02 9.04080331e-01 1.11620031e-01 5.15526235e-01 -6.27268851e-01 -2.59702742e-01 3.55157822e-01 1.04404438e+00 -9.50010002e-01 -5.28079569e-01 -7.30904281e-01 4.44469899e-01 4.82504219e-01 7.32157230e-01 -1.74718574e-01 -8.00376594e-01 3.36481541e-01 1.49121165e-01 -7.04214200e-02 -1.90124020e-01 -9.17581260e-01 -8.32045496e-01 4.25758749e-01 -1.04184926e+00 1.00375998e+00 -1.24865353e+00 -8.59846413e-01 3.98849726e-01 1.36604294e-01 -1.24612439e+00 -3.67257863e-01 -5.35644829e-01 -9.37282741e-01 8.87650430e-01 -1.27654290e+00 -4.30131197e-01 -7.86372364e-01 1.66020066e-01 6.57395065e-01 4.26428705e-01 9.27337170e-01 -1.56938896e-01 -2.58636594e-01 -1.64480880e-01 -1.40681207e-01 1.81013033e-01 5.69157422e-01 -1.77225351e+00 8.96806240e-01 7.82697201e-01 1.61127999e-01 6.36650622e-01 9.74105358e-01 -5.35025358e-01 -1.36387789e+00 -5.65835536e-01 9.70000088e-01 -7.28647709e-01 6.15279615e-01 -6.88291609e-01 -1.38257945e+00 4.63160515e-01 6.65704548e-01 -2.10631952e-01 1.03627574e+00 1.36098310e-01 -5.30445039e-01 6.79136738e-02 -9.72463846e-01 8.17386925e-01 4.77104217e-01 -7.64523923e-01 -8.04945409e-01 8.05669799e-02 7.62089729e-01 -2.99379081e-01 -4.34257030e-01 -2.65235424e-01 4.32350010e-01 -1.34951019e+00 6.55132711e-01 -9.34372425e-01 3.35976988e-01 -7.36877918e-01 3.79107237e-01 -1.10998952e+00 -1.23112582e-01 -8.52119446e-01 -3.66319895e-01 9.26422298e-01 3.61698717e-01 -2.08837077e-01 6.58881843e-01 1.02151573e+00 6.21385388e-02 -7.63258696e-01 -3.90798479e-01 -1.80624425e-01 -1.65343598e-01 -5.41810513e-01 7.21091688e-01 7.08079159e-01 6.70302689e-01 6.23715937e-01 1.79395583e-02 -1.56713232e-01 2.13012636e-01 5.97608507e-01 1.01792777e+00 -1.37689912e+00 4.68161464e-01 -5.67415953e-01 3.75873037e-02 -1.03497934e+00 -6.71984673e-01 -5.01230121e-01 -3.61745477e-01 -2.41026187e+00 -9.65615213e-02 3.44040066e-01 1.56120241e-01 4.55485471e-02 -5.28918803e-01 -2.93079257e-01 6.87985718e-01 1.97745398e-01 -8.70877028e-01 1.50661886e-01 1.59280157e+00 5.72377071e-03 -4.61917639e-01 -1.57133952e-01 -1.05093491e+00 4.87157345e-01 7.93456018e-01 -2.45469958e-02 -8.51498902e-01 -4.73670006e-01 9.02826726e-01 3.76926780e-01 3.36740226e-01 -5.77974439e-01 4.40007448e-01 -1.68892398e-01 5.19472301e-01 -1.12552965e+00 -5.79983890e-02 -7.30021298e-01 -5.76614320e-01 2.53703237e-01 -6.70329154e-01 1.10837364e+00 2.67225891e-01 6.82425022e-01 -5.61582565e-01 2.60441750e-02 3.58026266e-01 -1.42216980e-01 -5.70142448e-01 -3.46167475e-01 -5.69097340e-01 2.95772910e-01 9.88535881e-01 -1.53886139e-01 -6.95562124e-01 -8.43051851e-01 -4.29617256e-01 8.86634648e-01 1.21310137e-01 5.58333457e-01 7.55815268e-01 -1.21719348e+00 -3.73247623e-01 -1.08023465e-01 5.47635615e-01 9.60188508e-02 2.15965316e-01 3.65711480e-01 -8.61519158e-01 7.02073574e-01 -1.62065208e-01 -3.74333858e-01 -1.17398846e+00 1.08747876e+00 -2.35855971e-02 -2.58535087e-01 -5.25277972e-01 3.67107391e-01 3.53115112e-01 -6.44350350e-02 3.85160685e-01 -8.06549549e-01 -7.65723467e-01 3.05991620e-01 1.11910903e+00 9.36126947e-01 5.58293946e-02 -1.13809824e-01 -3.88107449e-01 2.22355202e-01 2.08725646e-01 -1.88950300e-01 7.67781734e-01 -1.67211056e-01 9.48714837e-03 8.43904734e-01 8.63999188e-01 -1.35011137e-01 -9.04191971e-01 7.77183697e-02 6.88263953e-01 -7.06890285e-01 -6.48326337e-01 -9.57903683e-01 -3.60576659e-01 1.53412855e+00 2.76997000e-01 1.15925753e+00 1.20812929e+00 2.52500296e-01 9.98317599e-02 4.85678673e-01 -3.59157711e-01 -1.08275497e+00 3.90839070e-01 2.38429248e-01 1.62275040e+00 -1.15562201e+00 -5.28500341e-02 -4.80279833e-01 -8.62621009e-01 1.39279985e+00 7.62032032e-01 -3.47781889e-02 3.62522930e-01 6.05451576e-02 6.57754004e-01 -8.45722616e-01 -8.84526312e-01 -1.90172657e-01 4.67635036e-01 8.37793171e-01 8.56987953e-01 -2.62028780e-02 -2.02880308e-01 1.63693383e-01 -5.94372928e-01 -1.60204872e-01 4.81460214e-01 1.16907060e+00 -6.50565445e-01 -6.46929204e-01 -1.00109899e+00 5.59833050e-01 -2.67405182e-01 2.17454389e-01 -1.01980770e+00 7.70185530e-01 -7.42830932e-01 1.40107632e+00 2.34566405e-01 1.48495212e-01 7.31987774e-01 2.42461264e-01 -1.24903031e-01 -4.42193508e-01 -6.83570862e-01 -2.72846878e-01 1.18757166e-01 -5.06272376e-01 -3.14277500e-01 -5.22903442e-01 -1.46166611e+00 1.13540888e-01 3.61484855e-01 4.66578841e-01 4.98266190e-01 4.74241614e-01 4.16396230e-01 7.05293000e-01 6.81731030e-02 2.90022139e-02 -9.60758701e-02 -8.45657766e-01 -6.44395724e-02 3.12290043e-01 7.83881009e-01 -3.30430120e-01 1.52063042e-01 2.05747053e-01]
[11.207382202148438, 2.0307939052581787]
3f99eed1-f230-450e-a2c2-5832644c9f52
deepmcat-large-scale-deep-clustering-for
2110.00109
null
https://arxiv.org/abs/2110.00109v1
https://arxiv.org/pdf/2110.00109v1.pdf
DeepMCAT: Large-Scale Deep Clustering for Medical Image Categorization
In recent years, the research landscape of machine learning in medical imaging has changed drastically from supervised to semi-, weakly- or unsupervised methods. This is mainly due to the fact that ground-truth labels are time-consuming and expensive to obtain manually. Generating labels from patient metadata might be feasible but it suffers from user-originated errors which introduce biases. In this work, we propose an unsupervised approach for automatically clustering and categorizing large-scale medical image datasets, with a focus on cardiac MR images, and without using any labels. We investigated the end-to-end training using both class-balanced and imbalanced large-scale datasets. Our method was able to create clusters with high purity and achieved over 0.99 cluster purity on these datasets. The results demonstrate the potential of the proposed method for categorizing unstructured large medical databases, such as organizing clinical PACS systems in hospitals.
['Daniel Rueckert', 'Ben Glocker', 'Wenjia Bai', 'Turkay Kart']
2021-09-30
null
null
null
null
['image-categorization']
['computer-vision']
[ 2.67480046e-01 3.35019499e-01 -3.34970132e-02 -6.82280123e-01 -1.09639692e+00 -5.43957770e-01 2.99937893e-02 9.01557744e-01 -5.07271409e-01 6.22384548e-01 3.25335823e-02 -1.18204184e-01 -3.83739501e-01 -5.02249122e-01 -2.99166024e-01 -9.26051855e-01 -5.42370714e-02 1.04396212e+00 -4.14989218e-02 5.94682455e-01 2.59894520e-01 4.62705195e-01 -1.65517831e+00 5.55397809e-01 8.46996605e-01 7.74987221e-01 2.01594859e-01 5.14819860e-01 -1.20343372e-01 1.14944410e+00 -6.27736628e-01 -1.50363624e-01 -2.61011743e-03 -4.60005462e-01 -1.07375896e+00 6.10713184e-01 6.66727126e-02 2.38096967e-01 1.86833024e-01 1.11024427e+00 7.96032846e-01 -4.40256089e-01 8.06974232e-01 -1.08770370e+00 -3.89027148e-02 6.68581903e-01 -5.06689191e-01 -6.94644079e-02 -2.01909766e-02 -2.39632189e-01 6.27010584e-01 -4.59168851e-01 9.24716651e-01 6.23439312e-01 6.55050278e-01 4.28680152e-01 -1.18659735e+00 -5.54328382e-01 -5.85365117e-01 1.08031956e-02 -1.47607970e+00 -4.38319355e-01 6.24657512e-01 -1.01004493e+00 4.10476506e-01 3.57014477e-01 2.77400702e-01 5.04722178e-01 1.44660741e-01 3.47848743e-01 1.46134245e+00 -6.09674871e-01 4.62603033e-01 3.96553040e-01 3.71435910e-01 6.59586191e-01 5.08344293e-01 -3.45412165e-01 -2.27188803e-02 -5.04033864e-01 4.04717267e-01 2.70475484e-02 -2.14837436e-02 -6.24942541e-01 -1.62040949e+00 7.23345876e-01 2.22087562e-01 6.53793514e-01 -4.78626370e-01 -5.00817776e-01 6.88491404e-01 2.40251962e-02 3.70493591e-01 7.36197591e-01 -2.99929678e-01 4.69270116e-03 -1.30600691e+00 -2.55229831e-01 6.48475289e-01 8.99979830e-01 4.67304230e-01 -4.57702667e-01 -1.15591832e-01 8.69373024e-01 9.61062219e-03 1.46391332e-01 8.03651989e-01 -8.63783360e-01 1.19307220e-01 6.69590235e-01 -3.06474492e-02 -1.11319935e+00 -8.57059717e-01 -3.14372838e-01 -1.22599554e+00 4.50696461e-02 5.33276260e-01 -7.81345069e-02 -1.10063541e+00 1.28427625e+00 3.88411373e-01 -2.67317235e-01 -3.28179225e-02 9.36467946e-01 1.02724195e+00 1.10664278e-01 3.65647167e-01 -6.76888585e-01 1.32865739e+00 -7.26855695e-01 -1.01920164e+00 4.14086372e-01 1.11216319e+00 -7.72195518e-01 8.35420251e-01 5.24894714e-01 -1.02156973e+00 -4.14986044e-01 -6.32301867e-01 3.88939321e-01 -2.33670264e-01 4.79249686e-01 3.66495103e-01 7.62234271e-01 -9.99207735e-01 4.71921653e-01 -9.75489080e-01 -3.71982485e-01 8.19979370e-01 5.65760434e-01 -4.95956242e-01 -4.03553285e-02 -6.02558076e-01 6.26780212e-01 5.82865059e-01 -1.52309701e-01 -5.97320616e-01 -5.73881149e-01 -5.57711899e-01 -1.78737104e-01 2.67547876e-01 -4.43230361e-01 8.74096274e-01 -9.26431298e-01 -1.06210184e+00 1.46957517e+00 -2.57245824e-02 -2.69141585e-01 5.00642538e-01 3.18555206e-01 -3.43238026e-01 3.40750158e-01 3.09356153e-01 5.36287904e-01 4.42005366e-01 -1.44711661e+00 -5.31007588e-01 -5.58231115e-01 -5.34769058e-01 -8.75148270e-03 -2.89555192e-01 1.38905153e-01 -1.37358487e-01 -5.62627792e-01 6.08620524e-01 -1.09780169e+00 -5.93622684e-01 -4.52235878e-01 -5.75712740e-01 -2.69840541e-03 3.03317070e-01 -5.25201023e-01 1.20645738e+00 -2.00229526e+00 -1.41358688e-01 3.57729614e-01 6.47770584e-01 1.02798320e-01 5.53305030e-01 1.50430843e-01 -2.95771956e-01 2.57745177e-01 -3.67309958e-01 -5.03653586e-02 -3.72131377e-01 1.28583610e-01 1.55454397e-01 6.66128516e-01 -1.17248386e-01 5.53054810e-01 -9.56648409e-01 -1.19262111e+00 1.81001455e-01 9.01562572e-02 -5.94846904e-01 3.67946625e-01 2.27059484e-01 8.91466618e-01 -1.75351083e-01 7.39889622e-01 5.27648032e-01 -7.14147925e-01 4.79876816e-01 -3.22658569e-01 2.24552676e-02 6.23171106e-02 -1.04415607e+00 1.77619433e+00 2.52096038e-02 2.94315070e-01 -6.18660785e-02 -1.29055882e+00 7.35404909e-01 6.13854885e-01 1.22083783e+00 -4.51447576e-01 3.70818526e-01 2.80087560e-01 1.18356384e-01 -8.74865234e-01 2.24667579e-01 -4.77377892e-01 -8.84574577e-02 6.09169066e-01 -1.00368103e-02 -8.87489617e-02 1.98764473e-01 1.79588825e-01 1.03934741e+00 -4.78585660e-01 3.28625917e-01 -5.88201284e-01 4.41376597e-01 5.25468349e-01 4.24983710e-01 6.10357225e-01 -3.67010534e-01 1.03939891e+00 3.53773952e-01 -4.32529151e-01 -1.15109241e+00 -7.83425689e-01 -5.26114821e-01 5.88464081e-01 -8.19910690e-02 -4.00213867e-01 -9.12032366e-01 -8.23750079e-01 -3.79633099e-01 2.12443233e-01 -4.40453976e-01 6.70591518e-02 -3.87858361e-01 -1.00029027e+00 4.46787715e-01 2.08939880e-01 -4.79663424e-02 -1.13054514e+00 -7.11179793e-01 2.70426363e-01 -5.05630851e-01 -1.11662829e+00 -4.97957245e-02 3.45748395e-01 -1.26053727e+00 -1.24104929e+00 -6.67620182e-01 -9.81801152e-01 1.32017517e+00 -1.61701024e-01 1.40695941e+00 -2.21968777e-02 -6.86960101e-01 1.01721108e-01 -4.24589425e-01 -3.98664415e-01 -6.37940049e-01 1.99170530e-01 -3.23376292e-03 -3.48543585e-03 3.74003470e-01 -2.38034189e-01 -6.29430056e-01 4.18122441e-01 -1.07534266e+00 1.14173684e-02 6.17568493e-01 8.34750056e-01 9.55813110e-01 3.05387437e-01 6.23517096e-01 -1.65431225e+00 2.93401539e-01 -4.91216093e-01 -2.55049944e-01 1.68701291e-01 -8.57916355e-01 -8.36686119e-02 5.15849710e-01 -1.92191213e-01 -6.83655143e-01 4.19203073e-01 -8.25387314e-02 -2.64494866e-01 -6.83554471e-01 4.42220181e-01 6.98169023e-02 1.90581501e-01 7.91916132e-01 -1.06159553e-01 1.12538584e-01 -2.86701083e-01 1.43309236e-01 1.20188129e+00 4.03926194e-01 -5.41252136e-01 3.67526442e-01 7.27611065e-01 -9.28344280e-02 -6.17543578e-01 -8.46417010e-01 -8.42380643e-01 -9.43687081e-01 -2.91606635e-01 1.07436240e+00 -8.82416189e-01 -4.13042814e-01 2.96506315e-01 -6.79735303e-01 -7.02256858e-02 -4.11452264e-01 5.80704331e-01 -4.53683048e-01 4.10763979e-01 -6.85108423e-01 -4.88349646e-01 -3.44774812e-01 -1.28594804e+00 9.57926095e-01 -1.05330078e-02 -3.93603176e-01 -7.85349667e-01 7.53029138e-02 6.64592683e-01 2.46716782e-01 6.99396312e-01 9.64935839e-01 -7.98825443e-01 -1.62317529e-01 -1.86838582e-01 -1.99080974e-01 2.07858860e-01 4.44574177e-01 -2.15745822e-01 -8.94999027e-01 -4.22363877e-01 2.48084471e-01 -5.17432928e-01 4.14879829e-01 4.90443707e-01 1.51319742e+00 -1.55332640e-01 -4.01932150e-01 2.78016925e-01 1.49476635e+00 1.53070360e-01 5.47670782e-01 2.68846691e-01 7.47537255e-01 8.54165256e-01 8.40167284e-01 5.95771432e-01 3.09759319e-01 2.30928406e-01 1.57203466e-01 -4.50662971e-01 9.67251882e-02 2.47695401e-01 -3.96407962e-01 1.29215848e+00 -1.66011769e-02 1.67727754e-01 -1.36821151e+00 8.71538818e-01 -1.67002201e+00 -5.44202507e-01 -4.75646406e-01 2.23378825e+00 1.06948745e+00 -2.70143840e-02 1.16387106e-01 5.20973563e-01 9.83913481e-01 -3.80392730e-01 -2.28912070e-01 5.59139512e-02 8.40032995e-02 1.65643141e-01 5.53326607e-01 -2.29591411e-02 -1.21637654e+00 4.56093431e-01 6.46588469e+00 5.93018770e-01 -1.12907803e+00 4.10871089e-01 1.08787000e+00 -7.22772479e-02 2.06834048e-01 -3.27225149e-01 -3.26996028e-01 5.24771035e-01 1.10460877e+00 1.21987961e-01 -2.61527568e-01 8.23479116e-01 1.51269019e-01 -1.63059533e-01 -9.65692937e-01 1.17276227e+00 9.20659006e-02 -1.14927518e+00 -1.65895745e-01 1.81534007e-01 9.53535974e-01 2.30966900e-02 -2.34002575e-01 -1.22799575e-01 1.58763230e-01 -1.11805952e+00 4.19274241e-01 3.25767994e-01 1.05568445e+00 -6.32295549e-01 1.10045099e+00 4.54821169e-01 -6.41213059e-01 1.45256966e-01 -2.59133250e-01 2.11445898e-01 -2.47269273e-02 1.03087807e+00 -1.02918375e+00 4.54892874e-01 8.15346718e-01 4.89070326e-01 -6.30915880e-01 1.14189386e+00 3.86940271e-01 6.88265622e-01 -1.53600559e-01 3.47824484e-01 3.15592922e-02 -1.38266087e-01 -5.60268871e-02 1.32072604e+00 1.76853478e-01 2.93224424e-01 3.60463947e-01 2.15921044e-01 4.34510075e-02 4.67777044e-01 -4.04459625e-01 -6.07377328e-02 1.68939680e-01 1.56196153e+00 -1.44355726e+00 -6.27740920e-01 6.38017757e-03 4.85741436e-01 1.67116344e-01 -1.34440213e-01 -8.06794643e-01 -1.76507264e-01 -9.99270901e-02 4.51289475e-01 3.41864652e-03 1.03469022e-01 -5.52734733e-01 -1.08656645e+00 -1.26288906e-01 -9.55688417e-01 7.52981961e-01 -4.67161894e-01 -1.37872040e+00 9.10514772e-01 -1.50548309e-01 -1.45129013e+00 -5.55866398e-02 -3.34631085e-01 2.09279686e-01 2.79322982e-01 -1.21776962e+00 -7.91177034e-01 -6.06658757e-01 5.93128383e-01 1.15087584e-01 -1.52574822e-01 9.31336462e-01 8.12513530e-01 -2.32964799e-01 3.91177446e-01 2.60757297e-01 3.02132547e-01 1.09470844e+00 -1.38225257e+00 -6.12824500e-01 2.46534526e-01 -1.20267868e-02 4.73541766e-01 5.77005148e-01 -3.97672623e-01 -8.05947363e-01 -1.22413135e+00 9.18482602e-01 -5.28250277e-01 2.23419279e-01 -1.49754807e-01 -8.71041179e-01 3.23109776e-01 6.38998598e-02 2.20746204e-01 1.28350723e+00 -1.01884581e-01 -3.02861053e-02 -4.64325063e-02 -1.56524038e+00 9.39626363e-04 7.09688485e-01 -2.33539373e-01 -4.97698009e-01 7.14313626e-01 2.53250182e-01 -4.53645468e-01 -1.40981245e+00 5.86575806e-01 2.46815369e-01 -1.00791883e+00 8.00308466e-01 -5.43396831e-01 4.59980488e-01 -4.10594672e-01 1.01950467e-01 -8.80696356e-01 -3.02011371e-01 -1.46710575e-01 4.41006601e-01 1.22461665e+00 3.33465606e-01 -4.48143184e-01 9.57612991e-01 5.63174725e-01 -6.79549128e-02 -6.80677116e-01 -6.86786115e-01 -3.24673057e-01 6.91381609e-03 -1.28653213e-01 3.09398741e-01 1.41848755e+00 1.75862938e-01 8.20462182e-02 -3.53286117e-02 9.98255238e-02 8.62852275e-01 3.09108853e-01 4.42121387e-01 -1.42232800e+00 3.18926945e-02 -2.42817909e-01 -5.38008511e-01 -3.38931121e-02 -1.24158375e-01 -1.11552083e+00 2.47622117e-01 -1.49889386e+00 6.00867510e-01 -1.07347667e+00 -4.42008704e-01 3.75264943e-01 -1.50989205e-01 7.57727265e-01 -1.47223681e-01 8.30996931e-01 -9.72629845e-01 -1.82013497e-01 9.89367545e-01 5.43362927e-03 3.48259485e-03 -1.23988159e-01 -6.02693319e-01 7.24917710e-01 1.02426541e+00 -1.03364563e+00 -1.96852013e-01 -5.25445081e-02 7.94402882e-02 1.26909942e-03 -8.02536681e-02 -1.22828698e+00 3.19286734e-01 1.47786155e-01 2.38491148e-01 -7.00339675e-01 -4.42306876e-01 -9.90726769e-01 4.96928096e-01 5.10115862e-01 -5.37342787e-01 -5.09718657e-02 -1.38636991e-01 3.57401311e-01 -6.06874764e-01 -2.22951651e-01 1.05334282e+00 -3.97510827e-01 -1.80247530e-01 6.76144883e-02 -3.22229952e-01 2.63105899e-01 1.27752030e+00 -1.34470891e-02 -3.67797054e-02 6.52471632e-02 -1.08883440e+00 3.81927341e-02 5.90402782e-01 9.42070112e-02 3.06865811e-01 -1.16161954e+00 -6.77904129e-01 -6.81456476e-02 3.73967588e-01 2.95195937e-01 3.39511573e-01 1.12378573e+00 -9.13553476e-01 5.65820515e-01 -2.40932941e-01 -1.26092446e+00 -1.40136015e+00 7.53657818e-01 -2.98375823e-02 -6.04232132e-01 -4.92089510e-01 2.82805860e-01 -4.24237400e-02 -7.03605235e-01 2.97040880e-01 -2.89014012e-01 -3.58364224e-01 2.26989269e-01 2.41020024e-01 2.48009026e-01 4.79608268e-01 -7.87856281e-01 -4.35661405e-01 5.45674205e-01 3.27102058e-02 1.70642719e-01 1.29716372e+00 -9.81044844e-02 -4.87628669e-01 6.09766901e-01 1.14934421e+00 -2.03107283e-01 -4.01735425e-01 -6.10887222e-02 3.13805431e-01 -1.53463185e-01 -9.13762487e-03 -7.21577168e-01 -1.18984556e+00 7.16298699e-01 8.17255199e-01 2.47724041e-01 1.09982872e+00 1.46593049e-01 4.90821451e-01 2.29526237e-01 5.84955037e-01 -1.21529841e+00 -1.39407963e-01 -1.52294144e-01 2.24226251e-01 -1.60085392e+00 9.47499946e-02 -5.65771282e-01 -8.20647657e-01 8.44333172e-01 2.67037541e-01 5.96420132e-02 8.62249672e-01 3.31547439e-01 4.88802105e-01 -5.65392971e-01 -3.95382404e-01 1.63734090e-02 4.17095274e-02 4.53437120e-01 7.63458967e-01 3.01898181e-01 -5.91740429e-01 3.78015339e-01 -9.05776471e-02 3.15101653e-01 4.94926870e-01 1.17377341e+00 -1.92903027e-01 -1.11282885e+00 -6.07096493e-01 8.06873083e-01 -9.36526358e-01 1.53394133e-01 -1.09727517e-01 4.77257282e-01 3.74462932e-01 1.05451047e+00 -7.68873319e-02 -2.05177262e-01 1.35910600e-01 4.79972064e-02 3.03122371e-01 -8.42520237e-01 -7.67076969e-01 2.68243760e-01 -8.89469311e-02 -3.64537090e-01 -9.43431377e-01 -7.06192851e-01 -1.47118008e+00 2.41117969e-01 -4.75022852e-01 5.52141666e-01 7.37927139e-01 5.87299526e-01 5.62856853e-01 6.41200066e-01 6.43014967e-01 -5.17107904e-01 -1.95976600e-01 -9.82305646e-01 -5.79988301e-01 9.55903053e-01 -1.55144287e-02 -4.26204890e-01 -3.50074053e-01 7.06017315e-01]
[14.800623893737793, -2.351714849472046]
559297fe-caef-45b3-8d04-707055d34f91
enhance-the-motion-cues-for-face-anti
1901.05635
null
http://arxiv.org/abs/1901.05635v1
http://arxiv.org/pdf/1901.05635v1.pdf
Enhance the Motion Cues for Face Anti-Spoofing using CNN-LSTM Architecture
Spatio-temporal information is very important to capture the discriminative cues between genuine and fake faces from video sequences. To explore such a temporal feature, the fine-grained motions (e.g., eye blinking, mouth movements and head swing) across video frames are very critical. In this paper, we propose a joint CNN-LSTM network for face anti-spoofing, focusing on the motion cues across video frames. We first extract the high discriminative features of video frames using the conventional Convolutional Neural Network (CNN). Then we leverage Long Short-Term Memory (LSTM) with the extracted features as inputs to capture the temporal dynamics in videos. To ensure the fine-grained motions more easily to be perceived in the training process, the eulerian motion magnification is used as the preprocessing to enhance the facial expressions exhibited by individuals, and the attention mechanism is embedded in LSTM to ensure the model learn to focus selectively on the dynamic frames across the video clips. Experiments on Replay Attack and MSU-MFSD databases show that the proposed method yields state-of-the-art performance with better generalization ability compared with several other popular algorithms.
['Zheng Ma', 'Xiaoguang Tu', 'Mei Xie', 'Yuefei Zhang', 'Yao Luo', 'Hengsheng Zhang']
2019-01-17
null
null
null
null
['motion-magnification']
['computer-vision']
[-1.27534121e-01 -6.01200581e-01 -2.94701219e-01 -3.01484764e-01 -1.69444203e-01 -1.59089744e-01 4.51021671e-01 -7.29960144e-01 -2.04128906e-01 4.20705944e-01 2.75703549e-01 8.86336565e-02 3.27937417e-02 -3.04811954e-01 -7.22774744e-01 -1.09515679e+00 -3.96343112e-01 -7.01295257e-01 1.57020129e-02 -1.34527817e-01 3.01138759e-01 6.45297289e-01 -1.38560545e+00 6.28961444e-01 1.40884891e-01 1.20145500e+00 -1.00146517e-01 5.19497097e-01 1.06144048e-01 1.12885964e+00 -5.45339406e-01 -1.15182795e-01 -4.66695838e-02 -4.33882773e-01 -4.40871596e-01 2.43043765e-01 5.06654918e-01 -8.95627677e-01 -8.91007483e-01 1.19115722e+00 4.86841232e-01 1.36324227e-01 1.51895449e-01 -1.26384151e+00 -7.61600435e-01 -6.64599836e-02 -8.43415380e-01 7.05675006e-01 4.54962194e-01 5.19650519e-01 2.43284225e-01 -8.71691823e-01 4.18236762e-01 1.66949558e+00 4.25453782e-01 8.88649225e-01 -5.48655093e-01 -1.01866138e+00 2.80709147e-01 7.66168952e-01 -1.27171278e+00 -9.45170283e-01 1.07787156e+00 -2.60764807e-01 4.87690359e-01 1.18685625e-01 5.58855951e-01 1.66903412e+00 4.77723151e-01 9.10550654e-01 5.60963452e-01 9.39667895e-02 -3.72832417e-01 -1.76493302e-01 -1.32244661e-01 7.89433599e-01 -1.44999370e-01 3.73132765e-01 -6.72190130e-01 -3.12520452e-02 9.45476770e-01 4.45493549e-01 -6.04426146e-01 2.74856240e-01 -1.19680429e+00 6.72301650e-01 3.41465205e-01 4.89688873e-01 -2.96806067e-01 2.56596953e-01 7.78026521e-01 3.84753942e-01 4.91427153e-01 -1.66322455e-01 -3.46982419e-01 -6.55178353e-02 -9.48385537e-01 3.36406864e-02 2.43307248e-01 5.97381473e-01 4.73123461e-01 4.77552801e-01 -2.81063408e-01 5.10516107e-01 4.56217527e-01 5.52414238e-01 8.37100565e-01 -9.96989131e-01 5.04524767e-01 2.26357222e-01 7.95688108e-02 -1.83680236e+00 5.75253181e-02 -5.15707545e-02 -9.77048576e-01 -1.01522885e-01 1.77039608e-01 -1.73640043e-01 -7.57274330e-01 1.73746467e+00 3.87601227e-01 9.02133226e-01 -7.46360719e-02 1.06922483e+00 8.63536000e-01 8.58857214e-01 2.22740937e-02 -8.56370568e-01 1.30313671e+00 -8.03464234e-01 -1.28830433e+00 -1.58599839e-01 4.62339640e-01 -7.00979829e-01 6.02465212e-01 2.78933257e-01 -6.60702825e-01 -8.49618435e-01 -9.62960958e-01 4.96292301e-02 -2.18599945e-01 1.17324650e-01 2.95154780e-01 4.98991758e-01 -6.97286844e-01 7.51179814e-01 -8.56038213e-01 -6.53096363e-02 6.11879110e-01 4.25564796e-01 -5.56893587e-01 2.17863470e-02 -1.42400777e+00 4.39297646e-01 1.92675471e-01 7.91608930e-01 -9.94947016e-01 -2.25072965e-01 -1.01177943e+00 -2.10669860e-01 1.73264101e-01 -1.90054607e-02 8.80090594e-01 -1.66523743e+00 -1.63554966e+00 5.86294770e-01 -3.40527862e-01 -1.59215048e-01 4.30879056e-01 -2.17377543e-01 -7.64125645e-01 5.69073021e-01 -2.03294322e-01 6.22550964e-01 1.52844107e+00 -8.21014106e-01 -3.68370503e-01 -4.16139632e-01 -1.43887371e-01 -7.20632598e-02 -7.07674325e-01 5.04787385e-01 -3.25922132e-01 -9.80625331e-01 -1.20182790e-01 -8.20769846e-01 4.66415256e-01 2.31925726e-01 -1.88147977e-01 -1.29406556e-01 1.78608692e+00 -1.04217947e+00 1.32728255e+00 -2.45837498e+00 1.18867390e-01 -2.88963556e-01 1.28020927e-01 6.96277797e-01 -1.79264575e-01 2.52532680e-02 -3.28004092e-01 -4.77549359e-02 1.62409514e-01 -3.48391742e-01 -4.19124305e-01 1.27705976e-01 -4.21108127e-01 1.04641628e+00 3.02478880e-01 7.87747800e-01 -8.15973043e-01 -5.94380677e-01 2.55069464e-01 8.06786895e-01 -2.97794431e-01 3.18507940e-01 7.35465065e-02 8.69214773e-01 -6.90631628e-01 6.68449581e-01 8.89607966e-01 1.08885780e-01 -2.34727085e-01 -3.34353715e-01 1.65677190e-01 -1.18041426e-01 -8.39597523e-01 1.50426948e+00 -8.01086724e-02 9.93218780e-01 2.31308624e-01 -9.71319139e-01 6.01096034e-01 6.34049118e-01 4.88110751e-01 -6.61269844e-01 4.93385583e-01 -6.58109263e-02 6.16968796e-03 -1.27521157e+00 5.56830876e-02 9.99319926e-02 2.60772437e-01 3.23275745e-01 5.47731221e-02 7.10871577e-01 -3.98738384e-01 -2.91634649e-01 5.93835711e-01 1.50309801e-01 -6.83002472e-02 -6.11133426e-02 8.36553514e-01 -6.52899265e-01 7.88834870e-01 7.74112791e-02 -7.50217319e-01 1.96506202e-01 2.83553600e-01 -8.37914407e-01 -6.11737311e-01 -3.06974828e-01 2.10693613e-01 1.10682762e+00 2.49791771e-01 -8.64760801e-02 -8.73510242e-01 -7.18907654e-01 -1.89530522e-01 1.62126184e-01 -8.89623106e-01 -7.44122505e-01 -9.23857629e-01 -4.72685486e-01 6.87311769e-01 3.04376125e-01 9.73805308e-01 -1.40632868e+00 -5.50341368e-01 8.72559473e-02 -2.54371792e-01 -1.26069355e+00 -8.23253453e-01 -6.50367022e-01 -7.67189801e-01 -1.02284789e+00 -6.31223798e-01 -9.57272768e-01 3.54240656e-01 5.84451377e-01 3.56463164e-01 2.78109521e-01 -1.12742238e-01 -1.35476157e-01 -2.52712756e-01 2.53965408e-01 1.17005281e-01 -6.14918232e-01 2.72636354e-01 7.09789574e-01 5.66102624e-01 -4.39167649e-01 -7.26454914e-01 3.48053724e-01 -9.62011039e-01 -2.02041402e-01 2.61858582e-01 9.94925141e-01 2.07946431e-02 1.38270617e-01 3.71585727e-01 -3.66975158e-01 2.29130596e-01 -6.65275812e-01 -6.52618632e-02 -5.72527759e-02 1.98492393e-01 -3.30015481e-01 6.93651497e-01 -9.58152175e-01 -9.91409600e-01 -2.27643937e-01 3.02118398e-02 -1.15782070e+00 -2.15979621e-01 1.99000776e-01 -4.38233256e-01 -3.24303836e-01 4.02761959e-02 5.88872015e-01 1.57826729e-02 -3.07964653e-01 -4.07999158e-02 7.84477770e-01 3.83712471e-01 -2.37198710e-01 6.00544691e-01 7.58824468e-01 -1.00005098e-01 -1.07935798e+00 -6.50670350e-01 -1.56835198e-01 -5.03161788e-01 -4.11675602e-01 1.12972248e+00 -8.14789832e-01 -1.18879342e+00 9.52107310e-01 -1.54374528e+00 -2.71695144e-02 5.61710656e-01 5.56092739e-01 -3.51681918e-01 6.49997532e-01 -1.09025359e+00 -7.87378073e-01 -2.15204492e-01 -1.29755187e+00 1.03344166e+00 3.97511750e-01 1.34281307e-01 -1.02970588e+00 -2.81545550e-01 1.95768774e-01 1.49025127e-01 4.77874219e-01 5.35280466e-01 -4.89189446e-01 -4.73767549e-01 -2.22281232e-01 -2.34319463e-01 3.58928353e-01 3.36947739e-01 3.40673417e-01 -1.09994459e+00 -5.61975777e-01 6.30528748e-01 -2.36698776e-01 8.65011692e-01 4.71501201e-01 1.51183128e+00 -8.93651307e-01 -3.51024747e-01 9.87923324e-01 9.79974866e-01 4.41648006e-01 7.78399050e-01 1.53867275e-01 1.03307092e+00 8.23066831e-01 7.08606958e-01 3.11302215e-01 2.42349785e-02 5.98462105e-01 5.32627344e-01 8.62952918e-02 2.06870779e-01 -1.57170728e-01 8.17705691e-01 7.64930844e-01 -1.30500831e-02 -1.30522028e-01 -3.47042084e-01 3.14988852e-01 -1.83499897e+00 -1.41038191e+00 1.35891169e-01 1.89766574e+00 4.54789340e-01 -7.11697787e-02 -8.24097618e-02 1.12972423e-01 1.11943209e+00 7.13216305e-01 -5.40686250e-01 -1.99319094e-01 -3.90832238e-02 -1.34486407e-01 1.64361402e-01 3.25005561e-01 -1.24755967e+00 9.61213589e-01 5.32311440e+00 1.11816728e+00 -1.75103676e+00 2.01539576e-01 8.50088358e-01 -2.78671592e-01 3.08818370e-01 -3.77416313e-01 -5.97449780e-01 9.44223464e-01 9.43902612e-01 3.78577739e-01 3.83835614e-01 5.69696128e-01 8.23669553e-01 3.28005642e-01 -7.64774621e-01 1.28183484e+00 2.65404671e-01 -1.03324652e+00 1.20992161e-01 6.27164468e-02 4.79516149e-01 -4.90599811e-01 4.21778440e-01 9.16015357e-02 -3.95462751e-01 -1.26809394e+00 5.46008587e-01 6.43160105e-01 8.68974388e-01 -8.49519432e-01 6.79324806e-01 2.62898237e-01 -1.32883942e+00 -2.69984275e-01 -5.27026951e-01 -4.08748463e-02 -1.03994180e-02 1.85645312e-01 -6.40740469e-02 3.10840487e-01 7.60471046e-01 1.34207487e+00 -2.56275505e-01 5.20873129e-01 1.11332824e-02 4.85316426e-01 1.46801785e-01 2.39988565e-01 3.45016688e-01 -6.48622513e-02 6.04802608e-01 1.26001167e+00 2.19672590e-01 2.16142088e-01 1.08121134e-01 3.76956433e-01 -8.95028710e-02 -2.95911252e-01 -7.73662329e-01 -2.27178559e-01 2.98607528e-01 1.09955335e+00 -2.53828019e-01 -2.93782502e-01 -4.49741751e-01 1.21416438e+00 -5.98305352e-02 5.59747338e-01 -1.07452679e+00 -3.62878025e-01 9.58618820e-01 -4.32037450e-02 3.50763738e-01 -2.79763371e-01 3.93678367e-01 -1.29483581e+00 3.43691036e-02 -1.03643990e+00 4.35918599e-01 -7.38758802e-01 -1.08509672e+00 6.24421716e-01 -2.40149498e-01 -1.39957166e+00 -2.17310131e-01 -5.79301357e-01 -8.84530485e-01 7.32142508e-01 -1.62137806e+00 -1.21037424e+00 -4.98195976e-01 1.07993972e+00 8.11180890e-01 -1.45739987e-01 3.94558966e-01 4.90300715e-01 -1.02631819e+00 6.07920766e-01 -2.11286843e-01 5.35376966e-01 8.51485014e-01 -3.59192014e-01 1.75996944e-01 9.69714940e-01 -2.91505307e-01 8.36117506e-01 5.77442348e-01 -6.53270543e-01 -1.50637043e+00 -1.24800098e+00 5.14843881e-01 4.40874463e-03 5.09267032e-01 -1.62239634e-02 -1.02865016e+00 8.28283370e-01 2.85841286e-01 5.57088792e-01 3.79115343e-01 -7.03957498e-01 -4.02075619e-01 -1.83002219e-01 -1.17718005e+00 4.04750109e-01 8.47636700e-01 -9.73596632e-01 -5.04366219e-01 1.55513316e-01 7.00292945e-01 -2.84022003e-01 -4.70148355e-01 3.75751555e-01 8.42772126e-01 -1.03145158e+00 9.78845119e-01 -8.90126348e-01 4.40658182e-01 -2.41828740e-01 -3.92139964e-02 -9.47013199e-01 -2.52286762e-01 -9.61259544e-01 -4.10920233e-01 1.11682701e+00 -4.03059125e-01 -5.68667531e-01 5.96457660e-01 1.21185489e-01 2.18634486e-01 -7.43480682e-01 -9.85793948e-01 -4.96084630e-01 -3.08637321e-01 -1.17362797e-01 4.84758824e-01 1.29058087e+00 -1.53184459e-01 -7.46531337e-02 -1.04457259e+00 3.21807981e-01 4.20565903e-01 -1.11940786e-01 5.53167105e-01 -7.73428500e-01 -7.10382387e-02 -2.12090284e-01 -6.28189087e-01 -1.20923090e+00 5.04282355e-01 -2.20159292e-01 -2.05424801e-01 -6.22823954e-01 7.68729597e-02 4.17491257e-01 -4.69709098e-01 3.72185349e-01 -2.24850640e-01 3.54520202e-01 -6.65100738e-02 3.69723231e-01 -4.46867526e-01 7.75544763e-01 1.59225154e+00 -1.93377033e-01 -3.06058303e-02 -1.56455576e-01 -1.65865973e-01 8.28442693e-01 5.52591205e-01 -3.68682057e-01 -1.44088820e-01 -5.48124909e-01 -4.49934036e-01 5.00446856e-01 7.11933732e-01 -8.36496890e-01 2.54496157e-01 -2.91607410e-01 9.37493622e-01 -4.11451072e-01 7.38259554e-01 -7.42662489e-01 -1.71707153e-01 8.40108931e-01 -1.74719363e-01 2.53577203e-01 3.62862915e-01 7.26356089e-01 -4.39304411e-01 2.19016477e-01 1.00097573e+00 -1.71214610e-01 -7.17072070e-01 9.07119989e-01 -3.46428424e-01 -3.03826213e-01 1.05107522e+00 -4.45484072e-01 -4.05638605e-01 -5.29004216e-01 -6.91462398e-01 -6.07257783e-02 1.60590708e-01 7.11697578e-01 9.05241728e-01 -1.56171739e+00 -4.95327771e-01 5.55163622e-01 -1.46842539e-01 -5.44736505e-01 6.72763884e-01 8.87503028e-01 -4.17893112e-01 4.85337406e-01 -5.76903403e-01 -6.10345483e-01 -1.46598959e+00 7.34235466e-01 5.99199474e-01 3.02831590e-01 -4.88852859e-01 9.78613317e-01 3.31066728e-01 3.06875437e-01 1.54509366e-01 -5.53474054e-02 -5.50137341e-01 2.24537998e-01 9.05257642e-01 2.89773077e-01 -3.73761296e-01 -1.14250600e+00 -5.51084638e-01 6.67178929e-01 -1.26468524e-01 1.34226426e-01 1.17482388e+00 -3.74338776e-01 -2.56939769e-01 2.95798540e-01 1.80432916e+00 -5.95470443e-02 -1.66423988e+00 -1.61630988e-01 -2.30474502e-01 -9.70181406e-01 2.04445869e-01 5.07268682e-03 -1.56191969e+00 1.21388376e+00 7.52840698e-01 2.32494660e-02 1.15878570e+00 -5.20137370e-01 1.21262658e+00 1.72091633e-01 1.02487482e-01 -7.28426397e-01 5.00395715e-01 2.60342091e-01 7.07265019e-01 -1.14615536e+00 -3.30578983e-01 -7.40555152e-02 -3.88448954e-01 1.47397625e+00 6.44019127e-01 -3.75077784e-01 7.14737833e-01 -1.67425603e-01 7.88637102e-02 -2.05887839e-01 -6.88470423e-01 2.28818789e-01 1.71762496e-01 5.05505621e-01 1.59502581e-01 -3.34886521e-01 1.68478847e-01 3.20329368e-01 3.05912614e-01 2.96345670e-02 3.27235013e-01 6.49057150e-01 -3.99135590e-01 -5.50811052e-01 -3.88308108e-01 5.16545735e-02 -8.15412104e-01 1.23364173e-01 -2.28581414e-01 6.54134989e-01 3.05144042e-01 1.07981014e+00 1.56417519e-01 -5.43200612e-01 -2.13081226e-01 -1.95789203e-01 3.88763487e-01 -2.62119740e-01 -3.56681556e-01 1.19755812e-01 -2.93790460e-01 -1.02076519e+00 -5.97486794e-01 -5.14449000e-01 -8.35615516e-01 -6.87134683e-01 -2.51621187e-01 -1.06677942e-01 2.68760443e-01 1.25514710e+00 3.60967398e-01 4.38936502e-01 1.06420040e+00 -1.12695110e+00 -2.97833651e-01 -1.14726675e+00 -4.90143985e-01 4.76359159e-01 1.05540442e+00 -8.41545522e-01 -6.24214053e-01 2.29284942e-01]
[13.209188461303711, 1.336948275566101]
5a2a10ba-0123-4bbd-bef7-898729a6259a
end-to-end-learning-to-index-and-search-in
2210.08410
null
https://arxiv.org/abs/2210.08410v2
https://arxiv.org/pdf/2210.08410v2.pdf
ELIAS: End-to-End Learning to Index and Search in Large Output Spaces
Extreme multi-label classification (XMC) is a popular framework for solving many real-world problems that require accurate prediction from a very large number of potential output choices. A popular approach for dealing with the large label space is to arrange the labels into a shallow tree-based index and then learn an ML model to efficiently search this index via beam search. Existing methods initialize the tree index by clustering the label space into a few mutually exclusive clusters based on pre-defined features and keep it fixed throughout the training procedure. This approach results in a sub-optimal indexing structure over the label space and limits the search performance to the quality of choices made during the initialization of the index. In this paper, we propose a novel method ELIAS which relaxes the tree-based index to a specialized weighted graph-based index which is learned end-to-end with the final task objective. More specifically, ELIAS models the discrete cluster-to-label assignments in the existing tree-based index as soft learnable parameters that are learned jointly with the rest of the ML model. ELIAS achieves state-of-the-art performance on several large-scale extreme classification benchmarks with millions of labels. In particular, ELIAS can be up to 2.5% better at precision@1 and up to 4% better at recall@100 than existing XMC methods. A PyTorch implementation of ELIAS along with other resources is available at https://github.com/nilesh2797/ELIAS.
['Inderjit S Dhillon', 'Cho-Jui Hsieh', 'Hsiang-Fu Yu', 'Patrick H. Chen', 'Nilesh Gupta']
2022-10-16
null
null
null
null
['extreme-multi-label-classification']
['methodology']
[ 4.57877181e-02 -5.62187731e-02 -5.71037412e-01 -5.95583737e-01 -1.09544432e+00 -6.01272464e-01 1.65750176e-01 3.58967215e-01 -4.17769790e-01 4.63965982e-01 -2.24525705e-01 -1.88608095e-01 -3.45208704e-01 -6.59389913e-01 -5.26515543e-01 -8.73175085e-01 1.55775622e-01 1.17538774e+00 6.28742352e-02 2.40668312e-01 2.76917160e-01 1.85215518e-01 -1.67663777e+00 4.55460787e-01 8.42308342e-01 1.30641079e+00 1.94526196e-01 4.99542594e-01 -2.71563232e-01 7.11128712e-01 -3.19357425e-01 -3.65926951e-01 2.35714450e-01 -2.12515205e-01 -8.96236658e-01 -5.58487773e-02 3.96516621e-01 3.50379527e-01 1.90894932e-01 1.05171072e+00 4.20066625e-01 5.33864908e-02 7.64647961e-01 -1.59256339e+00 -1.79100066e-01 8.50879252e-01 -8.69186997e-01 -1.38982251e-01 -5.25288023e-02 8.35488290e-02 1.41190994e+00 -7.81799078e-01 4.95219827e-01 1.08543622e+00 6.11925364e-01 3.59601498e-01 -1.49239635e+00 -9.22681749e-01 2.13364586e-01 2.77671874e-01 -1.71012306e+00 -2.21566092e-02 6.28820240e-01 -4.28366274e-01 8.18127096e-01 4.17620778e-01 2.87684023e-01 5.37483215e-01 -6.40059561e-02 7.15917170e-01 1.15775323e+00 -5.54463685e-01 3.72050226e-01 4.95815426e-02 5.78886449e-01 1.09004021e+00 -6.92136437e-02 -2.25414276e-01 -4.48771924e-01 -5.78526139e-01 1.01621598e-01 5.11574559e-02 -8.79667401e-02 -4.22737956e-01 -9.15560246e-01 1.05899823e+00 5.04928470e-01 1.83473766e-01 -9.00098383e-02 2.77664393e-01 4.65848863e-01 3.52124125e-01 5.94606400e-01 3.79081666e-01 -6.25702441e-01 8.05837065e-02 -1.06708443e+00 7.17056468e-02 8.15044045e-01 7.70589173e-01 9.51601684e-01 -6.83240354e-01 -4.71238524e-01 1.14062047e+00 4.23039824e-01 1.62010595e-01 4.16781247e-01 -9.80661988e-01 5.15441000e-01 7.72788107e-01 -2.03540534e-01 -7.25916862e-01 -7.21286058e-01 -9.36457276e-01 -8.22745740e-01 1.38324022e-01 4.61643964e-01 5.45743853e-02 -9.64082778e-01 1.86647964e+00 6.75162673e-01 2.07264170e-01 -3.71411920e-01 6.31684721e-01 4.96921778e-01 6.81590676e-01 2.72884220e-01 -2.23358631e-01 1.34748697e+00 -1.42547202e+00 -3.10068101e-01 -1.93073839e-01 1.02249873e+00 -5.29274285e-01 1.22445583e+00 5.37639856e-01 -8.23539674e-01 -3.14271748e-01 -7.93907583e-01 1.62921511e-02 -3.17771405e-01 1.49960577e-01 5.20133615e-01 4.54909474e-01 -9.73510146e-01 4.93375957e-01 -5.76314270e-01 1.28664374e-01 3.67231280e-01 5.43731213e-01 1.05088606e-01 -3.05989355e-01 -1.04672861e+00 6.10568881e-01 7.23668575e-01 -3.41738552e-01 -4.36519712e-01 -7.10571706e-01 -4.93274242e-01 2.21084595e-01 7.18318582e-01 -5.11218011e-01 1.21261299e+00 -5.69802225e-01 -1.05970633e+00 1.16332304e+00 -6.91460297e-02 -1.18239239e-01 4.55644488e-01 1.71996117e-01 -4.59148064e-02 3.20912222e-03 1.94609076e-01 7.54983962e-01 6.24978662e-01 -1.27592623e+00 -9.42453384e-01 -4.05465811e-01 -1.76295683e-01 2.94213831e-01 -4.86117780e-01 -4.74467762e-02 -9.15625215e-01 -5.40777624e-01 1.70681477e-01 -1.12278712e+00 -2.40265638e-01 -9.24114287e-02 -4.75206435e-01 -8.79348040e-01 5.39961457e-01 -1.73449680e-01 1.52925706e+00 -1.98609948e+00 3.11686814e-01 5.01110494e-01 4.58897144e-01 3.83687317e-02 -2.17003465e-01 2.73394287e-01 -4.50505828e-03 1.48599446e-01 -1.86933041e-01 -7.78430521e-01 3.03447008e-01 4.91024554e-02 6.26678243e-02 5.19393802e-01 -4.35122192e-01 8.31604302e-01 -8.88789117e-01 -7.57919729e-01 6.62730336e-02 1.69562593e-01 -5.73566675e-01 1.44377083e-01 -6.08733237e-01 2.14470655e-01 -3.46046835e-01 7.04756677e-01 5.44177175e-01 -8.52151453e-01 2.12678790e-01 6.47650510e-02 2.79211491e-01 -9.48735252e-02 -1.28948987e+00 1.53504586e+00 -5.28513551e-01 -8.74275118e-02 -1.22934617e-01 -9.81292605e-01 7.50632226e-01 2.99761239e-02 7.37568438e-01 -4.03328091e-01 2.32074171e-01 2.95716077e-01 -4.26149845e-01 -1.47501022e-01 6.69092759e-02 -1.58724129e-01 -3.30284268e-01 7.63126016e-01 -1.66180544e-02 1.14342988e-01 3.50483149e-01 9.20916125e-02 1.00269234e+00 -2.18346700e-01 1.87939689e-01 -2.10719317e-01 4.92006093e-01 -5.98846041e-02 7.52229989e-01 7.46467412e-01 1.69270203e-01 5.83944380e-01 4.91000175e-01 -4.76179928e-01 -6.86170876e-01 -7.74854481e-01 -2.77117282e-01 1.63193405e+00 -2.60097273e-02 -7.18969464e-01 -8.09996486e-01 -9.71705139e-01 2.72264391e-01 9.80683744e-01 -6.12086177e-01 -2.82824367e-01 -3.34539592e-01 -8.15268159e-01 3.05963874e-01 2.12102622e-01 4.37915415e-01 -1.04649115e+00 -3.14324051e-01 5.78710623e-02 -2.88235188e-01 -7.99552321e-01 -7.18814909e-01 6.32053256e-01 -6.02937162e-01 -1.05682075e+00 -4.32525635e-01 -8.45911622e-01 7.61218846e-01 -1.24799162e-01 1.37530279e+00 4.01769280e-01 -2.94419467e-01 -4.89967689e-02 -3.18823040e-01 -2.16968395e-02 -2.57302225e-01 6.07973397e-01 -1.89689234e-01 9.81186032e-02 4.12027359e-01 -3.16730082e-01 -4.60745782e-01 4.45318729e-01 -7.77995944e-01 2.07994506e-01 2.79002279e-01 8.20954561e-01 9.82611716e-01 2.95420676e-01 5.62138617e-01 -1.31550062e+00 3.40494901e-01 -7.98183441e-01 -6.99394524e-01 5.18893838e-01 -1.11735392e+00 2.50564814e-01 7.24174738e-01 -5.61567605e-01 -5.31574488e-01 3.72734725e-01 -1.44509345e-01 -5.60364246e-01 3.48857455e-02 5.26584506e-01 -4.07259986e-02 5.47834449e-02 4.49392259e-01 -1.54678687e-01 -2.60385334e-01 -5.93163073e-01 3.64204615e-01 8.46610129e-01 2.11254343e-01 -7.81811893e-01 5.01483977e-01 -1.02360090e-02 9.51910540e-02 1.46146994e-02 -1.51650035e+00 -7.62715042e-01 -6.16406620e-01 -2.41429240e-01 5.90945780e-01 -6.65506124e-01 -7.63808906e-01 6.15913272e-01 -4.18444753e-01 -7.98653483e-01 -1.38565928e-01 6.13803379e-02 -5.67425489e-01 8.66082907e-02 -7.86153793e-01 -5.34479976e-01 -6.33042097e-01 -1.16463387e+00 1.27739847e+00 6.35007396e-02 -1.32875592e-01 -1.07487726e+00 -3.15970071e-02 6.36381984e-01 1.46241754e-01 1.47948086e-01 1.33652270e+00 -7.25227535e-01 -4.31121320e-01 -2.60131329e-01 -1.35892302e-01 1.13750435e-03 -2.11338833e-01 -2.60050088e-01 -7.79859543e-01 -6.43304229e-01 -2.57356673e-01 -7.16787577e-01 1.15438652e+00 1.87602773e-01 1.64959526e+00 -1.56715155e-01 -7.32267201e-01 8.32318306e-01 1.67957914e+00 -4.98786531e-02 1.53000340e-01 3.77271861e-01 7.01706290e-01 3.14142495e-01 8.21643412e-01 5.12946963e-01 5.15356123e-01 8.31845164e-01 5.10951698e-01 -6.68513775e-02 5.63239679e-02 -1.10755533e-01 -1.07433848e-01 6.43128455e-01 5.20782590e-01 -5.19516349e-01 -1.08221078e+00 1.57632336e-01 -1.96564794e+00 -4.96905446e-01 4.54813838e-02 2.31659102e+00 1.24615562e+00 2.68849850e-01 1.50503874e-01 1.77515298e-01 8.11718345e-01 3.42053361e-02 -8.20051134e-01 -1.82360396e-01 1.88846797e-01 -2.35350132e-02 5.08835137e-01 4.77708101e-01 -1.18879128e+00 9.78600621e-01 5.13569641e+00 1.27477062e+00 -1.09620667e+00 2.26841986e-01 1.12171018e+00 -4.46506053e-01 -7.90991187e-02 -9.18138325e-02 -1.20596564e+00 7.30648518e-01 1.02896571e+00 1.41861185e-01 5.28488934e-01 8.83928180e-01 -3.30588907e-01 -1.21508300e-01 -1.26185942e+00 1.06368721e+00 -4.96911705e-02 -1.04950917e+00 -2.92386532e-01 2.12418839e-01 8.24079275e-01 1.68200016e-01 2.98546880e-01 6.08684480e-01 6.08130693e-01 -9.12443280e-01 6.95303380e-01 3.66760314e-01 1.17759800e+00 -9.28941786e-01 4.61632460e-01 8.98694694e-01 -1.14586532e+00 -5.33230901e-01 -3.61137599e-01 3.25706303e-01 -1.31429225e-01 8.62734318e-01 -5.53998351e-01 9.24725235e-02 6.22221947e-01 3.73764127e-01 -7.75929987e-01 1.05914998e+00 -6.72678202e-02 8.56943130e-01 -6.05900526e-01 -3.62768844e-02 3.72836292e-01 -1.78471878e-01 6.41883686e-02 1.15788221e+00 2.27579743e-01 -3.02268472e-02 8.17455649e-01 5.70219040e-01 -3.78360242e-01 3.41789037e-01 -2.11821981e-02 3.66635829e-01 7.02449381e-01 1.52476025e+00 -1.03552616e+00 -3.99499953e-01 -3.41359079e-01 7.44997919e-01 1.00988257e+00 8.31903219e-02 -1.03941107e+00 -1.79989472e-01 2.34951288e-01 -8.58802255e-03 2.23383814e-01 1.99448109e-01 -2.82029748e-01 -9.57782209e-01 -3.91791798e-02 -8.54802310e-01 9.88315344e-01 -3.61220658e-01 -1.31321347e+00 5.36482096e-01 6.48996904e-02 -9.66583431e-01 -2.47969985e-01 -3.93779278e-01 -3.38264316e-01 6.84396327e-01 -1.45392632e+00 -1.05155730e+00 -3.62321258e-01 3.28233063e-01 5.74557841e-01 2.57781073e-02 8.97587240e-01 2.71604896e-01 -8.53846788e-01 1.03636432e+00 4.87866104e-01 -1.10314973e-01 8.00159514e-01 -1.52656043e+00 7.20782951e-02 1.52305290e-01 2.42984056e-01 1.91701829e-01 3.79950404e-01 -5.19983649e-01 -8.26437891e-01 -1.31550121e+00 1.04771566e+00 -2.32949838e-01 4.30972099e-01 -4.48879391e-01 -8.64897370e-01 5.88116705e-01 -1.61323577e-01 3.08061957e-01 8.34465206e-01 3.44894290e-01 -4.47871566e-01 -2.56306678e-01 -1.20248556e+00 3.14961821e-01 9.89706039e-01 -2.00824350e-01 -1.23836122e-01 7.82362163e-01 4.93607968e-01 -4.39177424e-01 -1.03900421e+00 4.07870144e-01 3.11330736e-01 -6.39923871e-01 8.33832979e-01 -4.20238554e-01 2.35478297e-01 -9.96364802e-02 -9.68524665e-02 -1.31070089e+00 -6.01235330e-01 -2.94337988e-01 -3.18271339e-01 1.26715362e+00 6.64562583e-01 -6.14854038e-01 9.31553245e-01 6.11090660e-01 1.78439125e-01 -1.44513714e+00 -7.45870411e-01 -5.58008790e-01 3.02807391e-01 -2.00301573e-01 5.94641328e-01 9.98319030e-01 -4.25634868e-02 5.77930510e-01 -7.97493309e-02 -1.35782491e-02 1.01260400e+00 7.23295212e-01 3.58764708e-01 -1.56691873e+00 -4.63901371e-01 -4.88880455e-01 3.81107442e-02 -8.64702523e-01 4.99879062e-01 -1.55935061e+00 3.65711711e-02 -1.44443798e+00 6.16335928e-01 -1.20281434e+00 -7.28397608e-01 9.07012165e-01 -4.13107127e-01 3.40644002e-01 2.70207494e-01 4.23009098e-01 -1.16187358e+00 2.81590611e-01 8.98525953e-01 -1.58797726e-01 -1.08763412e-01 2.32194826e-01 -5.77348590e-01 6.22197926e-01 1.13766992e+00 -9.27118659e-01 -3.58025402e-01 -2.14130282e-01 3.00222665e-01 5.22536226e-02 -1.94844216e-01 -8.99283051e-01 4.63454962e-01 -2.26059660e-01 2.34608889e-01 -5.70678890e-01 2.26769760e-01 -7.28049219e-01 1.47539586e-01 3.60370547e-01 -8.85339916e-01 -1.41763389e-01 -3.04701030e-01 5.45213461e-01 1.24494314e-01 -6.22222304e-01 9.58324969e-01 -1.70694232e-01 -3.69682372e-01 4.58726615e-01 2.49419082e-02 3.95837396e-01 1.22139311e+00 1.61677346e-01 -2.24693418e-01 -9.98219550e-02 -7.90070295e-01 8.75324965e-01 5.84039807e-01 2.31310517e-01 1.78299278e-01 -1.34561133e+00 -6.41561687e-01 3.55632268e-02 2.33373910e-01 1.22988947e-01 -3.27405185e-02 6.41925156e-01 -1.74265847e-01 3.00349176e-01 3.03483695e-01 -6.19893432e-01 -1.60348928e+00 8.88422370e-01 3.52051377e-01 -8.41586649e-01 -5.48926890e-01 9.79252994e-01 -1.31866373e-02 -6.37510717e-01 7.20392108e-01 4.77364939e-03 -2.09685490e-01 1.05389729e-01 3.23149502e-01 4.21074301e-01 -4.17282768e-02 -6.32107317e-01 -1.89926475e-01 7.89559007e-01 -1.57691583e-01 2.44260788e-01 1.31778789e+00 -9.27504227e-02 -2.77095854e-01 5.98690629e-01 1.52757251e+00 -3.61098886e-01 -1.19919264e+00 -7.38745630e-01 3.88406396e-01 -2.49199659e-01 2.65786052e-01 -1.06424737e+00 -1.33247864e+00 5.28187811e-01 5.15019774e-01 8.57451633e-02 1.18383074e+00 3.12989920e-01 8.18960190e-01 4.30377722e-01 4.48404282e-01 -1.26457059e+00 1.69569194e-01 4.19255644e-01 5.38885057e-01 -1.20638490e+00 6.03513606e-02 -5.57250857e-01 -5.10572672e-01 8.33137333e-01 6.13308489e-01 1.90242156e-01 7.72467494e-01 3.91464293e-01 1.98928472e-02 -3.53875190e-01 -1.13412619e+00 -1.15068160e-01 2.03376755e-01 -5.92632666e-02 2.06231639e-01 1.77582130e-01 -3.34159940e-01 4.69840407e-01 -4.60457280e-02 -1.34414047e-01 -2.89060652e-01 7.08261549e-01 -5.99815488e-01 -1.25388503e+00 -2.77829766e-01 8.65879476e-01 -6.27466559e-01 -6.09145649e-02 -1.16942614e-01 2.24490225e-01 3.23139697e-01 1.00436723e+00 2.82709748e-02 -3.99002880e-01 -1.68628544e-02 4.21319306e-01 2.17579871e-01 -6.79350078e-01 -6.93552673e-01 4.21064347e-02 -8.91469568e-02 -4.87549424e-01 5.54680377e-02 -6.61927044e-01 -1.52568352e+00 -1.50503948e-01 -5.83816707e-01 3.62743586e-01 5.15616059e-01 6.33174658e-01 3.54961932e-01 3.04939061e-01 9.29093421e-01 -6.04774773e-01 -9.56992388e-01 -8.82117212e-01 -5.93494833e-01 4.74203438e-01 -4.81902175e-02 -6.88665986e-01 -4.67029631e-01 -2.42752820e-01]
[9.498454093933105, 4.357867240905762]
068a0ae8-40ed-470b-b6ed-ff7a7abba3f0
on-counterfactual-inference-with-unobserved
2211.08209
null
https://arxiv.org/abs/2211.08209v2
https://arxiv.org/pdf/2211.08209v2.pdf
On counterfactual inference with unobserved confounding
Given an observational study with $n$ independent but heterogeneous units, our goal is to learn the counterfactual distribution for each unit using only one $p$-dimensional sample per unit containing covariates, interventions, and outcomes. Specifically, we allow for unobserved confounding that introduces statistical biases between interventions and outcomes as well as exacerbates the heterogeneity across units. Modeling the underlying joint distribution as an exponential family, we reduce learning the unit-level counterfactual distributions to learning $n$ exponential family distributions with heterogeneous parameters and only one sample per distribution. We introduce a convex objective that pools all $n$ samples to jointly learn all $n$ parameter vectors, and provide a unit-wise mean squared error bound that scales linearly with the metric entropy of the parameter space. For example, when the parameters are $s$-sparse linear combination of $k$ known vectors, the error is $O(s\log k/p)$. En route, we derive sufficient conditions for compactly supported distributions to satisfy the logarithmic Sobolev inequality. As an application of the framework, our results enable consistent imputation of sparsely missing covariates.
['Gregory W. Wornell', 'Devavrat Shah', 'Raaz Dwivedi', 'Abhin Shah']
2022-11-14
null
null
null
null
['counterfactual-inference']
['miscellaneous']
[ 4.39730957e-02 3.20263952e-01 -8.30069721e-01 -5.51816642e-01 -1.15913486e+00 -3.30512077e-01 -1.97667748e-01 1.34248123e-01 -5.45332015e-01 1.21180785e+00 4.49389189e-01 -2.15245932e-01 -3.21017891e-01 -8.89862001e-01 -1.30303979e+00 -8.58180940e-01 -6.44054472e-01 2.37947062e-01 -8.73372614e-01 5.45045435e-01 1.30333090e-02 2.06021555e-02 -1.13480437e+00 -2.95915484e-01 1.18975008e+00 8.60276937e-01 -3.45978051e-01 3.67110550e-01 2.44236335e-01 6.73931420e-01 -2.04003245e-01 -3.89245570e-01 4.90932792e-01 -3.07449043e-01 -1.60467029e-01 -6.91550598e-02 4.31275070e-01 -3.18280905e-01 -3.23321134e-01 1.13511705e+00 4.80152339e-01 1.30629063e-01 1.23434138e+00 -1.23922372e+00 -8.03488791e-01 9.76286530e-01 -8.33416641e-01 -2.99599916e-01 8.21476895e-03 1.40544191e-01 6.79670453e-01 -7.28754580e-01 3.59785944e-01 9.28537607e-01 7.90710986e-01 3.88831407e-01 -1.64261115e+00 -1.21178973e+00 1.31902933e-01 -2.74971217e-01 -1.39703333e+00 -5.40093303e-01 7.53125191e-01 -8.19747150e-01 4.52126712e-01 -1.57245994e-02 4.10735011e-01 1.07392228e+00 4.09478068e-01 4.18630689e-01 7.91852415e-01 -1.72554016e-01 6.94115937e-01 2.01280624e-01 -3.92934214e-03 6.07029736e-01 8.90515268e-01 2.23023847e-01 -3.84987324e-01 -4.67157185e-01 8.23351979e-01 3.53026509e-01 -2.95361519e-01 -7.45886505e-01 -1.05020893e+00 1.14593697e+00 1.32896364e-01 -6.87490106e-02 -6.61963820e-01 4.53693300e-01 2.22742915e-01 6.65505230e-02 4.22341079e-01 2.91317940e-01 -5.15295744e-01 1.04142249e-01 -8.34139407e-01 3.46049845e-01 6.35746360e-01 1.00794530e+00 6.93675637e-01 1.83527887e-01 -2.05910340e-01 4.46620584e-01 -1.96317211e-02 1.22936058e+00 1.19860724e-01 -1.46875548e+00 8.74996603e-01 3.43628258e-01 8.15141141e-01 -9.07709718e-01 -2.18739673e-01 -1.72524005e-01 -1.36452425e+00 -6.94695339e-02 6.94546163e-01 -7.81462252e-01 -4.53642309e-01 2.61918855e+00 2.74089605e-01 2.85757959e-01 -2.04144910e-01 5.24406195e-01 4.96555567e-02 3.60605508e-01 2.94833183e-01 -7.24423528e-01 1.00736904e+00 -1.79930821e-01 -7.14321613e-01 -1.10099837e-01 7.56647348e-01 -2.66368449e-01 1.07019520e+00 5.62589839e-02 -1.57762897e+00 1.30581539e-04 -7.03087926e-01 3.02803129e-01 3.03839725e-02 -7.12583438e-02 5.77929616e-01 7.02495754e-01 -7.38882601e-01 4.33157861e-01 -5.78185618e-01 4.08591986e-01 7.01059043e-01 5.81071973e-01 -4.16728795e-01 -3.37422371e-01 -1.03678143e+00 4.50193703e-01 -1.72282428e-01 -1.59397438e-01 -9.30161476e-01 -1.41313982e+00 -1.05082285e+00 4.99673694e-01 2.74066269e-01 -9.63951826e-01 7.68084407e-01 -5.36649227e-01 -9.35285807e-01 5.20272195e-01 -3.98233086e-01 -3.63009602e-01 5.21240115e-01 4.15090024e-01 -2.09758747e-02 -3.99744213e-02 4.53078538e-01 8.44908804e-02 8.36517096e-01 -9.76228535e-01 -4.50072199e-01 -7.48487830e-01 -3.49935859e-01 1.42721161e-01 -2.98650742e-01 -1.97661728e-01 4.16890115e-01 -7.28216469e-01 -1.76401064e-01 -5.00209928e-01 -4.48697090e-01 9.53544229e-02 -1.53578594e-01 1.00111134e-01 -6.75689057e-02 -7.32172430e-01 1.31312287e+00 -2.34014535e+00 1.67322718e-02 2.03823358e-01 2.64340669e-01 -7.09527314e-01 -3.55020463e-02 -4.88313101e-02 -1.44312069e-01 1.60881802e-01 -7.08828211e-01 -4.36233193e-01 2.39236161e-01 -1.97169647e-01 -2.33085100e-02 9.56306040e-01 -1.84830010e-01 6.52849138e-01 -8.15246820e-01 -2.53291041e-01 -6.82066381e-02 3.30495566e-01 -7.86457717e-01 6.70547262e-02 6.62935749e-02 9.50514972e-02 -3.55661482e-01 4.35915381e-01 1.05348933e+00 -3.60270143e-01 1.05810463e-01 2.84901977e-01 1.34596378e-01 -1.40046060e-01 -1.52340257e+00 1.45012450e+00 -5.15686691e-01 1.22351617e-01 4.95735794e-01 -1.59412980e+00 5.08812666e-01 4.80397910e-01 8.67380917e-01 -2.31444910e-01 1.77349776e-01 4.73216087e-01 -3.94462228e-01 -4.64251935e-01 -1.22065932e-01 -9.38751638e-01 -4.88007396e-01 5.35645247e-01 1.51189966e-02 1.82306886e-01 -2.23716050e-01 -3.78281772e-02 1.24540102e+00 -5.63821852e-01 3.18634838e-01 -4.88435090e-01 3.76584716e-02 -3.26869339e-01 9.72159326e-01 9.36042070e-01 -2.31554151e-01 3.30408067e-01 8.06892812e-01 -3.31228524e-02 -1.14877677e+00 -1.51012611e+00 -3.28394830e-01 5.94617546e-01 -1.56291083e-01 3.92934263e-01 -6.24786615e-01 -5.74584961e-01 4.64412302e-01 7.77367651e-01 -1.20110214e+00 -1.70638174e-01 -4.46984440e-01 -1.08894825e+00 5.42469174e-02 5.90691388e-01 1.93709463e-01 -7.71055281e-01 -2.84376711e-01 1.49037719e-01 -1.74970269e-01 -3.43335718e-01 -1.08634543e+00 2.69195497e-01 -9.55770671e-01 -1.21119070e+00 -1.10201132e+00 -6.99944556e-01 8.35463166e-01 -3.04054581e-02 1.04264605e+00 -9.03938115e-01 -3.01767230e-01 5.62273979e-01 3.79324049e-01 -5.56768000e-01 1.86759621e-01 -4.25912380e-01 3.36581945e-01 -9.82938800e-03 4.05205280e-01 -6.01033390e-01 -1.14516962e+00 -1.64433122e-01 -7.09229827e-01 -6.13735676e-01 3.43792081e-01 8.24228764e-01 5.72188735e-01 3.36516909e-02 9.77279902e-01 -8.54963303e-01 4.41470027e-01 -1.02834725e+00 -7.22849607e-01 2.17963144e-01 -5.13996601e-01 4.17652763e-02 7.42890179e-01 -8.08543861e-01 -8.71974885e-01 -1.31463498e-01 6.26938462e-01 -3.76831323e-01 2.09904671e-01 5.33937335e-01 -4.24822748e-01 3.65874231e-01 7.57715344e-01 2.44165827e-02 7.42783844e-02 -3.14796001e-01 4.80822921e-01 6.59463406e-01 3.84638011e-01 -8.37423027e-01 3.38049024e-01 6.90893829e-01 9.77989882e-02 -5.43422282e-01 -5.95639288e-01 -1.77777857e-01 2.08058637e-02 3.62254053e-01 7.56834269e-01 -1.11715877e+00 -1.14014816e+00 1.19553857e-01 -7.00022519e-01 -6.03608668e-01 -9.46235895e-01 1.05248320e+00 -1.00496244e+00 -9.45307165e-02 -3.23384970e-01 -1.14012980e+00 -1.12289667e-01 -8.69933963e-01 7.30168998e-01 -1.01046450e-01 -7.33403713e-02 -1.02464175e+00 2.63140082e-01 1.05953053e-01 3.30003887e-01 3.83979231e-01 1.21756816e+00 -1.28440082e-01 -3.70256752e-01 -4.77809727e-01 -3.00177991e-01 3.69877934e-01 2.99109310e-01 -5.46567202e-01 -4.50840086e-01 -3.07773381e-01 4.81726289e-01 -1.51035592e-01 5.02631247e-01 1.42184806e+00 1.44230270e+00 -9.85124588e-01 -3.16160053e-01 6.24655128e-01 1.56875074e+00 2.83084810e-01 1.80982068e-01 -3.65709931e-01 3.16445440e-01 6.39043987e-01 4.65126447e-02 8.28320742e-01 6.26895964e-01 9.77664627e-03 1.05919383e-01 8.41972306e-02 6.25729084e-01 -2.61535138e-01 2.71779150e-01 3.57980758e-01 3.47175092e-01 9.38046873e-02 -5.96416771e-01 9.94681418e-01 -1.61137402e+00 -1.23911011e+00 3.47837150e-01 2.89957070e+00 1.00346100e+00 -2.37945899e-01 4.51420069e-01 -2.88058430e-01 7.55648017e-01 -1.85223371e-01 -1.02353966e+00 -2.57551432e-01 -1.48936823e-01 4.93018538e-01 1.03966057e+00 7.58482397e-01 -7.21677005e-01 -1.32584572e-02 6.89750671e+00 7.42050886e-01 -7.07989275e-01 7.40225986e-02 1.04632270e+00 -5.56768477e-01 -7.85409868e-01 -2.93538719e-01 -4.64730352e-01 1.01225185e+00 1.09139931e+00 -5.99234223e-01 4.13292289e-01 6.36103213e-01 5.43875396e-01 -2.11681142e-01 -1.27267766e+00 8.43945324e-01 -1.88955754e-01 -1.30841148e+00 -3.57747912e-01 4.30345625e-01 1.01493025e+00 -2.57689804e-01 4.18847442e-01 2.53708124e-01 7.50139594e-01 -1.30305505e+00 3.72209072e-01 3.81609708e-01 1.28518677e+00 -8.71094286e-01 6.39607370e-01 4.19094622e-01 -9.16490495e-01 -2.92376041e-01 -5.31565249e-01 -1.91084981e-01 1.64650247e-01 8.11354995e-01 -3.49478900e-01 -1.75508950e-02 6.11171544e-01 4.27499384e-01 4.09435302e-01 8.93150806e-01 2.68176019e-01 7.04767346e-01 -5.83333910e-01 1.87617511e-01 -2.50739574e-01 -5.11245608e-01 1.43367022e-01 8.35172355e-01 7.56417692e-01 5.58180511e-01 -1.94184884e-01 8.76325846e-01 -4.94066805e-01 2.82974720e-01 -9.11759973e-01 3.50447074e-02 7.96625018e-01 3.01475197e-01 -5.74796908e-02 -4.46008682e-01 -5.86055160e-01 5.91278493e-01 3.52972806e-01 6.68577373e-01 -7.38892734e-01 -3.03634703e-01 9.57093358e-01 2.03568354e-01 2.56855249e-01 -5.23123657e-03 -9.19235945e-01 -1.39464509e+00 2.94213563e-01 -5.27218044e-01 6.38508677e-01 -3.03002238e-01 -1.64459896e+00 -3.20477605e-01 1.92675926e-02 -8.29046547e-01 -3.01623065e-02 -2.09381491e-01 -4.15690660e-01 1.27558649e+00 -1.03692126e+00 -4.44050163e-01 4.26285386e-01 6.71359420e-01 -7.70881400e-02 7.16661811e-02 8.77151489e-01 2.23014504e-01 -7.48109698e-01 8.47023487e-01 6.66715980e-01 -5.35318814e-02 7.65352607e-01 -1.31333208e+00 -3.81600618e-01 3.82549763e-01 -5.23501217e-01 7.73327053e-01 6.99940503e-01 -7.46090412e-01 -1.16586173e+00 -1.00177097e+00 9.81221676e-01 -2.76580304e-01 7.60106087e-01 -4.03497249e-01 -7.99933672e-01 8.66330087e-01 -2.45728046e-01 1.00218393e-01 1.17238569e+00 2.59601295e-01 -4.70206648e-01 -2.25194946e-01 -1.76194501e+00 6.38205826e-01 9.48196590e-01 -3.71342480e-01 -2.00083807e-01 2.23154753e-01 7.60596871e-01 -7.65308887e-02 -1.17159164e+00 4.85364288e-01 6.21936262e-01 -5.40987372e-01 1.02819872e+00 -1.02512646e+00 5.27864456e-01 1.82896838e-01 -5.82255721e-01 -1.25603390e+00 -1.97079793e-01 -5.00520527e-01 -1.49834082e-01 7.81847715e-01 6.78929329e-01 -7.79429257e-01 1.04399168e+00 1.37767041e+00 4.38191116e-01 -7.05918074e-01 -1.31025219e+00 -6.44619942e-01 8.00472736e-01 -2.88149774e-01 6.46599591e-01 1.19278502e+00 3.21744770e-01 -8.50444287e-02 -5.67603230e-01 1.80899531e-01 1.23449171e+00 5.77043258e-02 3.95286024e-01 -8.04656327e-01 -4.46837723e-01 -2.53910452e-01 2.93018669e-02 -6.52589917e-01 3.63145173e-01 -6.75039113e-01 -2.42075650e-03 -9.37132835e-01 7.77819633e-01 -7.83264101e-01 -3.51199955e-01 -3.89879127e-03 -4.74620849e-01 -2.03018904e-01 -1.01301596e-01 -3.22658300e-01 -2.66593844e-01 7.51378238e-01 9.83173609e-01 -1.61678478e-01 -3.00951689e-01 1.03661448e-01 -1.14579427e+00 5.56186736e-01 6.32088184e-01 -5.70365310e-01 -4.57172722e-01 -3.47699821e-01 2.14939907e-01 7.29683399e-01 1.09000079e-01 -3.59164506e-01 -5.40306047e-02 -6.98546290e-01 2.88455755e-01 -1.05643325e-01 1.71185791e-01 -9.00340796e-01 2.07001925e-01 4.24127877e-01 -5.71934223e-01 -2.82369941e-01 -1.32128969e-03 6.63437784e-01 2.04058722e-01 -6.13804869e-02 8.51994336e-01 -1.23287976e-01 5.60222983e-01 5.95861197e-01 -2.47061446e-01 5.43523490e-01 9.72400308e-01 1.12615630e-01 -1.64068162e-01 -5.39206028e-01 -8.00027013e-01 5.27024329e-01 5.18343747e-01 -3.38349283e-01 3.05331826e-01 -1.46629596e+00 -7.73145258e-01 1.69323832e-01 -2.52248585e-01 3.03001612e-01 8.63349080e-01 6.23639762e-01 1.43794268e-01 3.05866957e-01 9.49526578e-02 -1.64376840e-01 -5.00179946e-01 9.78806198e-01 2.98854202e-01 -8.74887928e-02 -2.06527933e-01 8.44834447e-01 5.46733141e-01 -4.38901156e-01 3.21097940e-01 -2.94478595e-01 4.29294229e-01 6.80438289e-03 6.25993729e-01 7.44916499e-01 -4.72247124e-01 -3.45018320e-02 -1.63198307e-01 4.21573550e-01 2.69030482e-01 -3.08539242e-01 1.48073900e+00 -2.91074783e-01 -6.56243712e-02 5.59938073e-01 1.58943415e+00 1.05800211e-01 -1.60974705e+00 -3.09349269e-01 -3.77275020e-01 -4.67804283e-01 -4.26276714e-01 -5.40763080e-01 -8.30524802e-01 6.34008169e-01 5.29261827e-01 -3.56496610e-02 1.02941394e+00 5.98955229e-02 2.97411621e-01 -2.84385055e-01 4.35313731e-01 -1.00923431e+00 -3.11698884e-01 -1.59993604e-01 7.37217009e-01 -1.44754887e+00 -1.33958057e-01 -2.20542371e-01 -2.80194491e-01 2.82663703e-01 2.71300435e-01 -4.56047058e-01 1.14108527e+00 3.50217253e-01 -3.87694836e-01 2.35030614e-02 -4.70588446e-01 4.52082962e-01 5.34372125e-03 6.78491294e-01 5.34871399e-01 6.14992976e-01 -5.55342555e-01 1.05089808e+00 -1.53453156e-01 4.23956186e-01 5.68408966e-01 5.27462423e-01 -2.48250514e-01 -7.66499996e-01 -4.71297443e-01 1.02056396e+00 -6.62239134e-01 -1.20441228e-01 5.28627932e-01 5.61319530e-01 5.75394444e-02 7.09199071e-01 3.44434947e-01 2.97296584e-01 4.03880626e-01 1.38945073e-01 4.56909865e-01 -5.53396523e-01 6.95956200e-02 -2.67141014e-02 -5.33084035e-01 -5.61478555e-01 -2.64252633e-01 -1.02986825e+00 -9.20990050e-01 -7.43401825e-01 1.33152619e-01 3.12242270e-01 4.38053101e-01 6.88894093e-01 2.38670409e-01 8.76584798e-02 1.04475713e+00 -4.98272687e-01 -1.05521393e+00 -7.96554089e-01 -1.14404404e+00 2.89209247e-01 6.41920090e-01 -5.58049202e-01 -9.86975193e-01 -1.34241685e-01]
[7.95032262802124, 5.169031143188477]
73c5ac56-73df-4d01-ac5c-509b9537b2ea
improving-video-instance-segmentation-by
2012.07504
null
https://arxiv.org/abs/2012.07504v2
https://arxiv.org/pdf/2012.07504v2.pdf
Improving Video Instance Segmentation by Light-weight Temporal Uncertainty Estimates
Instance segmentation with neural networks is an essential task in environment perception. In many works, it has been observed that neural networks can predict false positive instances with high confidence values and true positives with low ones. Thus, it is important to accurately model the uncertainties of neural networks in order to prevent safety issues and foster interpretability. In applications such as automated driving, the reliability of neural networks is of highest interest. In this paper, we present a time-dynamic approach to model uncertainties of instance segmentation networks and apply this to the detection of false positives as well as the estimation of prediction quality. The availability of image sequences in online applications allows for tracking instances over multiple frames. Based on an instances history of shape and uncertainty information, we construct temporal instance-wise aggregated metrics. The latter are used as input to post-processing models that estimate the prediction quality in terms of instance-wise intersection over union. The proposed method only requires a readily trained neural network (that may operate on single frames) and video sequence input. In our experiments, we further demonstrate the use of the proposed method by replacing the traditional score value from object detection and thereby improving the overall performance of the instance segmentation network.
['Hanno Gottschalk', 'Fabian Hueger', 'Serin Varghese', 'Peter Schlicht', 'Matthias Rottmann', 'Kira Maag']
2020-12-14
null
null
null
null
['video-instance-segmentation']
['computer-vision']
[ 6.44873798e-01 3.70982230e-01 1.39998337e-02 -8.29584181e-01 -8.06795120e-01 -3.71163130e-01 4.61925238e-01 4.16387677e-01 -5.90817392e-01 9.82930183e-01 -6.84306622e-01 -2.83518791e-01 -4.50356543e-01 -8.84064734e-01 -1.20821238e+00 -5.41199386e-01 -2.17610762e-01 4.33352172e-01 6.74211800e-01 1.51241690e-01 3.57301921e-01 5.65670133e-01 -2.11685777e+00 3.55701983e-01 8.90243769e-01 1.73328638e+00 1.52585477e-01 8.95771265e-01 2.74527688e-02 7.42677629e-01 -7.15330660e-01 -3.67807001e-01 3.23266298e-01 2.53419757e-01 -3.13209981e-01 5.94028905e-02 3.95905674e-01 -3.73091727e-01 2.48781934e-01 1.19992983e+00 1.92017516e-03 3.62912357e-01 5.03008008e-01 -1.53020847e+00 1.68999046e-01 5.32887876e-01 -2.16484010e-01 2.74088830e-01 -1.13801591e-01 2.87529767e-01 7.54441381e-01 -7.71906495e-01 1.92210153e-01 1.01392102e+00 5.45725584e-01 4.70560715e-02 -8.90121341e-01 -6.14335895e-01 2.69451559e-01 6.73524261e-01 -1.16762471e+00 -4.97584045e-01 6.20254278e-01 -6.38594508e-01 7.86865890e-01 2.47445554e-01 5.31848311e-01 7.07954884e-01 1.00180693e-01 6.40424073e-01 1.06042564e+00 -2.98090041e-01 5.27629912e-01 3.48396122e-01 2.42204845e-01 2.97137111e-01 5.03227174e-01 4.28367108e-01 -4.15600002e-01 3.10813546e-01 4.74797249e-01 -1.97931275e-01 1.48965921e-02 -1.16929464e-01 -8.33350062e-01 4.89968240e-01 2.38871336e-01 3.34950257e-03 -5.30849397e-01 3.52962047e-01 3.44611168e-01 -3.23090144e-02 5.16935468e-01 1.59066498e-01 -4.27615881e-01 -3.14567685e-01 -1.11812186e+00 2.30464488e-01 5.59685230e-01 9.58190084e-01 7.08869636e-01 1.31089777e-01 -1.43199533e-01 6.76773429e-01 1.51331633e-01 3.61988604e-01 1.55770436e-01 -1.01731837e+00 4.43301409e-01 4.42854971e-01 4.52748150e-01 -1.09264588e+00 -3.68163884e-01 -5.51540017e-01 -5.51210821e-01 4.53636706e-01 5.23795664e-01 -5.58619462e-02 -8.83035362e-01 1.55204666e+00 3.55976611e-01 5.53919315e-01 1.08877167e-01 7.25400865e-01 2.59063184e-01 6.33244634e-01 7.49631152e-02 -3.38063478e-01 1.05012298e+00 -5.41654885e-01 -6.63207829e-01 -2.29434043e-01 2.94927448e-01 -4.77486074e-01 6.94703460e-01 8.15299869e-01 -1.07143462e+00 -9.36231494e-01 -1.21454859e+00 5.29645383e-01 -5.10858297e-01 3.16030115e-01 1.39790267e-01 6.88042402e-01 -6.59749687e-01 8.78429353e-01 -9.78631556e-01 -3.75688933e-02 2.02093914e-01 6.74351275e-01 9.53165963e-02 2.05810517e-01 -1.27486980e+00 1.04897237e+00 9.97473001e-01 6.33668363e-01 -7.76352584e-01 -4.46964800e-01 -7.87761390e-01 -5.93801700e-02 7.38831341e-01 1.69262365e-02 1.35279751e+00 -1.00777864e+00 -1.15115643e+00 1.36587903e-01 1.61353171e-01 -1.04678309e+00 8.49590540e-01 -3.34803730e-01 -4.33345765e-01 2.34940629e-02 -2.12716118e-01 6.46863401e-01 8.26438546e-01 -1.47930193e+00 -1.33759081e+00 -1.65064201e-01 3.78416330e-01 -8.24654698e-02 6.66406378e-03 -5.09150028e-01 -2.28240937e-01 -2.01214049e-02 5.27700409e-02 -8.64089251e-01 -1.70740142e-01 -9.06413272e-02 -3.74239624e-01 -1.02434769e-01 7.03254759e-01 -7.39374757e-01 1.02780604e+00 -1.95468855e+00 -3.81015748e-01 5.70720613e-01 -1.10558711e-01 2.30892152e-01 2.76211709e-01 -9.60599408e-02 1.54952649e-02 -5.88533729e-02 -4.28059876e-01 -1.20234296e-01 -9.23026577e-02 4.23988372e-01 -1.07521370e-01 3.17556530e-01 7.53683925e-01 3.32858682e-01 -7.90152013e-01 -5.61373353e-01 6.74086750e-01 2.21443042e-01 -2.34941512e-01 3.26614708e-01 -4.94488716e-01 2.02406228e-01 -3.39221865e-01 4.76120472e-01 6.36611044e-01 7.47734010e-02 -1.87198967e-02 -4.10777092e-01 -2.79768288e-01 -1.26581475e-01 -1.58772004e+00 9.44220006e-01 -5.01434088e-01 6.55769050e-01 -2.40162626e-01 -1.26752210e+00 1.00676680e+00 2.42076680e-01 3.19850475e-01 -6.36783779e-01 2.97470868e-01 3.50642979e-01 3.61157358e-02 -6.10021532e-01 8.47665012e-01 1.71470284e-01 3.60713303e-02 -1.19686060e-01 -2.28303939e-01 -1.60612583e-01 4.11314458e-01 -1.94162667e-01 7.34655797e-01 3.16567570e-01 2.65956521e-01 2.51509957e-02 7.44112551e-01 1.11714356e-01 6.32243812e-01 7.59382308e-01 -2.41076216e-01 3.75620455e-01 5.32325864e-01 -4.37738627e-01 -1.12863028e+00 -7.97324419e-01 -2.28704914e-01 5.25173843e-01 1.80644125e-01 3.49517882e-01 -8.66592646e-01 -4.62412417e-01 -1.35312201e-02 1.11847305e+00 -5.42702913e-01 -2.82165647e-01 -4.65210348e-01 -5.40456414e-01 3.28999251e-01 6.41124547e-01 3.75424296e-01 -8.25392902e-01 -9.46787477e-01 3.59179616e-01 -9.16099772e-02 -1.35957575e+00 3.26318562e-01 4.74625558e-01 -8.40853095e-01 -1.12137747e+00 -2.48922691e-01 -2.51710325e-01 5.91040075e-01 -2.02580303e-01 1.00933349e+00 1.37240559e-01 -6.86848089e-02 3.06097090e-01 -2.40004838e-01 -7.81140387e-01 -5.89497924e-01 -2.42041424e-01 2.46733755e-01 1.94147438e-01 1.74878910e-01 -2.72041649e-01 -2.94269681e-01 3.88137639e-01 -1.14321065e+00 -1.19887672e-01 5.63484848e-01 7.72128165e-01 9.03711200e-01 3.73255461e-01 7.62882411e-01 -7.91928649e-01 4.97542679e-01 -3.98392498e-01 -1.37243438e+00 3.03151220e-01 -7.52149940e-01 4.11030687e-02 5.74953973e-01 -4.52614278e-01 -1.03729522e+00 2.07314312e-01 -3.08573581e-02 -6.11947000e-01 -3.56426597e-01 4.67163324e-01 7.80182853e-02 2.85708308e-02 5.77868223e-01 -7.32572824e-02 -7.76606724e-02 7.58249462e-02 1.05626360e-01 5.48466384e-01 7.22295225e-01 -5.49426913e-01 5.86147547e-01 1.48791209e-01 3.18463504e-01 -7.48791218e-01 -7.71652997e-01 -3.15012246e-01 -6.32065713e-01 -9.64777946e-01 7.28338242e-01 -7.21315324e-01 -8.38156164e-01 3.28219444e-01 -1.18708670e+00 -7.07671717e-02 -3.27790022e-01 5.14081955e-01 -7.44934738e-01 1.75460428e-01 -8.87733698e-02 -1.51265478e+00 1.32789522e-01 -1.30963147e+00 8.70781243e-01 2.57802695e-01 8.21557567e-02 -7.05641150e-01 -5.17543495e-01 2.48983905e-01 1.25281289e-01 4.70017433e-01 5.15119791e-01 -9.22009826e-01 -1.00852251e+00 -5.21526277e-01 -2.22359374e-01 7.94081509e-01 -2.76573449e-01 3.08135659e-01 -1.19969165e+00 1.89828083e-01 -2.16116034e-03 -1.55491874e-01 5.96887767e-01 7.37186670e-01 1.30585027e+00 -1.08483963e-01 -7.23535568e-02 -2.79763304e-02 1.54528940e+00 6.02463722e-01 5.87911427e-01 4.93797302e-01 3.21015835e-01 1.02889872e+00 1.35669994e+00 5.59673548e-01 7.02370554e-02 5.22422493e-01 9.86182332e-01 2.47983113e-01 4.16228801e-01 3.07531863e-01 2.41779447e-01 4.03348595e-01 -1.83909938e-01 -3.13849270e-01 -7.97422886e-01 6.90633237e-01 -2.02911711e+00 -9.60362613e-01 -1.97983429e-01 2.52258921e+00 4.28110451e-01 8.99532259e-01 6.37319610e-02 5.63390851e-01 9.94130254e-01 -1.98539168e-01 -6.35444283e-01 -4.67629045e-01 3.63707274e-01 -1.24673940e-01 7.99471736e-01 5.02162993e-01 -1.05265653e+00 4.60263550e-01 5.29407883e+00 7.84301281e-01 -8.41739297e-01 -2.13775560e-01 9.90220428e-01 -1.13064408e-01 8.60502124e-02 -1.62132993e-01 -7.90755987e-01 4.80027944e-01 1.19169033e+00 -2.15962585e-02 1.97186291e-01 1.10114193e+00 4.82971370e-01 -6.34996831e-01 -1.27913713e+00 6.95986331e-01 -2.36760825e-01 -1.16897964e+00 -3.38886976e-01 -1.46818131e-01 5.29362082e-01 -3.86458427e-01 6.71699643e-02 1.47489086e-01 -7.47912675e-02 -9.06030357e-01 1.12271821e+00 9.46653843e-01 4.82945621e-01 -1.12563753e+00 1.10113335e+00 5.67280114e-01 -1.11384082e+00 -2.74539977e-01 -3.37499917e-01 -5.74524216e-02 3.80279601e-01 9.39693809e-01 -1.28280139e+00 5.59472144e-01 5.43600500e-01 3.26657116e-01 -2.90390849e-01 1.18369198e+00 7.67350867e-02 5.71573555e-01 -6.04885578e-01 -2.52983540e-01 1.96646824e-01 -3.31879929e-02 4.95614678e-01 1.11917138e+00 4.54579294e-01 -5.29328659e-02 1.48983210e-01 9.59159553e-01 3.26784670e-01 -3.40208292e-01 -5.14616251e-01 1.24772541e-01 4.95018393e-01 1.08643317e+00 -8.59940708e-01 -5.21261752e-01 -2.66327560e-01 2.79837608e-01 -5.38556650e-02 1.78270772e-01 -1.18995428e+00 -2.08520070e-01 5.59662223e-01 -5.39668277e-03 2.28700206e-01 -2.18686879e-01 -5.25194883e-01 -4.22059387e-01 1.97819650e-01 -5.60100496e-01 8.93298537e-02 -8.23476434e-01 -9.26586449e-01 8.03917885e-01 3.10507476e-01 -1.55316377e+00 -7.24839926e-01 -8.12858880e-01 -3.17258328e-01 7.28632331e-01 -1.54791677e+00 -7.64186084e-01 -3.70829612e-01 2.22674236e-01 7.85360396e-01 2.01595053e-01 2.32103303e-01 2.00585693e-01 -6.05138183e-01 3.59747827e-01 -1.64832875e-01 -2.78141767e-01 2.94103026e-01 -1.09129834e+00 8.58195499e-02 1.14104176e+00 -1.36502847e-01 5.36707900e-02 1.27370417e+00 -8.18615556e-01 -7.88528562e-01 -1.30066061e+00 4.13480610e-01 -2.97813684e-01 5.60780346e-01 1.47338286e-01 -9.46473300e-01 3.77220809e-01 -2.27008238e-01 8.81753936e-02 9.28703919e-02 -3.21179807e-01 2.84629434e-01 -4.39843774e-01 -1.38024867e+00 2.49010354e-01 6.36240840e-01 -3.09195578e-01 -3.60997289e-01 2.05474019e-01 6.93837404e-01 -6.68119133e-01 -8.47031951e-01 7.55440414e-01 6.17844939e-01 -1.34076679e+00 8.21218550e-01 -2.81360596e-01 3.39133263e-01 -4.85524565e-01 -2.29175478e-01 -9.61614311e-01 1.60040051e-01 1.61910653e-01 -2.04055071e-01 1.13385391e+00 6.83203697e-01 -4.18807954e-01 8.49835396e-01 1.07680643e+00 -2.60858119e-01 -7.14743733e-01 -1.10180044e+00 -8.74363601e-01 -6.04907572e-01 -1.14158285e+00 5.50122559e-01 3.08052957e-01 -5.84158778e-01 -5.00533283e-01 -3.99652392e-01 8.76139700e-01 7.95605421e-01 -2.92646974e-01 5.78472018e-01 -1.31866443e+00 -2.80508727e-01 -7.75444880e-02 -8.15050006e-01 -6.05833054e-01 -8.38310197e-02 -1.49852410e-01 5.11473238e-01 -1.19055665e+00 -3.17887962e-01 -3.79198819e-01 -5.30222833e-01 2.88960920e-03 5.48364222e-02 6.09463193e-02 7.43316710e-02 -3.78453545e-02 -5.01235843e-01 2.31048763e-01 7.53734112e-01 -1.08358085e-01 -6.64791688e-02 4.08490419e-01 6.71205390e-03 8.92587364e-01 8.20243239e-01 -4.42893267e-01 -5.84218204e-01 -1.50692612e-01 3.21808517e-01 1.79001540e-01 5.86411297e-01 -1.54970872e+00 2.06177086e-01 -3.86120528e-01 4.51796561e-01 -9.51034606e-01 5.31385720e-01 -1.39990056e+00 2.58085072e-01 3.67059082e-01 -4.41587776e-01 7.96073079e-02 2.89918333e-01 8.58500361e-01 -2.92548954e-01 -6.52042210e-01 7.05249667e-01 -3.93443182e-02 -1.06079805e+00 4.00221497e-02 -2.62386292e-01 -4.41500932e-01 1.27074325e+00 -5.36894679e-01 1.92451365e-02 -3.13960999e-01 -8.42998862e-01 2.48602062e-01 1.26165330e-01 3.90021503e-01 6.97315991e-01 -1.00556362e+00 -3.06765348e-01 -3.54346894e-02 1.39131561e-01 2.74227679e-01 3.51654202e-01 6.95620358e-01 -2.98489481e-01 2.89904237e-01 -1.82500437e-01 -1.06018710e+00 -1.09546638e+00 3.83637488e-01 4.01008576e-01 -1.13400534e-01 4.18666713e-02 7.06297338e-01 -1.04540639e-01 1.07149919e-02 5.17043352e-01 -8.72212231e-01 -4.88761842e-01 2.50464845e-02 4.65570152e-01 6.05466843e-01 2.85036325e-01 -5.06140471e-01 -1.65526152e-01 1.84386849e-01 2.16256127e-01 -1.44409627e-01 1.14176750e+00 -2.55966514e-01 2.43369505e-01 8.52013230e-01 6.79616034e-01 -4.64082867e-01 -1.60360181e+00 9.89085138e-02 3.40820432e-01 -4.23028022e-01 1.90946565e-03 -8.41911972e-01 -9.37452853e-01 8.28115761e-01 9.99447763e-01 4.24554795e-01 1.22378516e+00 -3.05441350e-01 5.30178010e-01 4.91233885e-01 5.30747950e-01 -1.56865108e+00 -2.88605481e-01 1.81184396e-01 7.36685812e-01 -1.38682520e+00 -1.72533900e-01 -4.23145026e-01 -5.92640460e-01 1.21511352e+00 8.29411089e-01 1.81779377e-02 4.87722069e-01 3.56530815e-01 -9.09895599e-02 1.85897052e-01 -7.63013840e-01 -1.24427706e-01 4.11338747e-01 4.52972591e-01 3.93910035e-02 1.92350209e-01 -1.87909007e-01 6.53027892e-01 -9.70595609e-03 2.43060842e-01 6.46262527e-01 8.01443160e-01 -7.94913411e-01 -7.88387179e-01 -5.48319399e-01 7.62280226e-01 -3.29376936e-01 2.15110391e-01 1.64168164e-01 7.42811084e-01 5.45469761e-01 1.20943606e+00 1.45160928e-01 -6.48771286e-01 5.98684371e-01 8.23036507e-02 5.88716790e-02 -2.39818603e-01 -2.38289669e-01 -3.36295366e-01 6.07828498e-01 -5.50985515e-01 -6.66178703e-01 -7.28188396e-01 -1.24662411e+00 9.78934988e-02 -6.50320351e-01 -2.43153842e-03 1.05922580e+00 1.18677175e+00 1.33078154e-02 8.09168875e-01 5.11761427e-01 -9.76243675e-01 -6.72425628e-01 -7.89851964e-01 -4.67132717e-01 2.00913414e-01 4.00351465e-01 -9.66183603e-01 -3.37670982e-01 2.42170766e-01]
[7.875953674316406, -0.8416284322738647]
e9e87080-4f90-4bc1-acae-86e796e1d909
fast-facial-landmark-detection-and
2101.10808
null
https://arxiv.org/abs/2101.10808v3
https://arxiv.org/pdf/2101.10808v3.pdf
Fast Facial Landmark Detection and Applications: A Survey
Dense facial landmark detection is one of the key elements of face processing pipeline. It is used in virtual face reenactment, emotion recognition, driver status tracking, etc. Early approaches were suitable for facial landmark detection in controlled environments only, which is clearly insufficient. Neural networks have shown an astonishing qualitative improvement for in-the-wild face landmark detection problem, and are now being studied by many researchers in the field. Numerous bright ideas are proposed, often complimentary to each other. However, exploration of the whole volume of novel approaches is quite challenging. Therefore, we present this survey, where we summarize state-of-the-art algorithms into categories, provide a comparison of recently introduced in-the-wild datasets (e.g., 300W, AFLW, COFW, WFLW) that contain images with large pose, face occlusion, taken in unconstrained conditions. In addition to quality, applications require fast inference, and preferably on mobile devices. Hence, we include information about algorithm inference speed both on desktop and mobile hardware, which is rarely studied. Importantly, we highlight problems of algorithms, their applications, vulnerabilities, and briefly touch on established methods. We hope that the reader will find many novel ideas, will see how the algorithms are used in applications, which will enable further research.
['Larysa Koriashkina', 'Kostiantyn Khabarlak']
2021-01-12
null
null
null
null
['face-reenactment']
['computer-vision']
[ 7.03698322e-02 6.58879131e-02 -1.41869962e-01 -6.38304651e-01 -3.79262596e-01 -4.61357027e-01 5.09439826e-01 -3.91923517e-01 -5.33932865e-01 3.83728355e-01 -2.30698973e-01 -6.57545775e-02 -1.63704827e-01 -3.95596236e-01 -4.52071577e-01 -7.72239029e-01 -1.80183128e-01 2.66953290e-01 3.03212609e-02 -3.68475139e-01 3.41510683e-01 1.04734302e+00 -2.47058201e+00 3.36066857e-02 7.41928294e-02 1.12037802e+00 -3.02726418e-01 5.08099616e-01 -3.10272388e-02 3.19350064e-01 -5.61497450e-01 -7.14339197e-01 2.64179707e-01 2.48906072e-02 -3.93790603e-01 -7.78775662e-02 1.07585156e+00 -1.63460970e-01 -1.44009605e-01 9.88123178e-01 9.43220675e-01 1.25137076e-01 1.73563480e-01 -1.60012734e+00 1.10435847e-03 -1.66369081e-01 -7.44728088e-01 1.77014053e-01 3.95039558e-01 1.15468599e-01 4.42791104e-01 -1.36410737e+00 7.49633908e-01 1.29618907e+00 1.01953733e+00 9.28943694e-01 -6.82102859e-01 -9.87603486e-01 3.33742611e-02 5.59412658e-01 -1.62678099e+00 -1.23455560e+00 7.76760697e-01 -1.55829504e-01 1.01073217e+00 3.75173211e-01 5.86397886e-01 9.23569083e-01 9.17132869e-02 6.41289294e-01 9.65322495e-01 -4.13289070e-01 -2.57915701e-03 2.35838711e-01 8.37410763e-02 1.20218873e+00 -2.15600207e-02 3.70293140e-01 -1.18221450e+00 -1.08557366e-01 2.92939186e-01 -1.37814477e-01 -2.08688751e-01 -1.80309206e-01 -6.05783343e-01 8.23337853e-01 9.63989869e-02 1.08167738e-01 -5.52818133e-03 -7.30191544e-02 5.48240900e-01 3.16431373e-01 6.71179891e-01 -1.48248523e-01 -4.27864552e-01 -3.57209057e-01 -1.17501295e+00 3.52298588e-01 5.58990061e-01 8.64710271e-01 6.97462201e-01 2.85392385e-02 8.76209438e-02 7.01895058e-01 3.66928130e-01 5.19200325e-01 3.33854586e-01 -9.18597102e-01 -1.35413021e-01 2.10813507e-01 -1.75185069e-01 -1.09308124e+00 -7.76608109e-01 3.41701619e-02 -7.26357520e-01 7.39377260e-01 2.96689034e-01 -1.59067050e-01 -7.97396064e-01 1.45286369e+00 6.24275863e-01 5.74702680e-01 -3.45864475e-01 6.27208769e-01 1.34218597e+00 1.35764495e-01 -7.27392063e-02 -3.62238675e-01 1.58733058e+00 -7.09618747e-01 -8.93133819e-01 -3.13375264e-01 3.72770995e-01 -9.33768809e-01 6.96027517e-01 4.16424841e-01 -1.11134863e+00 -7.40067422e-01 -9.57842946e-01 -7.58512616e-02 -7.12664187e-01 2.05589041e-01 9.22634840e-01 1.21841502e+00 -1.51598620e+00 5.56060910e-01 -6.39405191e-01 -6.04018688e-01 7.52032697e-01 8.23592365e-01 -6.35513067e-01 7.35617802e-02 -1.00252247e+00 9.53630388e-01 -1.70610264e-01 6.21906698e-01 -6.08313501e-01 -5.53461075e-01 -1.01341963e+00 -3.51609230e-01 5.90388775e-01 -1.70346931e-01 1.22508967e+00 -6.56502485e-01 -1.75776899e+00 1.36353409e+00 -7.27404416e-01 -6.14811964e-02 4.50975507e-01 4.65078233e-03 -6.80219054e-01 6.90710451e-03 -3.17513585e-01 8.22950423e-01 1.21412754e+00 -8.34919095e-01 -7.51305878e-01 -6.87542796e-01 -2.19098791e-01 -1.12701152e-02 -2.10502461e-01 5.98546684e-01 -7.80034244e-01 -1.01408511e-01 -1.04174033e-01 -6.83902562e-01 1.67350456e-01 3.43122125e-01 -3.51132005e-02 -5.96568882e-01 1.10811436e+00 -4.57479209e-01 1.15366483e+00 -2.39745808e+00 -3.52206290e-01 3.79631758e-01 3.53251547e-01 5.66509664e-01 -1.58313558e-01 1.25987157e-01 -2.31611535e-01 -1.74799323e-01 2.50956719e-03 -8.09980273e-01 7.14187920e-02 -9.65395570e-03 -7.82337636e-02 9.30257916e-01 2.17050716e-01 1.02442336e+00 -4.24031347e-01 -5.21022558e-01 2.88004190e-01 7.36341119e-01 -3.09498161e-01 -8.78462270e-02 3.69645298e-01 -2.17071120e-02 -2.71838112e-03 1.06315267e+00 9.44009721e-01 5.78367233e-01 -3.22000086e-01 -2.77173162e-01 -2.38268271e-01 -3.45563203e-01 -1.31428909e+00 1.45826912e+00 -2.67350256e-01 1.12403214e+00 5.15733004e-01 -6.85342014e-01 9.62364972e-01 4.33635414e-01 3.12032968e-01 -5.39314508e-01 3.35071683e-01 1.47132501e-02 -1.99797273e-01 -5.32313228e-01 4.81370032e-01 -1.11283153e-01 4.90127385e-01 2.54391521e-01 4.43937592e-02 1.79333434e-01 1.39224946e-01 -1.82862833e-01 7.28135109e-01 6.86420053e-02 2.65131712e-01 -1.21520557e-01 6.79201365e-01 -4.54229891e-01 4.18817192e-01 6.57030463e-01 -8.10231090e-01 5.18100500e-01 4.82381463e-01 -5.81402898e-01 -3.84892881e-01 -6.81860387e-01 -5.33836961e-01 1.14674556e+00 -1.10735640e-01 -6.45994067e-01 -1.04251111e+00 -6.18727684e-01 2.81399358e-02 3.60173546e-02 -8.35840702e-01 -9.16149095e-02 -5.58994710e-01 -8.88551891e-01 8.01526785e-01 5.66824436e-01 5.67722976e-01 -1.21268618e+00 -8.85303557e-01 -3.35239589e-01 2.94276357e-01 -1.06100333e+00 -1.58436134e-01 -1.44089773e-01 -6.45158887e-01 -1.27536702e+00 -4.63812530e-01 -8.01937580e-01 6.21713996e-01 2.99197823e-01 1.15671551e+00 6.35403395e-01 -8.00405502e-01 5.98679841e-01 3.45194899e-02 -6.82168067e-01 1.94759965e-01 -3.17713767e-01 4.90415901e-01 2.35863611e-01 9.97374356e-01 -1.97289437e-01 -3.70043695e-01 4.28661793e-01 -4.92163599e-01 -4.68018442e-01 2.94857711e-01 7.97840536e-01 3.28348875e-01 8.77061337e-02 1.60901219e-01 -9.19756711e-01 4.62532669e-01 2.43159086e-02 -8.16265881e-01 1.32377774e-01 -5.89273453e-01 -3.77416909e-01 6.86475709e-02 -1.99231952e-01 -1.05622220e+00 2.44916946e-01 -6.88311577e-01 -3.14044774e-01 -4.73765463e-01 2.84162164e-02 -4.36811566e-01 -8.20577681e-01 6.35595798e-01 -1.48553535e-01 3.38473678e-01 -2.49685317e-01 3.86571348e-01 7.42406785e-01 6.16111100e-01 -2.65517384e-01 8.60332906e-01 7.59602547e-01 2.66940624e-01 -1.34919226e+00 -3.22265118e-01 -5.15927017e-01 -8.55329633e-01 -5.72139740e-01 4.40346926e-01 -4.75498796e-01 -1.32089257e+00 6.91672683e-01 -1.08487511e+00 -8.24508537e-03 -1.27990186e-01 1.28033623e-01 -3.59365284e-01 1.23894125e-01 -3.08818161e-01 -1.10514915e+00 -1.15412071e-01 -1.23031020e+00 1.36669338e+00 5.63947201e-01 -1.44281149e-01 -1.00306904e+00 -2.06223011e-01 1.44970492e-01 5.33325076e-01 9.47250500e-02 2.94943929e-01 -2.05579951e-01 -1.79053888e-01 -6.13978624e-01 -1.68099359e-01 -1.98243652e-02 -1.05999187e-01 4.27026540e-01 -1.70638728e+00 -3.44587743e-01 2.53519649e-03 -2.32445762e-01 7.44598269e-01 4.19377923e-01 1.28075981e+00 1.24909632e-01 -5.44133186e-01 8.41429412e-01 9.36700702e-01 1.44257665e-01 8.19450200e-01 5.16734272e-02 3.56888950e-01 1.04701507e+00 6.32058680e-01 2.86481440e-01 -1.98156014e-02 8.37261081e-01 2.93835610e-01 -2.86832213e-01 -1.13488756e-01 2.01696455e-01 3.51281404e-01 9.08720493e-02 -5.07162213e-01 1.23083971e-01 -7.46270955e-01 3.30458842e-02 -1.44691670e+00 -1.10618722e+00 -7.40244091e-02 2.11577177e+00 3.30653906e-01 1.99008430e-03 2.38963291e-01 3.39887142e-01 5.65320969e-01 2.71745712e-01 -3.25960696e-01 -3.99397582e-01 -3.99475545e-02 5.96611440e-01 3.07600856e-01 6.95196211e-01 -1.12240255e+00 1.17272627e+00 6.76873541e+00 9.35249209e-01 -1.25158763e+00 1.41916335e-01 7.49793589e-01 -2.83437431e-01 2.57083565e-01 -4.17618245e-01 -1.37883127e+00 1.22622162e-01 9.33309317e-01 2.09114239e-01 3.85273278e-01 9.74099338e-01 7.25381374e-02 -3.68517727e-01 -1.08744907e+00 1.60751295e+00 6.84178352e-01 -1.16441822e+00 -6.31261826e-01 -5.32330349e-02 9.46868360e-02 -4.15135436e-02 3.67366165e-01 3.32907408e-01 -5.44040620e-01 -1.26010835e+00 4.18346643e-01 3.57153118e-01 1.02203631e+00 -9.85791802e-01 9.02085781e-01 -2.93425694e-02 -1.31023967e+00 1.76610515e-01 -4.64296013e-01 3.79186147e-03 -1.93512905e-02 3.52767080e-01 -4.79955524e-01 1.96534052e-01 9.61711347e-01 6.03608608e-01 -9.14516866e-01 8.83747935e-01 -9.21499208e-02 4.23038185e-01 -5.34281850e-01 -1.71272412e-01 -6.38934001e-02 3.56364809e-02 2.64902830e-01 1.13429558e+00 1.01213641e-01 4.16169725e-02 -2.40758076e-01 3.77496749e-01 7.61414319e-02 1.00764390e-02 -7.24492669e-01 3.40705842e-01 2.95533717e-01 1.68133867e+00 -9.30779517e-01 -5.86582161e-02 -6.12799585e-01 8.24738562e-01 9.04424489e-02 2.52310008e-01 -8.90399218e-01 -5.52536786e-01 1.14433634e+00 1.68820366e-01 -1.72045320e-01 -1.05752043e-01 2.51342878e-02 -5.88403881e-01 8.33567679e-02 -8.42051387e-01 3.16624582e-01 -5.29782534e-01 -9.35670733e-01 8.15924942e-01 7.78016001e-02 -7.76252270e-01 -3.10814649e-01 -1.21698129e+00 -6.77987933e-01 7.58457243e-01 -1.73644865e+00 -8.55607986e-01 -5.85998535e-01 7.58139312e-01 5.80690861e-01 -5.06338716e-01 8.67153645e-01 5.24431586e-01 -9.02211308e-01 1.03729582e+00 -1.31946608e-01 9.50476080e-02 8.03786993e-01 -6.49828196e-01 6.19438052e-01 7.63195753e-01 4.08510953e-01 6.46158576e-01 5.16210556e-01 -3.43904406e-01 -1.64283574e+00 -7.78321624e-01 9.68095899e-01 -7.20930099e-01 1.74069360e-01 -5.85167289e-01 -7.30594456e-01 5.53671062e-01 1.52714252e-01 3.04660738e-01 6.46258473e-01 3.99335265e-01 -8.16679746e-02 -3.48238528e-01 -1.41405821e+00 6.22288585e-01 1.20379412e+00 -6.08211815e-01 -4.62694466e-02 3.43517095e-01 -5.40135056e-02 -7.47891188e-01 -1.54322788e-01 3.83559167e-01 8.53019357e-01 -1.35522807e+00 1.04516399e+00 -4.77738500e-01 -3.97865623e-01 -1.64764345e-01 2.28270024e-01 -8.96920502e-01 -1.17966056e-01 -9.36392188e-01 -1.32752851e-01 1.23934364e+00 5.89586236e-02 -8.59024525e-01 1.22829163e+00 7.61176825e-01 5.24350908e-03 -9.07730401e-01 -1.24028218e+00 -5.63746750e-01 -4.60220575e-01 -7.74844706e-01 6.21189952e-01 5.72810948e-01 -1.04127400e-01 7.56006241e-02 -1.95448637e-01 6.20188937e-02 5.58678150e-01 -3.80299389e-01 8.59724998e-01 -1.43453181e+00 3.21941167e-01 -6.18038774e-01 -8.19821596e-01 -6.38682127e-01 5.27694881e-01 -5.96828759e-01 1.84524502e-03 -7.89391220e-01 -2.12700605e-01 -1.00090399e-01 1.07121125e-01 6.42638743e-01 3.99363637e-02 8.84943306e-01 -4.39320430e-02 -2.18616053e-01 -2.42813736e-01 3.09174061e-01 7.54413426e-01 -6.94526732e-02 6.75717220e-02 1.84544772e-02 -4.57325071e-01 1.09689128e+00 8.28685105e-01 -2.16684401e-01 -2.94762582e-01 -9.06687751e-02 1.39657959e-01 -5.63289404e-01 4.32158887e-01 -1.09944201e+00 4.41020399e-01 5.43293878e-02 6.28067195e-01 -6.42074704e-01 7.95065224e-01 -8.46319079e-01 -7.29191750e-02 1.89590245e-01 2.31126443e-01 2.19811678e-01 5.55902064e-01 1.27189755e-01 -3.02533209e-01 -2.97932953e-01 9.41246510e-01 9.14442167e-02 -1.16363752e+00 6.69022501e-01 -1.70461565e-01 -2.30075955e-01 1.16526628e+00 -5.81479549e-01 -7.82896504e-02 -3.66956860e-01 -5.69475174e-01 2.24443339e-02 1.20504409e-01 5.33196926e-01 8.48466098e-01 -1.14953828e+00 -6.22932673e-01 9.77883279e-01 1.23974241e-01 -2.52211094e-01 3.10919464e-01 9.27424550e-01 -4.85320926e-01 3.52132648e-01 -1.74238294e-01 -6.60999835e-01 -1.93044448e+00 4.74794865e-01 4.38493282e-01 4.84560221e-01 -4.69402999e-01 1.29625142e+00 -1.20091967e-01 -2.17024595e-01 6.37425661e-01 7.19113201e-02 -4.37992483e-01 4.65790421e-01 1.03285503e+00 3.97520542e-01 6.11546218e-01 -9.82379496e-01 -6.13766134e-01 1.05658090e+00 6.38621077e-02 3.22948545e-02 9.74613369e-01 1.65543351e-02 -3.06274086e-01 9.54555720e-03 1.14677882e+00 3.04017607e-02 -1.01095057e+00 1.15943156e-01 -5.51180765e-02 -7.49902070e-01 1.08540647e-01 -4.26470011e-01 -1.36688995e+00 1.23386419e+00 1.21222293e+00 -2.43213803e-01 1.40591657e+00 -1.07098781e-01 3.96024227e-01 4.49078619e-01 4.27392751e-01 -1.19376111e+00 -2.14921221e-01 5.29091060e-01 8.63439202e-01 -1.34103119e+00 2.84198541e-02 -5.69571376e-01 -1.88777909e-01 1.36451173e+00 6.62020981e-01 3.18024755e-01 9.61662412e-01 6.79966390e-01 3.07164490e-01 -4.76550817e-01 -3.90162855e-01 -3.10075879e-01 2.66450077e-01 8.13340604e-01 4.95543510e-01 -4.01928216e-01 1.75402880e-01 2.33194813e-01 -3.33304554e-01 -5.02621941e-02 -5.01015736e-03 8.09934974e-01 -2.90304154e-01 -1.06051886e+00 -6.55523539e-01 3.90376806e-01 -3.22071016e-01 1.88991707e-02 -4.23624307e-01 9.85171437e-01 4.51572001e-01 9.93644059e-01 1.05936766e-01 -3.82857621e-01 5.02910733e-01 3.81980598e-01 5.76252520e-01 -1.86954439e-01 -5.22807539e-01 -2.12261170e-01 -1.06438287e-02 -1.00504386e+00 -3.71314853e-01 -7.11992800e-01 -9.77313161e-01 -6.87386513e-01 -3.21601182e-01 3.63434665e-02 8.28011096e-01 8.09153676e-01 2.64169604e-01 2.19570398e-01 3.73713315e-01 -1.30274487e+00 3.01579498e-02 -8.15788329e-01 -6.88203573e-01 2.68330034e-02 4.31145459e-01 -1.05575716e+00 -4.97446328e-01 -4.66076061e-02]
[13.335107803344727, 0.730558454990387]
4be95811-f26a-4999-be1f-59299bda6a48
katsum-knowledge-aware-abstractive-text
2212.03371
null
https://arxiv.org/abs/2212.03371v1
https://arxiv.org/pdf/2212.03371v1.pdf
KATSum: Knowledge-aware Abstractive Text Summarization
Text Summarization is recognised as one of the NLP downstream tasks and it has been extensively investigated in recent years. It can assist people with perceiving the information rapidly from the Internet, including news articles, social posts, videos, etc. Most existing research works attempt to develop summarization models to produce a better output. However, advent limitations of most existing models emerge, including unfaithfulness and factual errors. In this paper, we propose a novel model, named as Knowledge-aware Abstractive Text Summarization, which leverages the advantages offered by Knowledge Graph to enhance the standard Seq2Seq model. On top of that, the Knowledge Graph triplets are extracted from the source text and utilised to provide keywords with relational information, producing coherent and factually errorless summaries. We conduct extensive experiments by using real-world data sets. The results reveal that the proposed framework can effectively utilise the information from Knowledge Graph and significantly reduce the factual errors in the summary.
['Jianhua Jiang', 'Edmund Lai', 'Weihua Li', 'Guan Wang']
2022-12-06
null
null
null
null
['abstractive-text-summarization']
['natural-language-processing']
[ 5.86340725e-01 4.47930157e-01 -3.10044497e-01 4.12355177e-03 -7.90820181e-01 -4.56976950e-01 6.55998766e-01 6.00365877e-01 -1.94619760e-01 9.88385618e-01 1.15567541e+00 1.78555265e-01 -2.93461293e-01 -5.01090288e-01 -4.10829246e-01 -3.28515857e-01 2.47133315e-01 1.13521941e-01 1.17428787e-01 -4.44311172e-01 9.72579420e-01 -6.28280779e-03 -1.41086936e+00 4.25212830e-01 1.55038619e+00 5.97692966e-01 2.76547700e-01 7.19938338e-01 -4.56537813e-01 9.71645236e-01 -9.72372830e-01 -6.55918598e-01 -1.79422289e-01 -6.68352842e-01 -9.32404935e-01 -1.07412506e-02 2.83182889e-01 -2.61284351e-01 -3.38862866e-01 1.16982400e+00 8.76330435e-01 3.43835115e-01 5.30337095e-01 -8.40767503e-01 -8.24307442e-01 8.99199128e-01 -4.98637319e-01 4.80091989e-01 8.96620035e-01 -2.71052849e-02 1.13742626e+00 -5.59768975e-01 7.52971172e-01 1.31008172e+00 4.10441130e-01 1.98057696e-01 -5.79206884e-01 -3.05264086e-01 1.66535661e-01 4.00607526e-01 -1.02930117e+00 -6.87555194e-01 8.69090855e-01 1.17259212e-01 1.27385044e+00 5.85911274e-01 7.73718417e-01 9.78023112e-01 3.72093618e-01 1.06841922e+00 6.51609838e-01 -3.42364848e-01 -1.73134208e-02 -6.35272488e-02 1.92910179e-01 5.09280682e-01 6.02466524e-01 -5.58415651e-01 -1.04896533e+00 -8.70246515e-02 1.20495632e-01 9.55971703e-03 -6.66241825e-01 2.58257002e-01 -1.26005054e+00 4.13273752e-01 3.17629606e-01 1.99185356e-01 -7.25724936e-01 -2.14789793e-01 6.03699565e-01 1.89310089e-01 7.69897223e-01 5.48349261e-01 -1.93209797e-01 -3.81607056e-01 -9.69346941e-01 2.84673393e-01 1.08081818e+00 1.12460470e+00 4.90822762e-01 7.48535618e-02 -6.83072627e-01 7.43217409e-01 2.03849450e-01 4.92787868e-01 7.43835449e-01 -9.50077415e-01 8.69947791e-01 1.02885318e+00 -9.37741622e-02 -1.50699413e+00 -2.26153955e-01 -6.37023449e-01 -1.08646905e+00 -7.91292310e-01 -3.52262974e-01 -2.46870965e-01 -5.99381328e-01 1.19733071e+00 1.22996852e-01 2.21645534e-01 4.74649131e-01 5.75751662e-01 1.45204484e+00 9.99467492e-01 -3.52227911e-02 -6.78412676e-01 1.21186614e+00 -1.19920385e+00 -1.24267447e+00 -1.70274734e-01 5.60811222e-01 -7.89648354e-01 5.65495431e-01 4.48817819e-01 -1.17489135e+00 -4.28069532e-01 -1.03488255e+00 -2.83808619e-01 -2.69779712e-01 2.02011876e-02 3.73430312e-01 2.28656843e-01 -1.12423134e+00 7.19377220e-01 -3.44227731e-01 -5.95452070e-01 4.42768604e-01 -5.62706264e-03 -1.95174560e-01 -2.65351206e-01 -1.26630449e+00 8.49599838e-01 9.38247979e-01 2.58690953e-01 -2.48983100e-01 -5.87195635e-01 -6.57916367e-01 2.15062708e-01 9.53246891e-01 -1.13696933e+00 1.23555028e+00 -7.72649348e-01 -1.55695105e+00 2.28555545e-01 -2.58804798e-01 -3.72737885e-01 3.43747526e-01 -5.46680331e-01 -3.98335129e-01 3.93312693e-01 6.99015567e-03 2.53738374e-01 6.08622789e-01 -9.99346495e-01 -6.62377715e-01 -3.64365667e-01 -2.96400283e-02 7.86272824e-01 -4.30636078e-01 -1.40137851e-01 -4.04597580e-01 -7.21274614e-01 -2.02269152e-01 -5.06937385e-01 -2.78967824e-02 -8.34668100e-01 -1.01700604e+00 -4.56295907e-01 7.25943029e-01 -1.18946469e+00 1.72567499e+00 -1.75955725e+00 4.24483418e-01 -1.49967268e-01 4.50096548e-01 6.63768411e-01 -1.36176720e-02 1.13560641e+00 4.80284035e-01 2.94396996e-01 -3.45608503e-01 -3.77202481e-02 -1.87278479e-01 5.87955154e-02 -2.27603361e-01 -1.29420072e-01 9.65194702e-02 1.10938561e+00 -1.26140070e+00 -7.42824852e-01 -1.37340814e-01 1.45689741e-01 -2.25310206e-01 2.14987248e-01 -2.74296314e-01 3.28856796e-01 -7.70156920e-01 4.16751355e-01 4.46635693e-01 -5.66540062e-02 5.00364117e-02 -1.63531095e-01 -1.00927122e-01 2.19658211e-01 -9.56582785e-01 1.91108322e+00 3.48047912e-02 5.44219017e-01 -2.69444287e-01 -1.12509453e+00 8.03979278e-01 3.29863548e-01 1.31231442e-01 -6.73310101e-01 1.57910556e-01 6.56611919e-02 -2.77515113e-01 -8.96545470e-01 1.14854634e+00 2.11158738e-01 -3.50812264e-02 2.30525732e-01 1.39176929e-02 -2.49665141e-01 5.34403086e-01 8.34508002e-01 1.13402092e+00 -2.06549820e-02 7.63856232e-01 1.14601865e-01 5.70215881e-01 2.37639800e-01 4.14030522e-01 8.24604213e-01 9.77021828e-02 3.54423136e-01 6.20253384e-01 8.04568008e-02 -6.69558823e-01 -6.07215345e-01 6.12862647e-01 6.69648170e-01 2.20711455e-01 -8.47972274e-01 -7.42009342e-01 -6.48628533e-01 -2.14231163e-01 1.01358867e+00 -1.05379634e-01 -4.29341853e-01 -4.30453628e-01 -4.32780862e-01 5.54735422e-01 3.63626271e-01 8.18142056e-01 -1.19867647e+00 -3.29339117e-01 2.11054713e-01 -6.52539730e-01 -9.77660179e-01 -5.00395656e-01 -4.60717469e-01 -1.08012497e+00 -9.69226718e-01 -6.32858217e-01 -6.27233505e-01 4.68802571e-01 5.70342422e-01 9.75767016e-01 1.45741850e-01 2.50528529e-02 4.80211318e-01 -9.18565392e-01 -6.64274871e-01 -4.76434082e-01 3.56290251e-01 -2.37484589e-01 -6.35640249e-02 4.17674594e-02 -4.84601647e-01 -6.14943981e-01 -4.76324588e-01 -1.16546249e+00 3.57798696e-01 1.06005836e+00 5.45083582e-01 4.11135018e-01 2.36289248e-01 1.25156212e+00 -1.26006591e+00 1.47880197e+00 -7.54275084e-01 2.17194870e-01 4.70528722e-01 -6.91044569e-01 1.34772182e-01 8.74766648e-01 2.84173321e-02 -1.51364875e+00 -4.43258971e-01 4.93634120e-02 2.26734281e-01 1.50313631e-01 1.09711730e+00 -5.08105233e-02 3.38367373e-01 3.10899407e-01 6.35170043e-01 -2.00638235e-01 -4.06337619e-01 6.37108982e-01 9.42586124e-01 6.25087261e-01 -2.63036549e-01 4.81206715e-01 9.12984014e-02 -1.80831954e-01 -1.11189830e+00 -1.08040237e+00 -7.30466306e-01 -3.70393723e-01 -4.11319703e-01 5.12854934e-01 -7.79633105e-01 -5.34970820e-01 3.37714374e-01 -1.35766697e+00 4.28940833e-01 -2.71532148e-01 3.03249478e-01 -2.87849396e-01 1.00103462e+00 -2.50742912e-01 -7.23256350e-01 -1.10821211e+00 -5.06097138e-01 9.26009834e-01 7.07289577e-01 -3.15440029e-01 -8.49121034e-01 6.98241815e-02 4.74921107e-01 3.69539440e-01 2.68838316e-01 9.87249911e-01 -1.14960313e+00 -4.86638457e-01 -2.09856883e-01 -2.01856762e-01 3.60701442e-01 1.15118161e-01 4.65329811e-02 -4.46115822e-01 1.10024707e-02 -8.28622356e-02 -1.61527872e-01 1.09594440e+00 1.92212194e-01 9.42949831e-01 -9.67941880e-01 -2.40078777e-01 1.02680326e-01 1.21998870e+00 6.53714389e-02 6.67636752e-01 8.39446709e-02 7.61692226e-01 6.32794619e-01 6.52936161e-01 5.52796423e-01 6.96250677e-01 1.70109719e-01 2.69844949e-01 4.10803735e-01 -2.41277292e-01 -5.52634537e-01 2.64294237e-01 1.67789996e+00 -2.51036704e-01 -7.80380726e-01 -4.80473071e-01 5.85310817e-01 -2.12530923e+00 -1.06149554e+00 -1.59690857e-01 1.70899558e+00 8.52152050e-01 1.24719171e-02 -2.51907855e-01 4.23946828e-02 7.16322482e-01 5.11263251e-01 -5.13723373e-01 -5.41238666e-01 -1.77308783e-01 -6.12822436e-02 2.03533620e-01 1.97714061e-01 -5.19238770e-01 1.00433624e+00 5.64487600e+00 9.20946538e-01 -7.01532841e-01 -3.17850173e-01 1.68645650e-01 -4.19149026e-02 -5.19956231e-01 -3.06368377e-02 -5.42255580e-01 4.50609535e-01 8.84755909e-01 -9.19414461e-01 2.47139841e-01 4.59131658e-01 4.55227077e-01 -3.84479165e-01 -7.46420324e-01 8.50214183e-01 5.82071006e-01 -1.24455190e+00 6.97388172e-01 -4.05277073e-01 8.33279014e-01 -3.77253324e-01 -2.84609646e-01 3.31358671e-01 1.77708909e-01 -7.01233685e-01 3.72812212e-01 9.77150917e-01 3.21762323e-01 -8.79588127e-01 1.05967665e+00 7.59861112e-01 -8.87568593e-01 6.72727749e-02 -3.08632225e-01 -2.08421245e-01 4.07024890e-01 7.09535778e-01 -9.83869135e-01 1.54106712e+00 3.63102138e-01 1.16904354e+00 -7.38564610e-01 1.11974013e+00 -4.37436044e-01 6.45137489e-01 -4.31610420e-02 -4.21732515e-01 1.06967039e-01 -2.94461757e-01 9.60201859e-01 1.36105967e+00 4.89988834e-01 4.15453255e-01 8.54695812e-02 3.17086190e-01 -4.45447177e-01 4.89664555e-01 -7.75574327e-01 -4.13492054e-01 5.70975900e-01 1.16439915e+00 -5.45907974e-01 -6.66964114e-01 -2.15131745e-01 1.05918610e+00 3.27274889e-01 3.53602082e-01 -3.73322845e-01 -6.94116831e-01 -1.70693099e-01 -4.01312888e-01 1.56497955e-01 2.14314796e-02 -6.29385337e-02 -1.22981620e+00 2.33768463e-01 -9.25150394e-01 4.22278047e-01 -9.83636856e-01 -1.09370041e+00 3.04011494e-01 2.34528199e-01 -9.00242865e-01 -1.58667818e-01 7.01846108e-02 -7.44542897e-01 4.40157890e-01 -1.76124275e+00 -8.96993637e-01 -3.62088978e-01 2.00543612e-01 9.62594450e-01 1.10102352e-02 5.09033263e-01 2.31243335e-02 -7.41110682e-01 1.91909373e-01 2.17482463e-01 -2.35125199e-01 7.22105801e-01 -1.21188056e+00 2.95970768e-01 1.10381305e+00 -1.22566232e-02 6.98737144e-01 8.78968060e-01 -1.11723685e+00 -1.67948389e+00 -1.11549711e+00 1.15943444e+00 -1.04457319e-01 3.85817319e-01 3.78553689e-01 -8.26273501e-01 5.06701052e-01 7.25713551e-01 -7.74944484e-01 6.79829299e-01 -2.86778897e-01 8.28258172e-02 -5.76816425e-02 -9.37227786e-01 6.73822105e-01 1.28109753e+00 -5.60638346e-02 -1.16355515e+00 3.37043494e-01 1.15123296e+00 -4.18170035e-01 -6.38172209e-01 2.50287712e-01 3.16142529e-01 -7.80288160e-01 6.19321823e-01 -6.11642957e-01 7.29133308e-01 -1.22920863e-01 3.12993765e-01 -1.91196239e+00 -8.45390782e-02 -7.93177426e-01 -5.68475604e-01 1.65980732e+00 1.37203157e-01 -5.94895363e-01 3.53138804e-01 1.26144662e-01 -5.73767364e-01 -5.46917975e-01 -6.42497301e-01 -4.99400318e-01 -3.98367941e-01 1.14279827e-02 4.09458816e-01 7.18399227e-01 4.50177819e-01 8.57020199e-01 -4.52748805e-01 -1.69607535e-01 3.99381399e-01 -4.94747832e-02 6.90734565e-01 -1.25538337e+00 1.89278066e-01 -3.73820424e-01 -2.27338418e-01 -1.18904197e+00 1.95365679e-02 -1.10151350e+00 -5.58041744e-02 -2.46637893e+00 5.37178874e-01 4.39162135e-01 1.24564050e-02 1.30997464e-01 -6.53964400e-01 -3.53779674e-01 1.89788640e-01 1.82073325e-01 -1.22248638e+00 7.72750556e-01 1.50487542e+00 -9.20937061e-02 -2.69505173e-01 -1.59511775e-01 -1.37866247e+00 6.39576077e-01 8.00275505e-01 -3.95895153e-01 -6.87330782e-01 -4.21884388e-01 5.59068918e-01 3.30626279e-01 9.54136439e-03 -7.32986748e-01 6.10455930e-01 -4.48820181e-02 2.03537792e-02 -8.31905484e-01 -4.24860790e-02 -5.17449915e-01 -4.93102856e-02 2.81672120e-01 -5.68031907e-01 1.76491097e-01 6.99122474e-02 1.06044734e+00 -3.64567041e-01 -2.15627953e-01 5.98972887e-02 -2.58410186e-01 -7.54081488e-01 -1.68062914e-02 -3.30362171e-01 4.51515973e-01 8.45686495e-01 -1.14410922e-01 -8.13917577e-01 -6.98949337e-01 -6.74820170e-02 5.07839561e-01 7.66543373e-02 4.04954970e-01 8.72771204e-01 -1.04836929e+00 -1.01574063e+00 -3.81680459e-01 1.48156166e-01 2.17875436e-01 5.79286158e-01 8.08131039e-01 -4.41478848e-01 6.80784583e-01 -6.28420934e-02 -1.05581962e-01 -1.21269560e+00 5.33572197e-01 -4.54886556e-01 -5.37969828e-01 -7.60735333e-01 4.48133141e-01 -3.90967637e-01 -8.10305253e-02 1.16521955e-01 -3.89672488e-01 -8.43054116e-01 3.28260630e-01 6.96118414e-01 8.80099356e-01 1.13021925e-01 -5.29023170e-01 1.25768268e-03 3.56799185e-01 -2.91911572e-01 1.47061914e-01 1.35424709e+00 -5.18383980e-01 -3.22620064e-01 3.05827737e-01 9.29572344e-01 2.14476034e-01 -6.93232358e-01 -4.29559916e-01 2.23519206e-01 -2.54710585e-01 -1.06254078e-01 -9.16488349e-01 -7.16081262e-01 5.39581656e-01 -3.26326221e-01 5.66171587e-01 1.20434976e+00 -2.42061660e-01 1.23637736e+00 7.06246912e-01 -4.28181291e-02 -1.19671142e+00 1.38406411e-01 4.75663781e-01 9.98576105e-01 -1.10059857e+00 4.14459139e-01 -4.41640139e-01 -9.53659236e-01 1.10681725e+00 3.47389966e-01 1.07232511e-01 -5.25076799e-02 -3.13633710e-01 -2.83294320e-01 -3.29968959e-01 -8.12663376e-01 -1.56699091e-01 3.69307846e-01 5.56969523e-01 2.45467559e-01 -7.15939179e-02 -7.80136287e-01 7.41077721e-01 -3.15973967e-01 8.33852589e-02 8.01697493e-01 9.32783127e-01 -8.20763767e-01 -7.01208115e-01 -9.16718170e-02 9.21195149e-01 -5.41120350e-01 -2.27086931e-01 -8.69487643e-01 4.07213449e-01 -4.87261653e-01 1.35009062e+00 -4.41039652e-01 -3.50976199e-01 5.91755867e-01 1.74217239e-01 2.16548145e-01 -7.47523367e-01 -5.92443168e-01 -3.73165935e-01 4.37640250e-01 -2.81066418e-01 -7.72972047e-01 -4.47903991e-01 -1.18839073e+00 -4.45013255e-01 -3.17286104e-01 3.57116342e-01 4.60610896e-01 1.09676409e+00 7.89440989e-01 9.24036801e-01 4.54076618e-01 -6.15058661e-01 -5.80119789e-01 -1.22032213e+00 -2.27369308e-01 3.16549748e-01 2.33821616e-01 -1.35029867e-01 -3.27821016e-01 1.56523749e-01]
[12.586559295654297, 9.519729614257812]
f78f45fa-17a1-4d03-b111-73db819b2715
does-crowdfunding-really-foster-innovation
2101.02683
null
https://arxiv.org/abs/2101.02683v2
https://arxiv.org/pdf/2101.02683v2.pdf
Does Crowdfunding Really Foster Innovation? Evidence from the Board Game Industry
Crowdfunding offers inventors and entrepreneurs alternative access to resources with which they can develop and realize their ideas. Besides helping to secure capital, crowdfunding also connects creators with engaged early supporters who provide public feedback. But does this process foster truly innovative outcomes? Does the proliferation of crowdfunding in an industry make it more innovative overall? Prior studies investigating the link between crowdfunding and innovation do not compare traditional and crowdfunded products and so while claims that crowdfunding supports innovation are theoretically sound, they lack empirical backing. We address this gap using a unique dataset of board games, an industry with significant crowdfunding activity in recent years. Each game is described by how it combines fundamental mechanisms such as dice-rolling, negotiation, and resource-management, from which we develop quantitative measures of innovation in game design. Using these measures to compare games, we find that crowdfunded games tend to be more distinctive from previous games than their traditionally published counterparts. They are also significantly more likely to implement novel combinations of mechanisms. Crowdfunded games are not just transient experiments: subsequent games imitate their novel ideas. These results hold in regression models controlling for game and designer-level confounders. Our findings demonstrate that the innovative potential of crowdfunding goes beyond individual products to entire industries, as new ideas spill over to traditionally funded products.
['Balazs Vedres', 'Johannes Wachs']
2021-01-07
null
null
null
null
['board-games']
['playing-games']
[-8.42260778e-01 4.97996330e-01 -6.08873606e-01 3.76574278e-01 -2.02362344e-01 -8.06687176e-01 4.48485225e-01 -6.03779964e-02 -3.25156808e-01 5.98199844e-01 7.61262298e-01 -8.71697009e-01 -1.75369456e-01 -9.74655986e-01 -6.62097275e-01 2.41464123e-01 2.23365217e-01 9.06491205e-02 -1.20568626e-01 -4.28385973e-01 6.46514893e-01 -2.52675712e-02 -1.29449201e+00 -9.25831571e-02 8.67952764e-01 6.90529227e-01 2.48915568e-01 2.88968056e-01 -4.17840630e-01 1.01116455e+00 -6.15090013e-01 -8.12916696e-01 9.03014719e-01 -1.88692048e-01 -2.10340649e-01 -6.93563819e-02 6.31797537e-02 -5.07642567e-01 -1.72999159e-01 5.21212280e-01 5.41739643e-01 -9.08308700e-02 3.41326177e-01 -1.02640998e+00 -1.24219036e+00 9.69305336e-01 -6.53433561e-01 5.93956649e-01 4.01451647e-01 8.17055404e-01 1.45002997e+00 -7.85245597e-01 1.09980130e+00 8.60794425e-01 9.11983788e-01 -8.13553575e-03 -9.85281646e-01 -1.10303986e+00 1.00810796e-01 -6.58582389e-01 -6.67874634e-01 -2.99819440e-01 4.04132992e-01 -1.21435630e+00 9.61275220e-01 -2.68106639e-01 1.40651238e+00 6.44655824e-01 2.86766917e-01 1.60643563e-01 7.25659430e-01 -1.87915072e-01 3.41948330e-01 1.06323160e-01 -4.78600860e-02 2.30566695e-01 1.11829698e+00 3.57936472e-01 -6.71060920e-01 -2.52298594e-01 1.48647761e+00 2.82336295e-01 1.59941372e-02 2.12103903e-01 -1.00048363e+00 9.94685590e-01 1.69273347e-01 5.57401896e-01 -9.74767745e-01 4.29591149e-01 1.00549288e-01 5.32465279e-01 5.21856248e-01 1.27165222e+00 -1.60764500e-01 -6.64657891e-01 -5.70212960e-01 6.15313828e-01 7.73695230e-01 6.99454069e-01 8.30166459e-01 -1.84566118e-02 -2.77131468e-01 4.50439125e-01 1.89385161e-01 1.70104653e-01 3.66505533e-01 -1.41932344e+00 5.20007312e-01 6.35241091e-01 4.44931865e-01 -1.14198458e+00 -9.11577269e-02 -9.35591280e-01 2.57527292e-01 4.77193266e-01 6.83019340e-01 -6.76031470e-01 1.21392652e-01 1.41679192e+00 -2.21647546e-01 -1.88769206e-01 -5.95013678e-01 6.36900425e-01 4.37198192e-01 2.92490363e-01 1.61152467e-01 2.41562761e-02 1.07743299e+00 -9.97227430e-01 -4.49349701e-01 -2.60194689e-01 6.22478783e-01 -5.59539378e-01 1.32020223e+00 3.08958024e-01 -1.41469681e+00 -4.48499024e-01 -6.24271989e-01 4.01465446e-01 5.35605215e-02 -7.99410939e-01 1.33054638e+00 1.18285871e+00 -1.29252267e+00 6.70939624e-01 5.21723703e-02 -3.08381349e-01 1.03764808e+00 2.05209494e-01 -9.90594476e-02 1.93027079e-01 -8.48330736e-01 5.41531146e-01 -3.33995163e-01 -7.43896484e-01 -3.03327739e-01 -1.10245156e+00 -2.93624282e-01 1.63052708e-01 4.51117873e-01 -1.13389361e+00 1.47265112e+00 -1.43826318e+00 -1.29384160e+00 3.36107850e-01 5.22690237e-01 -2.46623456e-01 1.42730653e-01 5.66456318e-02 -8.35085213e-02 -2.51604944e-01 9.29784954e-01 2.50629485e-01 4.39533472e-01 -5.77092171e-01 -7.39204884e-01 -2.43734628e-01 6.27145350e-01 -1.04886949e-01 -7.19526291e-01 3.01486671e-01 1.97057724e-01 -9.14463222e-01 -2.81169981e-01 -4.63518351e-01 -4.91922677e-01 -2.86166400e-01 1.42338037e-01 -2.93815821e-01 -2.76763201e-01 -4.45329696e-01 1.12922621e+00 -2.00291944e+00 -6.76371932e-01 1.32307662e-02 7.77871311e-01 -2.72083640e-01 -1.91461593e-02 6.05595767e-01 7.25325048e-02 8.40704739e-01 9.50516999e-01 3.81385721e-02 2.96005458e-01 -4.13022369e-01 -2.69370228e-02 2.67249197e-01 -3.49810384e-02 1.46190155e+00 -7.92048514e-01 1.50508508e-01 -3.82314056e-01 -4.68401551e-01 -1.07932329e+00 -4.37165022e-01 1.17075250e-01 2.28249654e-01 -5.86468339e-01 7.42092133e-01 4.76487696e-01 -5.83080292e-01 -1.69982910e-01 9.78771389e-01 -6.59174919e-01 2.75690347e-01 -8.37712824e-01 1.25835800e+00 -2.57472068e-01 7.04608381e-01 1.90372821e-02 -5.76355994e-01 8.76739264e-01 1.60839960e-01 3.36870015e-01 -1.14603269e+00 3.60189080e-01 4.30791974e-01 1.94592789e-01 -3.41071784e-01 8.75555396e-01 -7.46189713e-01 -8.82797167e-02 1.02207100e+00 -2.34486416e-01 7.80716464e-02 -1.52092591e-01 3.96393329e-01 1.60475230e+00 -2.69047230e-01 4.08079565e-01 -3.61358762e-01 -5.43557167e-01 2.12846965e-01 8.33135188e-01 1.01755905e+00 -3.14583451e-01 2.73095611e-02 7.08103240e-01 -1.35938317e-01 -1.13842535e+00 -8.68989587e-01 3.56052458e-01 1.26374960e+00 1.19320929e-01 -3.28343928e-01 -3.93387794e-01 -1.71993792e-01 7.88604856e-01 3.71239632e-01 -4.51345950e-01 2.60117650e-01 1.22652531e-01 2.40206331e-01 3.26608032e-01 5.51008940e-01 5.41362643e-01 -7.82359481e-01 -3.68372321e-01 3.95637810e-01 2.38001317e-01 -3.84215474e-01 -6.48940086e-01 -5.29704154e-01 -1.08328390e+00 -1.07501090e+00 -1.03490150e+00 -4.36026275e-01 2.85794381e-02 7.51848936e-01 1.29595852e+00 9.29684415e-02 -3.65314819e-03 5.12619376e-01 -3.26505959e-01 -9.49635684e-01 -1.90762177e-01 -1.92416295e-01 8.19091424e-02 -4.07519102e-01 4.75761831e-01 -9.94363010e-01 -5.09383678e-01 4.32071775e-01 -3.82251471e-01 -3.33140403e-01 6.75153077e-01 5.05595267e-01 1.86139699e-02 3.77089605e-02 1.48859370e+00 -7.62980819e-01 1.24224317e+00 -1.19480896e+00 -4.71311361e-01 -4.94907498e-01 -8.89590085e-01 -6.29900992e-01 2.32340060e-02 -5.73370993e-01 -9.54834223e-01 -5.84451258e-01 4.67578560e-01 6.49289340e-02 3.17330033e-01 7.67711401e-01 1.67581961e-01 -2.61516690e-01 1.19626403e+00 -9.19838428e-01 3.52460593e-01 -3.09830546e-01 4.48025346e-01 5.02990961e-01 6.08160682e-02 -6.21728659e-01 7.75409877e-01 4.40782636e-01 -6.99886084e-01 -4.02222902e-01 -1.16896173e-02 -4.88197863e-01 5.60036302e-01 -6.23147428e-01 6.98248386e-01 -1.42955053e+00 -1.16837740e+00 7.95670599e-02 -9.68681753e-01 -4.04293776e-01 -9.05728877e-01 9.05526996e-01 -3.54210496e-01 -7.85641000e-02 -5.99569559e-01 -9.00119126e-01 1.77781865e-01 -5.94241858e-01 1.28683463e-01 6.46475196e-01 -4.15701389e-01 -7.95986414e-01 2.88717806e-01 1.00184751e+00 9.04203296e-01 2.70827442e-01 3.15476090e-01 -4.42917734e-01 -9.87405479e-01 -2.32450396e-01 -8.40691850e-02 -1.91405900e-02 1.67136248e-02 -3.01827669e-01 -5.81194699e-01 3.17460507e-01 3.09963733e-01 2.40466022e-03 5.11278570e-01 7.67243624e-01 1.03252262e-01 -3.83487731e-01 -2.81766176e-01 4.24903817e-02 1.07138717e+00 2.36336559e-01 5.16615450e-01 9.45347548e-01 1.34950623e-01 7.53271818e-01 7.27722645e-01 8.29702795e-01 3.32779527e-01 3.34709257e-01 1.46153346e-01 4.96985205e-02 2.89258599e-01 -3.48280638e-01 3.26652557e-01 3.41023207e-01 -6.44383430e-01 1.72142968e-01 -9.22365904e-01 5.97123504e-01 -1.68766916e+00 -1.24487424e+00 -2.99124092e-01 1.96220076e+00 1.67922720e-01 5.06095111e-01 9.66757715e-01 -2.96458781e-01 9.39580977e-01 -1.59996867e-01 -7.80659616e-01 -2.65269011e-01 -1.87910706e-01 5.73756516e-01 8.82149279e-01 9.91402343e-02 -5.32972366e-02 7.90529728e-01 7.42599869e+00 4.64599609e-01 -6.73295319e-01 7.59319603e-01 8.05606663e-01 -7.68229485e-01 -1.07701457e+00 3.84219140e-01 -2.98164845e-01 1.39041081e-01 6.02835238e-01 -8.18914592e-01 4.40799415e-01 1.09587848e+00 5.72462857e-01 -3.03483218e-01 -5.83393633e-01 8.32291365e-01 -4.01422292e-01 -2.02994704e+00 -6.05561197e-01 9.27700877e-01 1.30849230e+00 -9.69789103e-02 3.43703151e-01 4.24593121e-01 7.65056610e-01 -1.06165278e+00 1.34857094e+00 5.83684027e-01 5.46074688e-01 -7.96591043e-01 4.61744577e-01 2.20258906e-01 -6.66456759e-01 -9.23342586e-01 -2.32968748e-01 -1.64858878e+00 3.28248858e-01 1.04067028e+00 -6.83831096e-01 -1.05433539e-01 3.76342297e-01 5.10160804e-01 -2.95058459e-01 9.93598402e-01 5.42304805e-03 7.02996254e-01 2.96240240e-01 -9.40598994e-02 1.14115961e-01 -1.70101792e-01 5.72624207e-01 1.80634826e-01 4.32767838e-01 8.37505236e-02 -4.92799848e-01 1.48576391e+00 -1.38027564e-01 2.57162422e-01 -1.02614748e+00 -9.59871948e-01 6.77276075e-01 9.08085287e-01 -8.62656832e-01 1.74520969e-01 -6.34742975e-01 3.60995233e-01 3.88754457e-02 3.00415486e-01 -4.96903121e-01 -4.14020211e-01 1.10714447e+00 9.56982672e-01 1.80422142e-01 -3.85987073e-01 -9.54431474e-01 -7.48747110e-01 7.57253021e-02 -7.17347205e-01 -1.87972382e-01 -6.14707053e-01 -1.31311262e+00 -2.68918604e-01 -7.40507483e-01 -7.16038048e-01 -4.80030384e-03 -1.62674651e-01 -9.25373614e-01 1.05525327e+00 -8.44008982e-01 -5.95456183e-01 -1.36075988e-01 1.76561698e-01 1.72404319e-01 -5.17494917e-01 -1.08623140e-01 1.35930419e-01 -1.99909478e-01 5.60035825e-01 -3.82473283e-02 -2.81961173e-01 5.45302749e-01 -6.37031794e-01 1.00597262e+00 6.06137812e-01 -1.14464164e-01 9.71599102e-01 3.18763345e-01 -1.03178251e+00 -1.19779027e+00 -4.11988974e-01 5.78911245e-01 -6.46277428e-01 9.49989319e-01 -2.19850391e-01 -8.97557288e-02 5.27881086e-01 4.12565432e-02 -5.88243008e-01 9.74924088e-01 7.82481015e-01 -2.76498199e-01 1.10708699e-01 -9.49935615e-01 5.89567780e-01 1.68572509e+00 -6.06153727e-01 -2.61235237e-01 2.49286219e-01 1.36740851e+00 4.45668548e-01 -9.80558217e-01 -7.39042640e-01 4.70831960e-01 -1.51818132e+00 7.27079272e-01 -4.63998556e-01 8.61476958e-01 3.28920692e-01 2.00881660e-01 -9.09060180e-01 -1.01062608e+00 -1.27211332e+00 6.24381244e-01 1.35514987e+00 6.04397178e-01 -1.22280300e+00 1.16425562e+00 1.00685477e+00 -3.85704249e-01 -6.56574845e-01 -7.19308734e-01 -1.05747509e+00 3.69093776e-01 -5.22232115e-01 1.12130654e+00 1.18295693e+00 6.97736919e-01 1.95614353e-01 2.44494826e-01 -6.17357492e-01 3.09209526e-01 -3.99110824e-01 1.19123924e+00 -1.62264001e+00 -9.03517842e-01 -1.07465577e+00 -6.23790145e-01 -5.76439679e-01 -5.73776588e-02 -7.51014173e-01 -5.93550146e-01 -1.68744242e+00 1.19963199e-01 -8.37005258e-01 2.71037668e-01 1.91279024e-01 1.08749710e-01 -2.76641920e-02 4.19172376e-01 4.99381781e-01 -3.36662352e-01 3.98551710e-02 1.54960227e+00 2.18495265e-01 -7.67277539e-01 -3.92188206e-02 -2.13051367e+00 2.93575734e-01 6.81354165e-01 -6.21217676e-02 -2.86075026e-01 6.85790107e-02 1.14082122e+00 2.50149518e-01 2.28175879e-01 -8.44560683e-01 1.29482865e-01 -5.28082132e-01 -1.88568339e-01 4.75033969e-01 -3.81464422e-01 -6.48108423e-02 7.11214066e-01 3.23927939e-01 1.52380452e-01 1.90898404e-02 4.13082801e-02 7.23056316e-01 2.32093588e-01 -2.13623762e-01 -2.03040894e-02 -1.50712892e-01 -1.32730603e-01 -6.11323901e-02 -9.97492433e-01 2.48791009e-01 1.09708178e+00 -1.02136254e+00 -4.99574572e-01 -8.39037418e-01 -1.44019902e-01 -2.34951720e-01 1.12242186e+00 2.61397928e-01 -1.07318863e-01 -1.35342801e+00 -5.30393600e-01 -2.38484725e-01 1.31949410e-01 -3.24323535e-01 1.75583184e-01 6.58614218e-01 -5.11605680e-01 2.07348973e-01 -4.03927505e-01 1.48201004e-01 -2.48483106e-01 4.73617435e-01 5.50634339e-02 -3.77725326e-02 -3.47876430e-01 1.12463963e+00 5.50155044e-01 1.73761714e-02 -2.81900316e-01 -2.78638154e-01 -4.54415157e-02 1.19461507e-01 6.22511446e-01 8.20432842e-01 -3.97245795e-01 -1.62172690e-01 9.81250852e-02 -7.20136836e-02 1.96773604e-01 -5.48878074e-01 1.37371898e+00 -7.71519169e-02 -1.26156121e-01 2.66085654e-01 2.65197396e-01 4.34926361e-01 -1.21606910e+00 2.49096509e-02 -5.31753190e-02 -1.06868505e+00 8.19325820e-03 -4.00883079e-01 -1.13815761e+00 1.81177899e-01 -1.59300026e-02 4.21933770e-01 5.16805947e-01 1.65544972e-01 6.47431612e-01 -2.37208102e-02 7.92021513e-01 -1.35544670e+00 7.72793293e-01 2.03699455e-01 6.36570752e-01 -5.04073679e-01 -3.07605922e-01 -2.59520471e-01 -5.09761214e-01 4.74682063e-01 4.36333418e-01 -2.43243679e-01 9.26412404e-01 -1.74258836e-04 -4.35072482e-01 -5.71363509e-01 -5.15865505e-01 -2.39763428e-02 -6.32965744e-01 8.07840168e-01 3.81538957e-01 3.97500843e-01 -8.10145736e-01 1.73190665e+00 -4.14292932e-01 5.09140015e-01 1.02236640e+00 7.76904166e-01 -9.12145913e-01 -9.39944088e-01 -6.89967692e-01 1.13901305e+00 -6.76268220e-01 -2.17508852e-01 -5.15505016e-01 6.45012915e-01 4.57536519e-01 1.25070405e+00 3.99929374e-01 -6.27134323e-01 3.44101191e-01 -3.94474089e-01 6.82771504e-02 -9.00972009e-01 -1.23222148e+00 -5.93281686e-02 2.65132248e-01 -5.38085580e-01 -1.76117178e-02 -1.01914012e+00 -9.00110841e-01 -9.39161241e-01 -6.79447591e-01 1.90051466e-01 6.41818404e-01 6.52795315e-01 8.29191029e-01 4.02293056e-01 4.86119717e-01 -7.51722217e-01 -3.50348175e-01 -7.67860234e-01 -1.51414800e+00 -1.25941053e-01 -4.98065561e-01 -9.04991746e-01 -5.11865437e-01 -4.89444137e-01]
[9.134956359863281, 6.195166110992432]
d96e354b-e18d-4b9f-9f02-6741bf3551e0
hierarchical-boundary-aware-neural-encoder
1611.09312
null
http://arxiv.org/abs/1611.09312v3
http://arxiv.org/pdf/1611.09312v3.pdf
Hierarchical Boundary-Aware Neural Encoder for Video Captioning
The use of Recurrent Neural Networks for video captioning has recently gained a lot of attention, since they can be used both to encode the input video and to generate the corresponding description. In this paper, we present a recurrent video encoding scheme which can discover and leverage the hierarchical structure of the video. Unlike the classical encoder-decoder approach, in which a video is encoded continuously by a recurrent layer, we propose a novel LSTM cell, which can identify discontinuity points between frames or segments and modify the temporal connections of the encoding layer accordingly. We evaluate our approach on three large-scale datasets: the Montreal Video Annotation dataset, the MPII Movie Description dataset and the Microsoft Video Description Corpus. Experiments show that our approach can discover appropriate hierarchical representations of input videos and improve the state of the art results on movie description datasets.
['Lorenzo Baraldi', 'Rita Cucchiara', 'Costantino Grana']
2016-11-28
hierarchical-boundary-aware-neural-encoder-1
http://openaccess.thecvf.com/content_cvpr_2017/html/Baraldi_Hierarchical_Boundary-Aware_Neural_CVPR_2017_paper.html
http://openaccess.thecvf.com/content_cvpr_2017/papers/Baraldi_Hierarchical_Boundary-Aware_Neural_CVPR_2017_paper.pdf
cvpr-2017-7
['video-description']
['computer-vision']
[ 4.95230079e-01 2.91507822e-02 -4.86457914e-01 -3.34662884e-01 -6.91055179e-01 -5.24973333e-01 6.69878900e-01 -7.53329545e-02 -1.10161655e-01 6.98408127e-01 7.05036044e-01 -1.18310342e-03 2.60961086e-01 -4.93024796e-01 -1.08542728e+00 -5.22355974e-01 -7.59928152e-02 8.28947946e-02 3.17506224e-01 5.22110499e-02 2.73949772e-01 9.53711383e-03 -1.59322536e+00 1.05179441e+00 1.36088267e-01 1.02115393e+00 4.14895982e-01 7.30117321e-01 -2.35490680e-01 1.53875983e+00 -3.62811327e-01 -2.79415667e-01 -2.65390843e-01 -8.10784578e-01 -9.94430184e-01 2.07514063e-01 3.21471423e-01 -4.36031103e-01 -1.00012028e+00 7.65414715e-01 3.30450267e-01 3.79851073e-01 4.87866491e-01 -1.01080823e+00 -7.95942366e-01 1.09994328e+00 -1.54602572e-01 3.67813349e-01 7.50872672e-01 -1.87097475e-01 1.14950526e+00 -6.14352226e-01 1.00254250e+00 9.51953769e-01 5.56664050e-01 8.40981722e-01 -9.71426010e-01 -5.04514337e-01 2.65898407e-01 7.53751099e-01 -1.31927538e+00 -6.91249907e-01 7.48941123e-01 -4.77201253e-01 1.17278683e+00 -7.86451399e-02 7.72683740e-01 1.67463851e+00 -3.53021920e-02 1.11464763e+00 2.48375773e-01 -2.65727013e-01 -1.23402150e-03 -1.99324667e-01 -1.98760062e-01 5.94077885e-01 -4.87124979e-01 -9.50855911e-02 -6.14680946e-01 3.04245085e-01 1.04037738e+00 3.35096344e-02 -4.46471423e-01 -2.28557542e-01 -1.25524318e+00 8.34893405e-01 2.68555254e-01 4.52687263e-01 -4.69419718e-01 5.37643135e-01 7.74749219e-01 1.48585498e-01 2.86558777e-01 3.06075007e-01 -8.15403834e-02 -5.81552982e-01 -1.09078944e+00 2.12322399e-01 6.18081391e-01 1.13356876e+00 2.61833966e-01 8.32995474e-02 -6.44864857e-01 8.23611736e-01 6.94904774e-02 -1.93750054e-01 8.53983045e-01 -1.20399249e+00 6.84436202e-01 2.88941234e-01 1.21961301e-02 -9.70459878e-01 -1.97082490e-01 1.15531482e-01 -7.73998499e-01 -7.67531097e-01 -4.48834971e-02 5.75353652e-02 -7.91238606e-01 1.69287682e+00 -3.37826699e-01 8.08968246e-01 2.12037876e-01 8.36186409e-01 1.12947965e+00 1.19827533e+00 4.80500748e-03 -2.83014178e-01 9.58421469e-01 -1.18118262e+00 -9.67258155e-01 8.99173878e-03 4.97217327e-01 -3.44123811e-01 6.83812261e-01 -4.28513661e-02 -1.46771216e+00 -6.45078063e-01 -9.23945308e-01 -1.34049281e-01 -3.87101695e-02 -2.17712503e-02 3.62425029e-01 -2.71262169e-01 -1.19845438e+00 6.44816875e-01 -7.42557108e-01 -2.94926226e-01 2.72991627e-01 7.50772133e-02 -3.49538714e-01 -1.43693075e-01 -1.30839956e+00 4.17288065e-01 8.46634805e-01 1.48506597e-01 -1.20602691e+00 -4.64608103e-01 -1.19942892e+00 3.64480257e-01 2.26947442e-01 -6.19272888e-01 1.38678181e+00 -1.29789364e+00 -1.56947196e+00 6.58088803e-01 -2.95035660e-01 -8.42579007e-01 1.75958887e-01 -1.32371001e-02 -3.28332096e-01 5.81660748e-01 -1.61250144e-01 1.24414885e+00 8.60200286e-01 -1.02172375e+00 -6.80902302e-01 2.28391767e-01 3.82464617e-01 1.06291048e-01 -2.39681289e-01 8.09523091e-02 -8.41386914e-01 -8.76071036e-01 -1.54649258e-01 -9.90262330e-01 1.40000712e-02 -3.96253467e-01 -4.72347170e-01 -1.76120535e-01 7.76997924e-01 -7.14497447e-01 1.51739597e+00 -2.25229001e+00 6.75773919e-01 -2.52780825e-01 -2.37535089e-02 5.01936190e-02 -2.63571113e-01 4.91622597e-01 -1.57131061e-01 2.30944812e-01 -3.58260542e-01 -5.04736960e-01 -7.44006410e-02 4.83516932e-01 -6.72658920e-01 1.44374475e-01 2.54426926e-01 1.07054019e+00 -9.53315079e-01 -4.96962577e-01 -1.19579554e-01 8.11576724e-01 -7.58234143e-01 5.59012234e-01 -4.18087661e-01 4.85752374e-01 -1.89920858e-01 3.98421794e-01 -1.77776679e-01 -4.74806905e-01 3.21947277e-01 -2.64556319e-01 -1.15275875e-01 4.99094665e-01 -5.75360835e-01 2.18713260e+00 -3.35200101e-01 1.14146042e+00 -4.77601439e-01 -1.04086685e+00 7.23323703e-01 9.47859168e-01 5.21512389e-01 -7.89522648e-01 -1.31303053e-02 -1.34169325e-01 -4.09679919e-01 -7.92484343e-01 7.39465117e-01 -1.70223769e-02 -2.59715676e-01 3.25246751e-01 2.00546771e-01 3.05107743e-01 4.45551813e-01 2.64745176e-01 1.06558573e+00 5.15737057e-01 9.01995823e-02 4.08805400e-01 5.30409873e-01 -1.42054155e-01 5.47366142e-01 5.87174177e-01 2.43814886e-02 1.05185437e+00 6.65950000e-01 -4.46075112e-01 -1.19143665e+00 -7.54879415e-01 1.43280447e-01 1.05400062e+00 9.07721594e-02 -5.96719563e-01 -6.96200967e-01 -5.29438317e-01 -4.64507192e-01 5.71116447e-01 -7.33334541e-01 -3.10061127e-01 -8.18045914e-01 -1.81948602e-01 5.99307954e-01 7.61434376e-01 5.53331017e-01 -1.46656597e+00 -5.69292009e-01 4.44488794e-01 -7.54364550e-01 -1.55749106e+00 -7.03934371e-01 -1.31388590e-01 -7.88884342e-01 -8.10081840e-01 -7.47122705e-01 -1.16705835e+00 4.97173816e-01 7.40743652e-02 1.18249011e+00 3.01537197e-02 1.04623429e-01 3.77619565e-01 -7.22986162e-01 4.57195878e-01 -6.06578112e-01 4.12647188e-01 -2.53219366e-01 1.04766868e-01 3.45990658e-02 -5.11419296e-01 -3.46426725e-01 7.15413392e-02 -1.26910770e+00 5.28279066e-01 2.17615515e-01 7.38346696e-01 6.65811956e-01 -2.81365275e-01 5.96223295e-01 -8.35080147e-01 3.90760124e-01 -7.52449274e-01 -3.95407706e-01 1.95166707e-01 -4.37825266e-03 1.89711243e-01 6.78054571e-01 -4.24826294e-01 -9.77069736e-01 2.87873417e-01 -1.72674015e-01 -8.92614245e-01 -8.75093192e-02 7.75181711e-01 2.16137901e-01 2.88968712e-01 3.60610560e-02 3.92648637e-01 -3.58020544e-01 -5.38237810e-01 3.27875495e-01 5.86506188e-01 9.59536433e-01 -2.43892848e-01 3.39242399e-01 2.01839626e-01 -2.33519778e-01 -5.31568050e-01 -8.24891627e-01 -3.36648315e-01 -9.09761548e-01 -2.25865468e-01 1.07672095e+00 -9.87911701e-01 -4.60911572e-01 1.14391617e-01 -1.62124157e+00 -3.12060714e-01 -2.16878861e-01 3.84070694e-01 -9.52376783e-01 2.02976868e-01 -9.43835795e-01 -3.15781325e-01 -7.09852530e-03 -1.34521723e+00 9.33747232e-01 1.44342542e-01 -2.90283531e-01 -1.13885999e+00 1.79970264e-01 1.64954290e-01 2.49425218e-01 3.80289286e-01 9.23193514e-01 -3.77655298e-01 -7.34565556e-01 -5.35170771e-02 -9.66124684e-02 5.27421534e-02 1.80001352e-02 1.28091946e-01 -7.88367689e-01 -2.83974141e-01 -2.98950553e-01 -4.95608717e-01 1.08057523e+00 3.90497148e-01 1.41529679e+00 -6.60451651e-01 -2.03369856e-01 8.58275175e-01 1.28412271e+00 4.48779225e-01 1.03194046e+00 4.64561552e-01 8.05027068e-01 5.18472195e-01 1.85479298e-01 6.19124115e-01 5.83928287e-01 9.55539584e-01 4.13530201e-01 2.70255595e-01 -1.65492430e-01 -6.55898869e-01 7.62125611e-01 1.27117109e+00 -1.26074553e-01 -4.86469090e-01 -7.26314545e-01 6.83562934e-01 -2.17562819e+00 -1.38369834e+00 3.38338077e-01 1.77017725e+00 7.89409220e-01 -1.03317402e-01 1.38169393e-01 -1.08933344e-01 8.37167561e-01 3.43696356e-01 -4.16116178e-01 -5.27103722e-01 -5.70415147e-02 -2.63961434e-01 1.77547380e-01 3.02337915e-01 -1.19198871e+00 9.83989477e-01 7.17604160e+00 4.48510677e-01 -1.17832971e+00 5.69727793e-02 5.18522263e-01 -3.95370752e-01 -2.47192904e-01 -2.63171196e-01 -6.98470712e-01 6.12159014e-01 1.48317659e+00 -1.86112076e-01 4.56537098e-01 5.50785780e-01 2.49913454e-01 3.87876272e-01 -1.38908339e+00 1.08718777e+00 4.04830575e-01 -1.88895273e+00 4.90193665e-01 -2.99641907e-01 8.09485912e-01 -1.32559150e-01 3.80704775e-02 4.52545524e-01 -1.20170154e-01 -1.04857540e+00 8.60138237e-01 7.25741208e-01 1.01819360e+00 -6.72753751e-01 6.57793999e-01 -3.53183527e-03 -1.38384795e+00 -1.02245748e-01 -2.81196237e-01 -8.51194654e-03 5.15612662e-01 -1.54917613e-01 -7.24969447e-01 2.76255846e-01 7.14070499e-01 1.46620512e+00 -4.78983402e-01 1.05118918e+00 -2.76836246e-01 6.37513995e-01 2.90840238e-01 2.72547722e-01 6.28157556e-01 -1.15516363e-02 4.38904464e-01 1.40352607e+00 3.64914984e-01 1.20418951e-01 5.55762388e-02 7.89042711e-01 -4.80415702e-01 -2.14145526e-01 -8.01850736e-01 -4.63125885e-01 3.20154548e-01 8.27300370e-01 -6.15501523e-01 -5.23361564e-01 -5.53177595e-01 1.24182415e+00 2.47472599e-01 6.28272474e-01 -1.21916759e+00 -2.42935047e-01 3.76706183e-01 5.34836343e-03 8.85598779e-01 -2.45981842e-01 4.19132739e-01 -1.43493891e+00 -7.49305487e-02 -7.37646878e-01 3.98293585e-01 -1.06624532e+00 -7.71788657e-01 9.07793343e-01 1.89000160e-01 -1.42115188e+00 -9.76189137e-01 -1.71451643e-01 -4.11011308e-01 1.87565953e-01 -1.53288543e+00 -9.06537354e-01 -2.53488809e-01 5.09709001e-01 1.00707698e+00 -2.82171190e-01 7.28738844e-01 5.32924056e-01 -7.28078246e-01 3.75507712e-01 1.24929816e-01 3.93824756e-01 3.84025007e-01 -8.38548601e-01 4.92956311e-01 6.94817364e-01 4.61276054e-01 3.77944589e-01 5.78416407e-01 -4.17544365e-01 -1.29154837e+00 -1.19624376e+00 9.95467603e-01 -1.86804950e-01 6.67503893e-01 -3.26771051e-01 -1.03070509e+00 1.14420402e+00 5.10633707e-01 -8.34208131e-02 6.75426960e-01 -5.12437820e-01 -3.18138868e-01 1.99529096e-01 -5.80578983e-01 4.94278610e-01 1.23426664e+00 -8.53047669e-01 -6.98022366e-01 1.83911696e-01 1.13418448e+00 -5.08007646e-01 -8.94134164e-01 3.16890359e-01 6.64619803e-01 -8.24029505e-01 9.12966907e-01 -8.25360298e-01 1.07671690e+00 -7.09689260e-02 -2.62825102e-01 -1.14951718e+00 -3.79295170e-01 -6.91483974e-01 -3.97921652e-01 1.36943567e+00 4.62986350e-01 1.94183007e-01 6.14451647e-01 3.86045426e-01 -4.20820475e-01 -6.20728016e-01 -1.01767874e+00 -4.38872933e-01 -4.17920202e-01 -2.78219074e-01 4.01235431e-01 8.43294740e-01 7.76693225e-02 4.08681303e-01 -7.10938215e-01 -2.66777098e-01 3.82884629e-02 1.73730910e-01 2.32859552e-01 -6.80926323e-01 -4.10065293e-01 -3.73832852e-01 -5.04314005e-01 -1.36802542e+00 7.55290270e-01 -1.07242715e+00 2.89896399e-01 -1.67139089e+00 3.04145336e-01 1.46031991e-01 -4.12613183e-01 6.45422757e-01 2.44526684e-01 4.62298781e-01 3.72682661e-01 4.78301227e-01 -1.06189537e+00 7.37305820e-01 9.47412252e-01 -3.95592660e-01 -1.75317541e-01 -3.60403776e-01 -3.29874605e-01 5.19716680e-01 4.63603884e-01 -4.65787411e-01 -4.01376247e-01 -8.13826680e-01 3.73790890e-01 5.58629572e-01 2.06065297e-01 -1.00510669e+00 2.95301646e-01 1.03531793e-01 2.55793512e-01 -3.99787664e-01 4.32725102e-01 -6.80076241e-01 4.02765751e-01 1.86700821e-01 -9.87796724e-01 3.90998006e-01 2.69684285e-01 5.63610137e-01 -7.34504759e-01 -3.33687931e-01 4.69180465e-01 -2.17559516e-01 -1.12076402e+00 3.78091753e-01 -6.34061038e-01 -6.17745519e-02 8.80708814e-01 -1.79934174e-01 -2.55734444e-01 -7.16899514e-01 -8.05245161e-01 1.59494147e-01 3.91804993e-01 8.44572306e-01 9.87023771e-01 -1.69510937e+00 -6.53594434e-01 1.63274277e-02 7.67103359e-02 -2.94490129e-01 1.73015505e-01 6.08602405e-01 -3.92780632e-01 6.49848580e-01 -3.74580741e-01 -5.18663883e-01 -1.01619661e+00 6.49877548e-01 2.33898491e-01 -3.99911851e-02 -8.83684278e-01 6.80129886e-01 1.46129176e-01 4.19999182e-01 4.39681411e-01 -3.79985541e-01 -7.05762029e-01 2.67780364e-01 7.22788155e-01 3.95873524e-02 -4.37292069e-01 -1.03424859e+00 -6.07686676e-02 4.51550126e-01 -1.69175461e-01 -5.47971502e-02 1.43558395e+00 -3.04643005e-01 -1.08862489e-01 8.46184254e-01 1.54898131e+00 -6.08424366e-01 -1.49300015e+00 -7.39231110e-02 2.05567792e-01 -2.33256117e-01 -6.92427829e-02 -2.30445743e-01 -1.25283551e+00 8.06167483e-01 1.47127882e-01 3.16409618e-01 1.01992249e+00 1.06925100e-01 1.26570129e+00 3.71216834e-01 1.35001078e-01 -8.49800408e-01 2.57793665e-01 7.45560706e-01 9.36857283e-01 -9.25089180e-01 -4.56568658e-01 1.08092604e-02 -8.16403329e-01 1.30634356e+00 3.97066057e-01 -9.16279256e-02 1.20711543e-01 6.82286918e-02 -2.71056801e-01 -7.44697824e-02 -1.22350347e+00 -1.15389787e-01 2.92292863e-01 2.84278095e-01 6.76763117e-01 -3.69027048e-01 3.06149367e-02 5.99279642e-01 8.26399028e-02 3.09299797e-01 8.65866959e-01 7.33054817e-01 -2.28890315e-01 -1.02594125e+00 1.07019469e-01 2.25691423e-01 -4.23833817e-01 -2.08327875e-01 -3.27890873e-01 5.24874628e-01 -1.56477898e-01 6.30581617e-01 5.51388025e-01 -5.08157670e-01 2.53470719e-01 2.24054769e-01 3.96583378e-01 -7.77529597e-01 -5.42654455e-01 2.01560501e-02 1.24441102e-01 -8.32861364e-01 -9.24056709e-01 -6.02333784e-01 -1.39232528e+00 7.51948655e-02 2.50757664e-01 3.37479591e-01 4.03074026e-01 9.15212810e-01 4.69776213e-01 8.01446199e-01 5.86564243e-01 -1.06345379e+00 7.26454332e-02 -8.19680452e-01 -3.03026080e-01 6.17334008e-01 4.94428337e-01 -4.57832992e-01 -1.85955673e-01 7.87664235e-01]
[10.58216381072998, 0.6662557125091553]
0f7a680d-1a46-41c1-9377-244253d51459
incremental-learning-for-animal-pose
2110.13598
null
https://arxiv.org/abs/2110.13598v1
https://arxiv.org/pdf/2110.13598v1.pdf
Incremental Learning for Animal Pose Estimation using RBF k-DPP
Pose estimation is the task of locating keypoints for an object of interest in an image. Animal Pose estimation is more challenging than estimating human pose due to high inter and intra class variability in animals. Existing works solve this problem for a fixed set of predefined animal categories. Models trained on such sets usually do not work well with new animal categories. Retraining the model on new categories makes the model overfit and leads to catastrophic forgetting. Thus, in this work, we propose a novel problem of "Incremental Learning for Animal Pose Estimation". Our method uses an exemplar memory, sampled using Determinantal Point Processes (DPP) to continually adapt to new animal categories without forgetting the old ones. We further propose a new variant of k-DPP that uses RBF kernel (termed as "RBF k-DPP") which gives more gain in performance over traditional k-DPP. Due to memory constraints, the limited number of exemplars along with new class data can lead to class imbalance. We mitigate it by performing image warping as an augmentation technique. This helps in crafting diverse poses, which reduces overfitting and yields further improvement in performance. The efficacy of our proposed approach is demonstrated via extensive experiments and ablations where we obtain significant improvements over state-of-the-art baseline methods.
['Anirban Chakraborty', 'Het Shah', 'Gaurav Kumar Nayak']
2021-10-26
null
null
null
null
['animal-pose-estimation']
['computer-vision']
[ 2.06671268e-01 -2.13041410e-01 2.72792391e-02 -2.19982564e-01 -4.27217454e-01 -5.07959604e-01 3.49500656e-01 1.53708160e-01 -5.40664554e-01 7.31708229e-01 -1.87940776e-01 3.19338620e-01 -8.62873867e-02 -5.47810793e-01 -1.11799073e+00 -5.98275661e-01 -1.06436260e-01 4.67436463e-01 5.12900949e-01 -1.27097175e-01 3.61365855e-01 5.65568089e-01 -1.52745986e+00 -4.25301641e-02 6.91044807e-01 5.89960337e-01 3.33274901e-01 3.86380434e-01 6.72636107e-02 3.88405889e-01 -5.52700222e-01 -4.55599785e-01 4.63472366e-01 -1.84785441e-01 -6.77891791e-01 1.60664678e-01 7.00173438e-01 -3.37538302e-01 -2.20266566e-01 7.49670148e-01 3.69740456e-01 3.98737907e-01 4.87342149e-01 -1.39313293e+00 -5.73433697e-01 2.03124180e-01 -1.03363264e+00 2.00454012e-01 7.08866715e-02 1.92298397e-01 5.19133091e-01 -1.11264575e+00 4.60538894e-01 1.26563013e+00 1.25614858e+00 5.91368973e-01 -1.68600631e+00 -8.49253356e-01 3.36160839e-01 3.54345888e-01 -1.41830397e+00 -2.42738903e-01 8.77521098e-01 -2.88411736e-01 6.19878352e-01 1.42439604e-01 9.89539444e-01 9.92914796e-01 3.87034804e-01 7.03172505e-01 1.09858298e+00 -1.41913012e-01 2.32367292e-01 -8.20034519e-02 3.03481594e-02 5.61645150e-01 3.78796041e-01 2.88614072e-02 -7.48053074e-01 -2.74264723e-01 8.69860590e-01 3.64690214e-01 -8.34799185e-02 -1.08629274e+00 -1.42378998e+00 7.40239739e-01 7.84696519e-01 -2.51264781e-01 -3.41579586e-01 2.46168211e-01 2.20653281e-01 2.12629750e-01 5.02518058e-01 7.97380269e-01 -6.63970232e-01 1.84218138e-01 -9.88725007e-01 6.18655086e-01 4.94908810e-01 8.77420902e-01 7.11379588e-01 -2.89236605e-01 -1.19542554e-01 1.12835836e+00 2.24169925e-01 4.55108494e-01 5.00501335e-01 -6.87796295e-01 4.70016561e-02 6.14544570e-01 9.80121270e-02 -1.16744781e+00 -5.78980446e-01 -7.89136648e-01 -6.79466307e-01 9.85604376e-02 2.51673460e-01 2.03834593e-01 -1.39476299e+00 1.94269967e+00 6.85810804e-01 5.32704651e-01 -2.00201243e-01 6.34428084e-01 6.01083755e-01 7.17248261e-01 1.86793402e-01 -1.44372717e-01 1.12348139e+00 -1.16444576e+00 -3.21386337e-01 -6.15904808e-01 9.23958793e-02 -6.86549127e-01 1.01824510e+00 2.85732925e-01 -7.32381701e-01 -5.91132581e-01 -1.03914404e+00 1.80796266e-01 -3.09502065e-01 -4.43776920e-02 7.03317702e-01 2.49437869e-01 -7.78157353e-01 5.71244240e-01 -9.80984032e-01 -3.08841616e-01 5.39853871e-01 5.57242095e-01 -4.00091976e-01 9.79553536e-02 -7.42923677e-01 9.86490965e-01 1.87203363e-01 1.86045811e-01 -8.73863399e-01 -9.96063709e-01 -7.27682769e-01 -4.02369887e-01 5.39339602e-01 -6.80259705e-01 1.16279757e+00 -7.02702940e-01 -1.36326277e+00 7.41635978e-01 -1.95533186e-02 -5.87145030e-01 5.97851455e-01 -6.58885837e-01 -6.28309920e-02 -1.24233114e-02 1.92473844e-01 1.28150523e+00 1.37098026e+00 -1.33571970e+00 -5.09757102e-01 -5.01434565e-01 -9.73649770e-02 2.08490059e-01 -2.44590119e-01 -4.75517809e-01 -4.45193589e-01 -1.09690964e+00 5.28363466e-01 -1.54957223e+00 -2.92029709e-01 2.73060411e-01 -1.86066497e-02 -4.24527861e-02 9.25404429e-01 -6.32707536e-01 1.17754018e+00 -1.89466798e+00 3.50329995e-01 -6.03542738e-02 2.04017878e-01 1.91846132e-01 -2.35695258e-01 1.43865868e-01 -1.51268587e-01 -3.79018068e-01 -3.63689482e-01 -2.42324576e-01 -4.14599717e-01 2.90003538e-01 -2.92948663e-01 5.45208037e-01 3.21805388e-01 8.92973721e-01 -8.11638236e-01 -3.42198879e-01 5.27951270e-02 4.22008574e-01 -8.58502865e-01 5.34119196e-02 -1.10273212e-01 5.06873071e-01 7.52914185e-03 7.00011492e-01 9.58991706e-01 -1.47115767e-01 -2.73719668e-01 -3.41457605e-01 9.03969258e-02 -2.34489903e-01 -1.17946076e+00 1.74067307e+00 -2.25071654e-01 2.35638425e-01 -5.41375220e-01 -9.26633596e-01 8.88008893e-01 -2.46608585e-01 2.69311875e-01 -2.23597541e-01 -2.55813777e-01 -3.27486359e-02 -3.30793560e-02 -1.91737071e-01 5.31085968e-01 -1.05051592e-01 -2.32765362e-01 1.29663981e-02 2.08131805e-01 -2.12955892e-01 9.92764756e-02 -9.69896242e-02 9.32354152e-01 6.63161099e-01 3.46033007e-01 -2.28215709e-01 1.88108876e-01 1.96369410e-01 6.67719066e-01 7.15643406e-01 -2.73170918e-01 6.10366106e-01 9.22216177e-02 -7.02989995e-01 -1.18212914e+00 -1.18828559e+00 -2.15310290e-01 1.17208695e+00 3.71053606e-01 -1.88554049e-01 -6.08363509e-01 -7.61495709e-01 3.55966568e-01 5.72022080e-01 -8.59339297e-01 -5.70594072e-01 -7.55638063e-01 -9.65758085e-01 2.20845297e-01 6.19151354e-01 6.36210203e-01 -9.50920045e-01 -8.37193489e-01 3.12433511e-01 -2.79180557e-01 -6.61710322e-01 -4.52382177e-01 2.46828303e-01 -1.14074326e+00 -7.23562598e-01 -9.84720111e-01 -8.91025603e-01 9.01255071e-01 4.47523296e-01 8.13460052e-01 -2.56188482e-01 -4.69971299e-01 2.19192594e-01 -2.93289542e-01 -5.58198929e-01 -1.50525153e-01 1.41178623e-01 3.54930788e-01 -2.12730855e-01 2.89663583e-01 -4.57344800e-01 -8.56180429e-01 4.59271222e-01 -8.62701416e-01 1.71693057e-01 8.31994891e-01 1.05851614e+00 8.31147254e-01 -1.35438845e-01 6.61684394e-01 -7.72276640e-01 3.01306814e-01 -4.30036485e-01 -4.80698317e-01 2.37197652e-01 -4.01010185e-01 3.78877074e-01 4.93311763e-01 -1.04291642e+00 -8.80441248e-01 2.70175755e-01 2.87402093e-01 -5.20664692e-01 1.56889349e-01 3.06232661e-01 2.37291515e-01 -4.83467638e-01 6.72328234e-01 5.33069372e-02 2.39520967e-01 -6.20553851e-01 1.13065377e-01 1.59884423e-01 6.40206099e-01 -4.10680354e-01 9.89803374e-01 3.56318235e-01 2.08341390e-01 -6.35093868e-01 -8.91341925e-01 -4.32422727e-01 -7.21883416e-01 -1.99158192e-01 3.67421746e-01 -9.17881489e-01 -4.08354908e-01 7.71907210e-01 -7.89630353e-01 -1.80014014e-01 -2.51854122e-01 4.31898206e-01 -5.92619240e-01 3.28696340e-01 -3.61470819e-01 -5.16607106e-01 -2.82192767e-01 -9.03922141e-01 1.05216944e+00 1.88758671e-01 -3.47266048e-01 -5.10217249e-01 2.36451820e-01 2.66277224e-01 4.81272936e-01 3.18946660e-01 6.75674736e-01 -4.07708526e-01 -4.19630051e-01 -3.69641334e-01 9.75776538e-02 2.48592138e-01 2.74525099e-02 -3.09954971e-01 -5.37120879e-01 -7.21060455e-01 -1.54601159e-02 -3.17395866e-01 8.59842896e-01 2.86072999e-01 1.15933359e+00 -1.60266295e-01 -5.67021251e-01 5.02381980e-01 1.24956095e+00 3.51437144e-02 5.37645817e-01 5.98714232e-01 7.22445428e-01 5.34304678e-01 1.08687913e+00 3.78307074e-01 2.96429038e-01 8.10711205e-01 2.44454041e-01 5.94417565e-02 -1.11679792e-01 -4.87869859e-01 1.35527104e-01 6.09707594e-01 2.20380705e-02 1.73184603e-01 -9.73912954e-01 6.41704440e-01 -1.91165435e+00 -7.35102296e-01 3.34049046e-01 2.40220428e+00 9.00768280e-01 7.26404265e-02 1.38534024e-01 -2.76151542e-02 7.16185391e-01 -2.12658569e-01 -9.01014984e-01 1.39003932e-01 1.99005857e-01 2.86053330e-01 6.87882662e-01 6.80179521e-02 -1.35343683e+00 9.61171865e-01 6.24844027e+00 7.54055798e-01 -1.10232270e+00 -5.11079244e-02 5.81762969e-01 -1.72908321e-01 3.38575989e-01 1.36179673e-02 -8.92470419e-01 4.68642563e-01 3.93873751e-01 1.77403521e-02 3.24823439e-01 1.03127158e+00 -1.68246761e-01 -3.20023328e-01 -1.22128057e+00 1.11474848e+00 3.61706764e-01 -8.66388559e-01 1.01429857e-01 -2.17449844e-01 7.33853698e-01 -3.23900968e-01 2.92957127e-01 4.50505376e-01 3.32824253e-02 -7.88192511e-01 8.09092641e-01 5.62608719e-01 5.71781278e-01 -8.95637691e-01 5.12323856e-01 3.60896558e-01 -1.16860235e+00 -1.52988717e-01 -5.42613387e-01 1.82558864e-01 -1.54696465e-01 2.13323817e-01 -8.52887928e-01 9.85082984e-02 9.41547394e-01 4.72430497e-01 -1.04264581e+00 1.44765186e+00 1.97184142e-02 3.70739698e-01 -6.99792802e-01 4.97674644e-02 -4.22701426e-02 1.16044231e-01 6.77717507e-01 7.71613419e-01 3.89283448e-01 -1.40048489e-01 2.62613744e-01 4.96904910e-01 6.86721876e-03 -1.99607294e-02 -4.18747693e-01 2.92331994e-01 5.57892203e-01 1.09192562e+00 -1.00057006e+00 -1.91955596e-01 -6.39181137e-02 1.21184611e+00 4.25801337e-01 7.40593076e-02 -1.00444829e+00 -2.87081301e-01 4.04031754e-01 3.32252175e-01 4.19670165e-01 -2.67850697e-01 -6.63322434e-02 -9.92245972e-01 4.76617776e-02 -8.20120156e-01 2.51818627e-01 -6.31946385e-01 -1.28584671e+00 3.37959975e-01 4.97181773e-01 -1.46103692e+00 -1.40810376e-02 -3.28825682e-01 -2.74110138e-02 3.67963314e-01 -1.15149546e+00 -1.47318363e+00 -3.63693237e-01 1.81174561e-01 8.39459896e-01 -1.15056932e-02 7.08812237e-01 2.00687513e-01 -2.17112765e-01 8.05841029e-01 4.17003110e-02 -3.97716492e-01 1.03630865e+00 -9.63672161e-01 4.59821850e-01 8.51388812e-01 6.04070956e-03 7.37415493e-01 1.02186894e+00 -9.21409547e-01 -1.33724511e+00 -1.31032181e+00 4.65646923e-01 -5.68350196e-01 3.83111387e-01 -3.88282657e-01 -1.00590849e+00 6.52016580e-01 -4.96222705e-01 9.86825451e-02 5.70274234e-01 -5.09143211e-02 -2.63163090e-01 -2.82940954e-01 -1.34589159e+00 7.25062788e-01 1.26198578e+00 5.11811906e-03 -7.39287078e-01 2.12794960e-01 7.44770169e-01 -4.87722367e-01 -5.95469236e-01 8.31849158e-01 7.45080829e-01 -5.31800210e-01 1.23551691e+00 -3.96445692e-01 3.51564020e-01 -6.12834990e-01 1.53715760e-01 -1.44519186e+00 -6.10021353e-01 -3.17507923e-01 -2.66592085e-01 9.17932153e-01 1.87186852e-01 -4.87294406e-01 8.78010154e-01 3.60584021e-01 7.03566745e-02 -7.54778326e-01 -7.71016121e-01 -9.34611559e-01 -3.22003402e-02 -1.48581220e-02 5.42657018e-01 8.17069232e-01 -3.27194840e-01 3.14809799e-01 -7.27754653e-01 2.55434275e-01 8.78625453e-01 1.48134679e-01 1.08429337e+00 -1.29336989e+00 -3.86875600e-01 8.96181613e-02 -7.27822006e-01 -1.06682694e+00 -2.00120986e-01 -5.17449677e-01 2.96154141e-01 -1.08460307e+00 4.25863415e-01 -4.99061882e-01 -9.48222652e-02 6.35756612e-01 -4.20296639e-01 6.07238114e-01 1.76541403e-01 4.65808153e-01 -6.54307365e-01 5.15347302e-01 1.15532565e+00 -6.02015890e-02 -4.60742325e-01 1.44682854e-01 -5.20601690e-01 7.71820962e-01 6.91706181e-01 -5.46372056e-01 -3.61915290e-01 -3.41030270e-01 1.00317933e-01 -2.58322328e-01 5.76453447e-01 -1.40339530e+00 1.72930047e-01 -4.73635271e-02 8.40062439e-01 -9.31364417e-01 4.75521713e-01 -8.20677638e-01 4.93789196e-01 6.93008900e-01 -1.68393493e-01 1.10878527e-01 4.47495937e-01 9.05107319e-01 4.06762026e-02 -1.47470459e-01 1.00382817e+00 -2.09170282e-01 -8.99324238e-01 3.06939572e-01 -6.60508946e-02 -3.67773443e-01 1.42132175e+00 -4.07621950e-01 2.75167022e-02 6.54579401e-02 -8.22956920e-01 2.77298570e-01 6.49715364e-01 7.06629932e-01 6.82563126e-01 -1.38607097e+00 -4.71108109e-01 3.23334157e-01 3.99787426e-01 1.99273169e-01 2.45057166e-01 7.42105067e-01 -7.54687846e-01 4.83631976e-02 -4.36231464e-01 -8.09054255e-01 -1.48450971e+00 7.00734496e-01 2.09954567e-02 -1.73867375e-01 -5.38134396e-01 1.09694707e+00 4.12517339e-01 -4.86283571e-01 2.52013415e-01 -2.32302606e-01 -2.28358269e-01 7.22239539e-02 2.65147895e-01 4.19886827e-01 3.57764587e-02 -5.98277688e-01 -3.95718902e-01 7.92795300e-01 -4.72285360e-01 1.60064638e-01 1.63187754e+00 -1.68515995e-01 6.24616211e-03 6.07719958e-01 9.27825928e-01 -2.10450873e-01 -1.58992469e+00 -2.85200536e-01 -5.75318038e-02 -6.91293478e-01 -1.95100650e-01 -7.29852080e-01 -6.89316154e-01 5.63446105e-01 1.10481489e+00 -4.13357139e-01 1.00121927e+00 -8.10981318e-02 8.54994297e-01 5.01181722e-01 5.60303688e-01 -1.20923638e+00 5.21052599e-01 3.94515067e-01 1.02242398e+00 -1.28134859e+00 2.64543682e-01 -3.34790587e-01 -5.85989296e-01 9.40908730e-01 8.74665558e-01 -3.31299245e-01 5.20595312e-01 1.21143103e-01 -2.35092714e-01 -9.65796784e-02 -3.88942331e-01 4.47550230e-02 3.10963243e-01 5.39683759e-01 -1.45226289e-02 5.33734774e-03 -2.79090017e-01 3.37070823e-01 -3.12273860e-01 1.36911094e-01 1.69637024e-01 1.24446619e+00 -5.40250957e-01 -9.74174321e-01 -6.26105011e-01 5.76050043e-01 -2.46670678e-01 6.38163611e-02 -1.84940860e-01 6.39297009e-01 3.16070467e-01 3.45211923e-01 7.40713626e-02 -3.42502505e-01 3.90902340e-01 2.64637880e-02 6.93079829e-01 -7.15471387e-01 -2.29450375e-01 9.05576907e-03 -4.04341310e-01 -2.77542919e-01 -4.63954479e-01 -6.87701166e-01 -9.26821291e-01 -2.70661026e-01 -5.28537333e-01 -1.53783932e-01 5.31368375e-01 5.88726103e-01 3.48917127e-01 3.99550587e-01 4.46950644e-01 -1.18088150e+00 -7.01701045e-01 -9.55391347e-01 -1.40620768e-01 4.69852000e-01 3.11279118e-01 -1.17031157e+00 -2.17369825e-01 2.48659670e-01]
[7.518299102783203, -0.9246379733085632]
60601b67-b2a0-46d8-ab75-0c93dfb196f1
neural-color-operators-for-sequential-image
2207.08080
null
https://arxiv.org/abs/2207.08080v2
https://arxiv.org/pdf/2207.08080v2.pdf
Neural Color Operators for Sequential Image Retouching
We propose a novel image retouching method by modeling the retouching process as performing a sequence of newly introduced trainable neural color operators. The neural color operator mimics the behavior of traditional color operators and learns pixelwise color transformation while its strength is controlled by a scalar. To reflect the homomorphism property of color operators, we employ equivariant mapping and adopt an encoder-decoder structure which maps the non-linear color transformation to a much simpler transformation (i.e., translation) in a high dimensional space. The scalar strength of each neural color operator is predicted using CNN based strength predictors by analyzing global image statistics. Overall, our method is rather lightweight and offers flexible controls. Experiments and user studies on public datasets show that our method consistently achieves the best results compared with SOTA methods in both quantitative measures and visual qualities. The code and pretrained models are provided at https://github.com/amberwangyili/neurop
['Errui Ding', 'Fu Li', 'Qi Zhang', 'Dongliang He', 'Kun Xu', 'Xin Li', 'Yili Wang']
2022-07-17
null
null
null
null
['image-retouching']
['computer-vision']
[ 4.17544320e-02 -1.43202797e-01 -2.83657044e-01 -3.03838462e-01 -5.20870209e-01 -6.97487712e-01 5.87091744e-01 -4.90237236e-01 -3.70715171e-01 4.35947567e-01 1.81985900e-01 -2.41394445e-01 4.10650849e-01 -6.19473815e-01 -1.08152103e+00 -7.47225046e-01 1.02397151e-01 -1.33756265e-01 2.09730670e-01 -2.61103719e-01 2.63014227e-01 2.95702875e-01 -1.29855669e+00 1.31704107e-01 8.05632412e-01 1.18152738e+00 -2.23735631e-01 9.43864405e-01 2.29508981e-01 8.05003285e-01 -2.35881463e-01 -5.68419874e-01 6.99491739e-01 -5.11649430e-01 -6.82288527e-01 4.79817353e-02 7.69630373e-01 -4.97169524e-01 -4.74610001e-01 1.30645525e+00 3.43298882e-01 4.75509316e-02 3.67871344e-01 -1.33783722e+00 -1.67458439e+00 5.10111570e-01 -8.18194747e-01 2.11344436e-02 1.77688658e-01 3.33316058e-01 1.17316997e+00 -1.08677292e+00 5.59079170e-01 1.17403150e+00 4.39574391e-01 5.05925179e-01 -1.61148310e+00 -8.54597509e-01 -1.55192586e-02 2.07719281e-01 -1.32861900e+00 -2.67568529e-01 7.48596370e-01 -3.35119784e-01 5.18777966e-01 3.02733034e-01 7.08280861e-01 1.01830363e+00 4.17413473e-01 8.11274707e-01 1.20255232e+00 -3.07311267e-01 2.05293804e-01 -5.80217279e-02 -2.19564453e-01 1.16642356e+00 6.41589344e-04 3.12236369e-01 -6.89730167e-01 9.16114971e-02 1.12392890e+00 1.11178800e-01 -8.62663463e-02 -8.20499241e-01 -1.44879651e+00 7.10187554e-01 9.09360766e-01 1.62536616e-03 -2.16599517e-02 6.78270400e-01 1.20522551e-01 5.94730139e-01 3.32321793e-01 3.85133326e-01 -5.35621345e-01 -9.58890989e-02 -6.33936822e-01 3.08777634e-02 4.41910595e-01 9.45204198e-01 1.12974858e+00 3.50333005e-02 -3.87680858e-01 8.39095592e-01 -9.88749564e-02 6.22841477e-01 3.21887225e-01 -1.21595466e+00 6.69366643e-02 4.74307686e-01 6.36108741e-02 -8.86691928e-01 -2.32771799e-01 -1.74009368e-01 -9.34721470e-01 6.11224473e-01 2.33515292e-01 -8.38147402e-02 -1.23910224e+00 1.94278455e+00 8.92923400e-02 4.10248816e-01 -5.02906501e-01 1.07563746e+00 2.93429255e-01 4.84981239e-01 8.35790262e-02 3.16650718e-01 1.22873414e+00 -1.08711517e+00 -5.36025584e-01 9.23937932e-02 3.54331881e-01 -6.31365895e-01 1.63694119e+00 3.85829151e-01 -1.36414874e+00 -5.79661906e-01 -1.27031171e+00 -4.31392044e-01 -5.29022515e-01 2.39395693e-01 6.48218274e-01 4.54340756e-01 -1.52786458e+00 4.32788283e-01 -7.71711886e-01 -3.33398223e-01 2.81825811e-01 3.16229224e-01 -4.44833279e-01 9.84354392e-02 -9.94781733e-01 6.45145535e-01 1.33653298e-01 -3.79162244e-02 -7.73169219e-01 -8.69926035e-01 -7.03745484e-01 -4.04605269e-02 2.78836757e-01 -7.39105701e-01 1.26774752e+00 -1.34751391e+00 -1.90858626e+00 9.39603984e-01 2.54998565e-01 -1.39277145e-01 7.15437889e-01 -3.50424707e-01 -4.12720978e-01 -6.70907646e-02 -1.05554469e-01 1.09072709e+00 1.14736152e+00 -1.31715429e+00 -6.83496475e-01 3.78888473e-02 2.29041204e-01 6.59566075e-02 -5.35819471e-01 -8.64720047e-02 -8.94307375e-01 -9.04578090e-01 -7.49804080e-02 -1.22166407e+00 2.41927858e-02 6.87541127e-01 -5.41902006e-01 2.41947666e-01 6.54218137e-01 -3.75105143e-01 1.21424496e+00 -2.34101248e+00 3.05037528e-01 3.15746278e-01 6.51422322e-01 -4.50317711e-02 -5.94891310e-01 1.77149937e-01 -2.30635494e-01 2.04563707e-01 -3.62979263e-01 -1.77720428e-01 9.47897881e-02 1.14965349e-01 -1.50077492e-01 4.86748576e-01 3.36047322e-01 1.21616769e+00 -9.76500213e-01 -2.48586521e-01 1.90700725e-01 6.40994310e-01 -8.45148683e-01 2.86283970e-01 -5.93306571e-02 1.74599349e-01 -1.83623657e-01 7.12166131e-01 9.45134759e-01 -4.67921436e-01 3.04604555e-03 -4.73266482e-01 -1.90104291e-01 -1.06841549e-01 -9.18627977e-01 1.95296061e+00 -3.22471827e-01 5.11072159e-01 -1.85596868e-01 -4.85886335e-01 7.08858728e-01 -1.15333915e-01 3.69458675e-01 -1.05780327e+00 1.13285698e-01 -3.93808484e-02 -1.99819729e-01 -1.78902626e-01 6.50447190e-01 8.79156664e-02 -5.64867519e-02 5.02699196e-01 8.67140219e-02 -1.21411383e-02 3.00295688e-02 2.96963483e-01 1.12666583e+00 3.48988980e-01 8.57530348e-03 -3.43384594e-01 1.95461690e-01 -3.55338454e-01 4.91697252e-01 6.39620245e-01 -2.98440248e-01 7.28866518e-01 5.80466211e-01 -6.11398280e-01 -1.26872671e+00 -1.45947218e+00 -7.44354539e-03 1.53060639e+00 3.09480637e-01 -3.24540794e-01 -6.89299107e-01 -6.02720261e-01 2.39604831e-01 3.09664994e-01 -1.19502461e+00 -4.86820847e-01 -4.63630229e-01 -5.16179383e-01 3.56841177e-01 6.64465368e-01 6.78547263e-01 -8.59556854e-01 -3.49646449e-01 -1.41346410e-01 2.09946454e-01 -8.40698421e-01 -1.03654420e+00 1.77097812e-01 -5.44829190e-01 -9.05703545e-01 -7.49732137e-01 -7.70198345e-01 7.55451560e-01 4.72406656e-01 1.05071282e+00 2.24319384e-01 -4.24045473e-01 3.12516868e-01 -3.80984068e-01 1.41402688e-02 -2.39139661e-01 8.58545080e-02 -7.42294788e-02 2.70957649e-01 1.33175090e-01 -5.36622226e-01 -1.09711719e+00 2.41331547e-01 -1.19408727e+00 4.48092610e-01 6.46431327e-01 9.55586791e-01 5.83422303e-01 -4.82225448e-01 4.89837769e-03 -8.50730479e-01 7.58720100e-01 -1.71897143e-01 -6.74996912e-01 2.75938988e-01 -7.75955617e-01 3.67426336e-01 5.06678700e-01 -7.15586901e-01 -8.89337897e-01 1.77537724e-01 3.00547183e-01 -4.03423578e-01 1.45490199e-01 1.86576560e-01 1.71050414e-01 -4.60306525e-01 7.33445108e-01 5.32263815e-02 -1.16850875e-01 -4.33125943e-01 1.05777240e+00 6.69932663e-02 7.85411000e-01 -6.07329786e-01 1.23739469e+00 5.20332873e-01 -2.51137346e-01 -1.06157690e-01 -5.16631901e-01 6.75800256e-03 -7.02969432e-01 -2.74484277e-01 8.19773138e-01 -8.46038520e-01 -8.23038757e-01 5.22100091e-01 -7.88828075e-01 -7.07748175e-01 -3.15676242e-01 3.70136425e-02 -5.12472630e-01 9.90619585e-02 -9.21830237e-01 -2.24313334e-01 -2.33391836e-01 -1.03338099e+00 1.06758785e+00 2.50419885e-01 1.16954841e-01 -8.28565240e-01 2.93218702e-01 -2.94690609e-01 6.79729581e-01 1.88608438e-01 9.68812525e-01 2.41216704e-01 -8.34280670e-01 -1.73604324e-01 -6.47307158e-01 3.90832216e-01 1.90444201e-01 3.28421474e-01 -8.07694018e-01 -4.09667879e-01 -5.54905057e-01 -4.57194000e-01 9.57907498e-01 1.74795076e-01 1.41736960e+00 -3.52419913e-01 1.08332172e-01 1.12597251e+00 1.66271591e+00 -9.81271587e-05 7.86378682e-01 5.15349209e-01 9.44593489e-01 -5.72129376e-02 -6.38684910e-03 4.90344733e-01 4.45042819e-01 5.32936573e-01 3.90957296e-01 -6.92734480e-01 -3.74301344e-01 -4.00824428e-01 4.83949721e-01 7.16891408e-01 -1.34760231e-01 7.02901855e-02 -6.53965354e-01 1.01454847e-01 -1.93190944e+00 -7.25283980e-01 3.80912036e-01 2.07798004e+00 1.04367661e+00 1.93750113e-02 1.44809946e-01 -2.75217652e-01 5.16320288e-01 1.27720490e-01 -7.90109873e-01 -4.42999572e-01 -3.01715314e-01 2.47326672e-01 8.50648642e-01 4.27558899e-01 -1.08323777e+00 1.20172787e+00 7.13332224e+00 4.03474092e-01 -1.35882533e+00 9.30697173e-02 7.83199310e-01 -2.52449363e-01 -5.00893891e-01 -8.48089829e-02 -2.81015057e-02 4.61617202e-01 4.94352102e-01 -8.12385380e-02 9.87102449e-01 4.47878897e-01 1.20471977e-01 7.84423426e-02 -1.04616201e+00 9.69358921e-01 1.29698617e-02 -1.39999247e+00 2.55457699e-01 -5.77731505e-02 9.78141129e-01 1.72463372e-01 8.17166209e-01 9.03465673e-02 7.09209144e-01 -7.89109230e-01 9.94447827e-01 7.55408585e-01 1.16324031e+00 -5.21957099e-01 1.05142891e-01 -5.14262557e-01 -1.13852942e+00 -1.53403148e-01 -2.77935475e-01 2.95803040e-01 -3.24008375e-01 1.79438502e-01 -1.62747175e-01 4.94429432e-02 9.14868355e-01 7.92784214e-01 -9.43091810e-01 1.00866389e+00 -2.85296917e-01 5.00296414e-01 4.27381434e-02 2.15787381e-01 2.43793204e-01 -3.50169450e-01 1.41880378e-01 1.29064977e+00 3.84490848e-01 -3.77583839e-02 -1.24843620e-01 1.09985828e+00 -5.41717172e-01 1.02836758e-01 -3.38294029e-01 1.74796745e-01 2.93665975e-01 1.42728662e+00 -5.45001268e-01 -3.12482327e-01 -5.88980138e-01 1.50234461e+00 4.91880268e-01 7.91657269e-01 -1.06624246e+00 -3.38194609e-01 8.69440317e-01 -1.17014557e-01 3.06478381e-01 -2.82195538e-01 -4.22895789e-01 -1.35388088e+00 1.92880258e-02 -8.69969249e-01 2.91642904e-01 -1.07450438e+00 -1.31508684e+00 4.82535392e-01 -3.67521495e-01 -1.38134992e+00 2.88618088e-01 -8.87022495e-01 -6.32177949e-01 6.02136254e-01 -1.45061111e+00 -1.30999160e+00 -4.64086533e-01 8.03732097e-01 1.33904755e-01 -2.17181295e-02 6.23718441e-01 2.60697395e-01 -7.04806805e-01 1.05127203e+00 1.71019837e-01 1.78701013e-01 1.14719343e+00 -1.43238366e+00 5.72259724e-01 9.59698617e-01 -3.55196488e-03 4.80654716e-01 6.11827970e-01 -2.46477708e-01 -1.70500481e+00 -9.66083407e-01 2.41839319e-01 -4.63150650e-01 1.01519191e+00 -5.46181858e-01 -9.53862071e-01 6.25439823e-01 5.02393484e-01 2.84351856e-01 5.13809919e-01 1.80561114e-02 -1.07350898e+00 -3.09103310e-01 -9.05508339e-01 9.67995286e-01 1.17911732e+00 -8.35314333e-01 7.50334188e-02 7.47777987e-03 8.20356786e-01 -3.55462193e-01 -6.86043382e-01 4.01377678e-02 7.47808099e-01 -1.01535559e+00 9.52096879e-01 -5.97373188e-01 5.28172076e-01 -4.57782805e-01 -6.07165098e-02 -1.37700868e+00 -9.60667670e-01 -7.12642074e-01 -1.55568421e-01 6.81138575e-01 5.70320666e-01 -6.33816242e-01 7.16507852e-01 9.74197447e-01 3.77725512e-02 -5.82323432e-01 -6.91295326e-01 -6.94659114e-01 1.74584344e-01 -1.96363017e-01 7.59243429e-01 9.63796616e-01 2.85033341e-02 1.45411948e-02 -6.93447769e-01 -1.44643141e-02 5.78883469e-01 -1.06227249e-01 6.77595675e-01 -7.04917371e-01 -2.57138312e-01 -8.85627568e-01 -4.62181807e-01 -9.77790773e-01 -5.79136238e-02 -9.11344230e-01 -4.05052267e-02 -1.00128853e+00 5.44502318e-01 -2.27781937e-01 -9.89758372e-01 8.96315336e-01 -2.41544381e-01 7.50252306e-01 3.14340651e-01 1.72817364e-01 -7.71216512e-01 5.33402205e-01 1.52957559e+00 -3.00448418e-01 -2.60856241e-01 -5.74585259e-01 -8.75134468e-01 2.92730540e-01 7.99907923e-01 -9.20329913e-02 -3.24542969e-01 -8.06340039e-01 5.47663271e-01 -2.92608917e-01 4.92123067e-01 -8.76031876e-01 2.48819023e-01 -3.90478015e-01 5.19859731e-01 8.81024301e-02 1.96802348e-01 -6.32215440e-01 6.75278232e-02 2.77476460e-01 -6.97098792e-01 5.46338797e-01 8.81977975e-02 4.45069373e-01 4.17574570e-02 5.90372086e-01 9.63259220e-01 2.70519763e-01 -8.78287911e-01 6.65197551e-01 -2.49941926e-02 -2.28345096e-01 8.32957864e-01 -2.06501305e-01 -4.86324638e-01 -5.30100524e-01 -4.89234716e-01 -7.24930316e-02 7.42372632e-01 7.81113148e-01 7.22559690e-01 -1.79077792e+00 -6.14630163e-01 4.62775052e-01 4.04459745e-01 -6.30014241e-01 5.34495898e-02 6.82785749e-01 -6.76335454e-01 -2.29441151e-02 -5.40299237e-01 -2.57416695e-01 -8.37103784e-01 7.74826288e-01 4.17822421e-01 2.95538396e-01 -7.22649932e-01 8.82069409e-01 4.58041817e-01 -1.02613665e-01 1.94503590e-01 -5.48217118e-01 3.08185905e-01 -4.01576728e-01 5.35884500e-01 1.61241069e-01 -1.69452950e-01 -5.10665894e-01 -1.07426740e-01 7.31074691e-01 -9.28017944e-02 -2.00675979e-01 1.19975543e+00 -2.09414348e-01 -2.06297815e-01 3.29752833e-01 1.58204544e+00 -1.31855488e-01 -1.79859495e+00 -4.96351004e-01 -1.68541938e-01 -6.19895875e-01 1.64044008e-01 -1.04996586e+00 -1.50288141e+00 6.30380988e-01 1.10191023e+00 1.14084132e-01 1.45651639e+00 -1.05057508e-02 7.43033290e-01 3.25931281e-01 1.57083496e-01 -1.25550199e+00 4.68598515e-01 4.15327549e-01 1.01358378e+00 -1.29297554e+00 -2.93229342e-01 -2.51263031e-03 -7.13595033e-01 9.60108221e-01 7.45884418e-01 -3.27897042e-01 8.28838944e-01 3.72543097e-01 4.42242444e-01 9.51494835e-03 -7.68910944e-01 -1.21391498e-01 4.55413133e-01 4.42146540e-01 5.11686802e-01 3.59285712e-01 2.65354868e-02 -4.87075709e-02 -1.84201494e-01 5.94212860e-03 3.95472854e-01 5.68133295e-01 -2.45707378e-01 -1.09041190e+00 -1.90040022e-01 3.28770846e-01 -2.35423625e-01 -2.16371149e-01 -2.81371117e-01 6.22323453e-01 6.55079186e-02 4.99653131e-01 1.46940306e-01 -8.60121787e-01 3.49866152e-01 -1.50171295e-01 3.93318266e-01 -2.02989548e-01 -3.94687802e-01 3.97148691e-02 -5.04466653e-01 -1.05480433e+00 -2.68664807e-01 -2.48477027e-01 -1.01292479e+00 -6.18800044e-01 9.61391255e-02 -4.44299042e-01 5.35856307e-01 3.53740185e-01 4.18605119e-01 4.14160192e-01 8.30985129e-01 -7.73038924e-01 -3.02695721e-01 -4.40864801e-01 -6.54715180e-01 8.72760534e-01 5.99429369e-01 -5.36504030e-01 -1.35418475e-01 3.53642195e-01]
[11.404088020324707, -0.9918814301490784]
0617cfd2-b76b-4801-9d95-f980fbb644af
data-augmentation-for-low-resource-dialogue-1
null
null
https://aclanthology.org/2022.findings-naacl.53
https://aclanthology.org/2022.findings-naacl.53.pdf
Data Augmentation for Low-Resource Dialogue Summarization
We present DADS, a novel Data Augmentation technique for low-resource Dialogue Summarization. Our method generates synthetic examples by replacing sections of text from both the input dialogue and summary while preserving the augmented summary to correspond to a viable summary for the augmented dialogue. We utilize pretrained language models that produce highly likely dialogue alternatives while still being free to generate diverse alternatives. We applied our data augmentation method to the SAMSum dataset in low resource scenarios, mimicking real world problems such as chat, thread, and meeting summarization where large scale supervised datasets with human-written summaries are scarce. Through both automatic and human evaluations, we show that DADS shows strong improvements for low resource scenarios while generating topically diverse summaries without introducing additional hallucinations to the summaries.
['Shashi Narayan', 'Gonçalo Simões', 'Joshua Maynez', 'Yongtai Liu']
null
null
null
null
findings-naacl-2022-7
['meeting-summarization']
['natural-language-processing']
[ 5.10931671e-01 9.76717293e-01 -2.70963442e-02 -3.34236294e-01 -1.39770997e+00 -7.01232553e-01 1.00017679e+00 3.94852191e-01 -2.97462910e-01 1.36898041e+00 1.27839696e+00 -9.11939610e-03 3.62259060e-01 -3.54960293e-01 -1.75332278e-01 -1.29662275e-01 3.89196903e-01 9.33597982e-01 -2.63667673e-01 -4.33092117e-01 3.87718022e-01 -7.61181563e-02 -1.16643369e+00 7.13794053e-01 1.29077578e+00 1.80226609e-01 7.87753016e-02 1.13949978e+00 -1.35916129e-01 7.27744281e-01 -1.50503576e+00 -3.79597247e-01 8.69905129e-02 -1.00965321e+00 -1.14296222e+00 4.56398159e-01 7.39329457e-01 -5.05331933e-01 -3.49752575e-01 4.81995344e-01 7.81111538e-01 5.38540900e-01 7.27643847e-01 -9.24971402e-01 -4.76734579e-01 1.02246761e+00 -2.72390723e-01 -5.00677414e-02 1.02870035e+00 2.61261195e-01 1.09993136e+00 -7.41383255e-01 7.52216995e-01 1.43461680e+00 4.91879344e-01 8.67825747e-01 -1.50938511e+00 -1.87974691e-01 -3.54276188e-02 -3.38934004e-01 -6.11798108e-01 -9.54829454e-01 6.31825209e-01 4.65143509e-02 1.17854202e+00 6.74031317e-01 3.74856442e-01 1.44128704e+00 -1.41025245e-01 1.05428743e+00 6.91617727e-01 -3.94116014e-01 1.14689074e-01 2.56746799e-01 4.15993154e-01 2.43252277e-01 9.04597193e-02 -8.23946893e-01 -8.15953434e-01 -6.44762278e-01 1.73153132e-01 -6.62278116e-01 -3.53978962e-01 4.10102993e-01 -1.53788912e+00 8.27944100e-01 -4.72240150e-02 1.30779827e-02 -6.41195118e-01 -3.54600996e-01 6.04286730e-01 3.23983699e-01 6.45154774e-01 1.27991843e+00 -2.46937916e-01 -3.47782373e-01 -1.02314317e+00 7.98623860e-01 1.39483905e+00 1.08671999e+00 3.60587716e-01 2.81008601e-01 -9.98488665e-01 1.03955305e+00 -3.88891548e-01 2.77983904e-01 6.44257665e-01 -1.36698461e+00 8.37317050e-01 5.08939505e-01 5.85033655e-01 -6.09164536e-01 -4.26713914e-01 -2.07377329e-01 -8.42004478e-01 -4.71753448e-01 3.41829807e-01 -5.49836755e-01 -4.74759340e-01 1.74755776e+00 1.12701375e-02 -2.51712531e-01 6.11152649e-01 4.17033464e-01 1.31784487e+00 1.02985811e+00 -2.37181306e-01 -7.11948395e-01 9.35181499e-01 -1.26873875e+00 -1.12277257e+00 -4.51666117e-01 7.69641221e-01 -8.31216991e-01 1.53511059e+00 2.28345156e-01 -1.54163444e+00 -3.26132327e-01 -1.07922661e+00 -3.35084409e-01 3.01244736e-01 2.39015117e-01 3.10784638e-01 1.75159425e-01 -1.12398016e+00 6.06868386e-01 -4.83727872e-01 -4.54722315e-01 -2.64031719e-02 6.84654936e-02 -3.62567484e-01 1.74436286e-01 -1.04806423e+00 9.81632233e-01 3.38371724e-01 -3.74892801e-01 -5.81877351e-01 -5.51398396e-01 -1.08837748e+00 1.23748653e-01 3.38230789e-01 -9.30215061e-01 1.84600425e+00 -7.10705578e-01 -1.59326017e+00 4.92271185e-01 -4.26035047e-01 -6.69133544e-01 6.42275453e-01 -4.29193854e-01 -3.26753408e-02 5.69181815e-02 3.61758590e-01 8.04042399e-01 3.10409307e-01 -1.18829906e+00 -1.42346263e-01 7.31152371e-02 -1.42000793e-02 8.36404443e-01 -2.79899180e-01 -1.91418961e-01 5.17812464e-03 -6.67948663e-01 -3.51649165e-01 -7.58278251e-01 -4.08114851e-01 -7.54449904e-01 -1.18454027e+00 -2.47604012e-01 6.09845459e-01 -9.17350292e-01 1.29696274e+00 -1.55625105e+00 2.16028884e-01 -4.05694157e-01 1.47708684e-01 2.99680889e-01 -5.15288830e-01 8.98148477e-01 3.27083915e-01 2.92124331e-01 -4.52264398e-01 -7.67807364e-01 -4.91477475e-02 3.99257913e-02 -4.33757752e-01 -1.31523892e-01 2.55380958e-01 6.96271658e-01 -9.98776972e-01 -3.94795030e-01 -3.63698602e-02 -1.36036411e-01 -6.29643023e-01 7.22831905e-01 -7.12133467e-01 5.35525084e-01 -2.31688842e-01 4.43660356e-02 2.67414868e-01 -4.17857319e-02 9.03478563e-02 -5.64596541e-02 3.19542252e-02 9.44902837e-01 -6.74057424e-01 1.82973874e+00 -5.87028265e-01 8.40020895e-01 -7.29358345e-02 -4.52549815e-01 9.33009684e-01 4.72104400e-01 -1.16375096e-01 -3.40418756e-01 -4.35964949e-02 -8.09071469e-04 8.53493512e-02 -4.55196917e-01 1.45539701e+00 -4.07355055e-02 -4.87847835e-01 1.06088603e+00 1.61388353e-01 -6.23732030e-01 5.72012603e-01 9.29175079e-01 1.14909005e+00 -1.86693534e-01 4.83142316e-01 -6.71106055e-02 2.81385750e-01 3.07156175e-01 3.28698337e-01 1.10762525e+00 1.18415341e-01 9.35387015e-01 8.74884427e-01 8.16835314e-02 -1.19931984e+00 -8.97044182e-01 4.53285575e-01 9.41934764e-01 -2.44086236e-01 -7.55495191e-01 -8.91159952e-01 -6.74761355e-01 -4.49752510e-01 1.54001689e+00 -3.06873858e-01 -7.37395138e-02 -6.60554111e-01 -5.08132279e-01 7.73036242e-01 2.62511104e-01 5.03972232e-01 -1.21597660e+00 -3.20476860e-01 3.41864824e-01 -7.98869073e-01 -1.09212708e+00 -9.09200191e-01 -9.88024548e-02 -7.48526871e-01 -8.64996374e-01 -5.90958118e-01 -5.94745994e-01 4.91143882e-01 1.23198502e-01 1.40624106e+00 -1.16959088e-01 1.10051870e-01 2.66390175e-01 -1.11396268e-01 -3.71674448e-01 -1.26343036e+00 3.89590681e-01 1.16542004e-01 -4.51000303e-01 -1.38109386e-01 -4.49223191e-01 -2.04021603e-01 -1.20361149e-01 -7.81626701e-01 5.21547139e-01 2.89704263e-01 1.18542838e+00 9.62803662e-02 -7.07325995e-01 1.29917967e+00 -1.24879181e+00 1.57004035e+00 -4.97767061e-01 2.23652214e-01 2.94195712e-01 -1.68603763e-01 1.21596649e-01 9.46985483e-01 -4.32064950e-01 -1.45047152e+00 -3.87287259e-01 -9.82456654e-02 2.04764768e-01 -1.63678885e-01 4.74048525e-01 -2.07345232e-01 8.71265352e-01 1.06389654e+00 2.40974382e-01 1.51447222e-01 -3.59436959e-01 5.33187628e-01 8.58788371e-01 7.74909496e-01 -5.51465809e-01 4.12753522e-01 -5.05412705e-02 -6.57264471e-01 -1.30297422e+00 -1.08110833e+00 -1.88897088e-01 -5.66376209e-01 3.67306098e-02 5.67367315e-01 -7.39339590e-01 -4.08042610e-01 1.38095468e-01 -1.64952207e+00 -4.48153377e-01 -7.14263380e-01 1.68411404e-01 -7.10045159e-01 4.80827481e-01 -5.62707067e-01 -8.88762951e-01 -7.06192434e-01 -8.03594112e-01 9.48282719e-01 5.03744066e-01 -1.23739326e+00 -9.66753244e-01 2.55827934e-01 5.00223756e-01 3.30750138e-01 3.76579553e-01 8.98703694e-01 -1.61599112e+00 -1.54070724e-02 -2.67112494e-01 1.46306470e-01 3.80999297e-01 5.10713935e-01 -1.74083337e-02 -8.07780325e-01 -1.06484570e-01 -4.25039455e-02 -9.08022940e-01 7.90056169e-01 9.37408879e-02 6.57264471e-01 -1.21466303e+00 9.51148644e-02 -8.81663635e-02 4.45463866e-01 4.81963269e-02 4.03748155e-01 -1.39269471e-01 4.81526852e-01 7.94503808e-01 6.34688377e-01 7.06473827e-01 4.20375854e-01 4.00739104e-01 -2.12406620e-01 -4.65833247e-02 -1.10536925e-01 -4.73852098e-01 4.09535527e-01 1.12535584e+00 4.67993945e-01 -1.02342093e+00 -6.22618437e-01 8.15404832e-01 -1.85331655e+00 -1.03537476e+00 7.91790783e-02 2.09854317e+00 1.31737840e+00 1.56771123e-01 2.13215962e-01 -9.24915820e-02 7.24335492e-01 6.19480073e-01 -5.40299833e-01 -8.57275844e-01 -3.42598826e-01 3.64687927e-02 -2.34664157e-01 8.14963043e-01 -7.60397196e-01 8.61612380e-01 6.58004475e+00 4.05816436e-01 -5.14776349e-01 -2.04234138e-01 7.53678501e-01 -5.94730914e-01 -7.33263016e-01 -1.58760071e-01 -4.38939869e-01 2.42904663e-01 1.15677989e+00 -9.07552361e-01 1.08743273e-01 4.98627454e-01 5.41301072e-01 -2.66211450e-01 -1.36978710e+00 4.60279971e-01 3.98341745e-01 -1.67129910e+00 5.09373844e-01 -3.81475180e-01 9.33385968e-01 -3.60311329e-01 -1.78781137e-01 4.22949970e-01 6.43093467e-01 -1.06201208e+00 2.67517090e-01 2.03807220e-01 7.80562997e-01 -7.13777781e-01 7.67154336e-01 6.08203590e-01 -3.27318668e-01 4.12157923e-01 -1.55562252e-01 -2.73232579e-01 4.34610844e-01 1.79936916e-01 -1.69736791e+00 4.77121621e-01 -2.39057496e-01 4.88845229e-01 -5.35066903e-01 4.82065976e-01 -1.16792157e-01 6.01713061e-01 -2.25053906e-01 -1.41616091e-01 3.43440771e-01 -8.66189525e-02 1.12440455e+00 1.43785703e+00 1.78206116e-01 1.35847718e-01 4.47785139e-01 7.50315964e-01 -5.48708677e-01 1.42173931e-01 -9.76469338e-01 -3.03004384e-01 8.99474680e-01 1.11362386e+00 -2.31851459e-01 -6.88266575e-01 5.87565824e-02 1.10366642e+00 2.29576007e-01 2.89363801e-01 -2.94987142e-01 -4.90921617e-01 4.27402765e-01 -4.23781462e-02 -4.61487234e-01 -3.62708718e-02 -5.06706655e-01 -1.13760817e+00 -1.51284456e-01 -1.32471132e+00 3.19446385e-01 -7.69212902e-01 -1.10650492e+00 8.90071630e-01 2.28117220e-02 -1.02710366e+00 -9.29296553e-01 3.62456352e-01 -1.26304460e+00 9.78641510e-01 -8.16900671e-01 -8.47638190e-01 -2.47443065e-01 7.36119077e-02 1.38096368e+00 -4.15707618e-01 1.12720597e+00 -3.74619037e-01 -5.20683050e-01 7.14365602e-01 -6.61339685e-02 -1.59571484e-01 1.12613380e+00 -1.54467773e+00 8.42036247e-01 6.79511368e-01 -7.82381892e-02 6.27897561e-01 1.29840922e+00 -8.98081183e-01 -8.57954323e-01 -1.05601871e+00 1.07485557e+00 -3.29111159e-01 4.22014743e-01 -2.71523476e-01 -1.10437179e+00 6.57103896e-01 1.04793835e+00 -9.29815829e-01 8.10498953e-01 4.90497127e-02 1.26165330e-01 5.61195910e-01 -1.25362980e+00 1.11107481e+00 7.90296137e-01 -2.40643337e-01 -1.11801982e+00 6.46269560e-01 1.02771688e+00 -6.37360752e-01 -4.49112147e-01 1.68671310e-01 1.92710862e-01 -6.94543064e-01 5.30984044e-01 -9.58491445e-01 9.63359833e-01 1.81387663e-01 1.69662863e-01 -1.87219000e+00 2.34661937e-01 -1.32636964e+00 -1.59530610e-01 1.62424076e+00 7.38073409e-01 -1.80427670e-01 7.69918978e-01 8.46075475e-01 -5.62175333e-01 -4.07344192e-01 -6.33178890e-01 -4.02829856e-01 -1.58654142e-03 2.91328251e-01 4.10366625e-01 7.52538025e-01 4.26997900e-01 1.08453798e+00 -6.30645096e-01 -4.51971889e-01 3.69473577e-01 -6.72835261e-02 1.16704094e+00 -9.32538569e-01 -1.89903796e-01 -2.96328753e-01 4.12678659e-01 -1.27046478e+00 4.65110213e-01 -7.26574719e-01 3.77347022e-01 -1.99315190e+00 2.45634452e-01 7.25364089e-02 5.77434123e-01 2.06924140e-01 -5.24211884e-01 -1.12366535e-01 1.44203782e-01 4.55988012e-02 -7.63400316e-01 9.25573885e-01 1.18231499e+00 -1.96557954e-01 -7.35681474e-01 2.39652351e-01 -1.16041994e+00 8.09646130e-01 8.43044579e-01 -2.47133121e-01 -6.37563407e-01 -2.88124651e-01 -3.49884838e-01 6.27823710e-01 -1.88807935e-01 -6.30305290e-01 1.59003049e-01 -1.90873012e-01 1.48314148e-01 -8.49051774e-01 4.79786813e-01 1.34531960e-01 -3.63079697e-01 1.27767920e-01 -1.29407692e+00 2.98290193e-01 3.65559816e-01 4.77417409e-01 -2.02118024e-01 -3.35393190e-01 6.33304179e-01 -3.44307244e-01 1.05951361e-01 -2.58044422e-01 -7.32580304e-01 7.02424645e-01 5.98065317e-01 -1.08650841e-01 -6.40902162e-01 -1.18978584e+00 -6.70959890e-01 5.99806249e-01 5.49820423e-01 2.89740622e-01 5.17635345e-01 -1.01087403e+00 -1.26482832e+00 -1.87070474e-01 -1.04709320e-01 2.62811482e-01 2.14137241e-01 2.34631315e-01 -4.21788663e-01 3.97202462e-01 -7.14597404e-02 -2.63114482e-01 -1.34136522e+00 6.34932751e-03 -7.59735005e-03 -5.76777399e-01 -6.59969270e-01 5.83163917e-01 1.97476387e-01 -6.11738622e-01 3.62986594e-01 -2.42410917e-02 -3.39727968e-01 3.21943939e-01 7.03641474e-01 6.19146705e-01 1.18507095e-01 -2.41649598e-01 5.61614372e-02 -5.05813301e-01 -4.32282299e-01 -6.46905780e-01 1.19633400e+00 -3.28099042e-01 1.38651744e-01 7.22852290e-01 7.60959983e-01 3.33800107e-01 -1.05871022e+00 -2.90641129e-01 1.58452466e-01 -1.35592699e-01 -5.63155115e-01 -1.05875885e+00 -1.83003977e-01 5.35034359e-01 -5.07214487e-01 3.70420724e-01 6.51130617e-01 -1.27443090e-01 9.24470603e-01 1.03457606e+00 -2.67006695e-01 -1.07083690e+00 5.48266292e-01 8.06047440e-01 1.52394640e+00 -1.17704415e+00 1.98671222e-01 -7.63353482e-02 -1.46865869e+00 9.32129085e-01 9.65409040e-01 -1.24804517e-02 -4.34959054e-01 3.28825355e-01 1.12652987e-01 1.52525574e-01 -1.42253351e+00 3.46162081e-01 1.10893846e-01 5.89424372e-01 7.62143850e-01 -1.42418757e-01 -3.66263747e-01 8.33746374e-01 -7.09687352e-01 -5.98775625e-01 1.51117468e+00 6.33099198e-01 -2.97357500e-01 -8.53140652e-01 -1.89915746e-01 8.38959038e-01 -3.15698117e-01 -2.33913243e-01 -1.14741623e+00 6.20821595e-01 -9.11897600e-01 1.22634816e+00 2.46815860e-01 -1.28833458e-01 4.48506951e-01 4.45131510e-01 2.71731704e-01 -1.29212797e+00 -1.06324375e+00 -3.11140660e-02 1.15857589e+00 -2.27632616e-02 -9.95778218e-02 -6.38479114e-01 -1.23031318e+00 -3.38008791e-01 -2.60436088e-01 5.62214315e-01 2.41678804e-01 8.59929681e-01 4.64255482e-01 7.32956111e-01 6.94355428e-01 -1.05571079e+00 -7.71250427e-01 -1.53174520e+00 -7.02676699e-02 4.81367767e-01 4.31025833e-01 3.92853618e-02 -3.22739512e-01 -8.32309574e-02]
[12.39822006225586, 9.134276390075684]
e6805ac1-93b2-48de-b205-a2e7a23dcb78
efficient-adaptation-for-end-to-end-vision-1
null
null
https://openreview.net/forum?id=CVN5cZBFFFG
https://openreview.net/pdf?id=CVN5cZBFFFG
Efficient Adaptation for End-to-End Vision-Based Robotic Manipulation
One of the great promises of robot learning systems is that they will be able to learn from their mistakes and continuously adapt to ever-changing environments, but most robot learning systems today are deployed as fixed policies which do not adapt after deployment. Can we efficiently adapt previously learned behaviors to new environments, objects and percepts in the real world? We present empirical evidence towards a robot learning framework that facilitates continuous adaption. We demonstrate how to adapt vision-based robotic manipulation policies to new variations by fine-tuning via off-policy reinforcement learning, using less than 0.2% of the data necessary to learn the task from scratch. We find that the simple approach of fine-tuning pre-trained policies leads to substantial performance gains over the course of fine-tuning, and that pre-training via RL is essential: training from scratch or adapting from supervised ImageNet features are both unsuccessful with such small amounts of data. We also find that these positive results hold in a limited continual learning setting, in which we repeatedly fine-tune a single lineage of policies using data from a succession of new tasks. Our empirical conclusions are consistently supported by experiments on simulated manipulation tasks, and by 52 unique fine-tuning experiments on a real robotic grasping system pre-trained on 580,000 grasps.
['Karol Hausman', 'Chelsea Finn', 'Sergey Levine', 'Gaurav S. Sukhatme', 'Benjamin Swanson', 'Ryan Julian']
2020-06-12
null
null
null
icml-workshop-lifelongml-2020-7
['robotic-grasping']
['robots']
[ 4.00702916e-02 7.10721686e-02 -5.93137182e-02 -2.56740451e-01 -2.67584264e-01 -8.21932495e-01 5.31554222e-01 -1.10941090e-01 -9.01620686e-01 1.13440824e+00 -9.87101719e-02 -1.06218174e-01 -2.75605321e-01 -6.17406428e-01 -1.21873510e+00 -5.30499578e-01 -4.21333671e-01 9.08328831e-01 6.56438649e-01 -6.13257766e-01 2.36299336e-01 7.05978930e-01 -1.54622436e+00 1.53015396e-02 4.21535492e-01 5.77466786e-01 7.18440831e-01 8.37253034e-01 4.65443581e-02 4.59637046e-01 -4.13554370e-01 2.18397364e-01 6.61705256e-01 1.04401417e-01 -9.57125068e-01 8.37719291e-02 1.58794656e-01 -5.46606123e-01 -4.84296918e-01 7.47627020e-01 3.16697687e-01 2.94810712e-01 5.13433278e-01 -1.03721297e+00 -7.65141070e-01 7.37802267e-01 -2.65548855e-01 4.00554053e-02 5.83161935e-02 9.00261223e-01 5.12349844e-01 -4.19853836e-01 6.48990691e-01 1.47758543e+00 8.39513123e-01 8.66684198e-01 -1.24346006e+00 -4.68845397e-01 3.06105644e-01 -5.67098707e-03 -7.87622631e-01 -3.77857387e-01 3.44774693e-01 -4.96192247e-01 1.23717856e+00 -5.30944169e-01 7.27335870e-01 1.33194482e+00 4.48338628e-01 4.16825294e-01 9.97160316e-01 -4.60637808e-01 3.74295801e-01 -7.51239136e-02 -3.76422793e-01 8.10106099e-01 3.99011314e-01 5.28003097e-01 -1.27222717e-01 -6.71362579e-02 1.18684697e+00 -1.10264570e-02 5.12811802e-02 -9.53767538e-01 -1.43224597e+00 5.61452687e-01 6.71928704e-01 2.53441811e-01 -5.49387395e-01 7.15139031e-01 6.12523973e-01 8.42172027e-01 -2.23717898e-01 1.10954368e+00 -1.08913672e+00 -2.19893083e-01 -3.88173349e-02 3.03996623e-01 9.68259037e-01 1.07283473e+00 1.10707450e+00 2.40075156e-01 2.52604514e-01 8.76997769e-01 -1.80928446e-02 6.38884306e-01 8.79258096e-01 -1.57025027e+00 9.37451236e-03 1.01997986e-01 3.58129114e-01 -4.63011295e-01 -4.58588034e-01 -5.06316684e-02 -1.83409646e-01 7.06257582e-01 4.22723532e-01 -5.90538025e-01 -1.16513836e+00 1.82698274e+00 2.04632178e-01 -3.57871324e-01 1.06817648e-01 5.60032308e-01 -1.66133821e-01 3.83363098e-01 2.12578520e-01 -6.41608611e-02 6.17814362e-01 -9.84311521e-01 -4.10482772e-02 -4.51624721e-01 4.42473203e-01 -4.31994885e-01 1.39353824e+00 5.52663326e-01 -9.36922252e-01 -7.83821046e-01 -1.00572431e+00 4.37944859e-01 -2.79922664e-01 -2.25671753e-01 9.42776024e-01 1.11118294e-01 -1.17307317e+00 8.62228513e-01 -1.12540388e+00 -8.91997695e-01 4.29810524e-01 7.11393476e-01 -4.35332239e-01 -8.59718025e-02 -6.46404564e-01 1.25889492e+00 8.56792450e-01 -2.09333211e-01 -1.52219558e+00 -3.17658871e-01 -3.35434884e-01 -1.76262781e-01 6.58481181e-01 -7.46299505e-01 1.72538638e+00 -1.21486187e+00 -1.90946960e+00 4.65841353e-01 4.08295393e-01 -3.06049526e-01 2.80352801e-01 -3.51812720e-01 1.31352201e-01 2.13064868e-02 5.28058372e-02 9.91920590e-01 1.06041682e+00 -1.72446406e+00 -9.12747860e-01 -8.78415927e-02 4.25735742e-01 1.83673874e-01 -3.59848559e-01 -4.99069512e-01 -1.15060911e-01 -3.05329293e-01 -1.60155237e-01 -1.20604527e+00 -4.75389361e-01 6.74989494e-03 3.42343569e-01 -3.52568626e-01 6.69754326e-01 -7.90878832e-02 1.61583766e-01 -1.93551731e+00 2.89408088e-01 6.68656304e-02 -1.52913153e-01 2.25522935e-01 -5.96151948e-01 4.89607841e-01 2.53379077e-01 -1.82888702e-01 -4.44299616e-02 2.23941490e-01 1.31500870e-01 8.76083910e-01 -4.20587599e-01 1.62991822e-01 5.75655736e-02 1.05314207e+00 -1.37668121e+00 -7.54706785e-02 1.35224819e-01 -1.31444737e-01 -7.61114657e-01 2.79069245e-01 -6.76171184e-01 4.71459359e-01 -5.41739345e-01 4.48236704e-01 -5.01605682e-02 -1.22419924e-01 3.28842402e-01 6.47566244e-02 3.49469855e-02 -1.88543871e-01 -7.05854118e-01 1.89022350e+00 -6.08425021e-01 2.85354257e-01 1.97944701e-01 -1.24653089e+00 8.12510848e-01 5.19942716e-02 5.87824643e-01 -5.12217760e-01 7.82905221e-02 3.86671454e-01 3.16530138e-01 -8.85700524e-01 2.92430818e-01 -3.23864788e-01 -1.39910311e-01 4.15692478e-01 5.76424837e-01 -6.96223319e-01 2.51311123e-01 -2.07270309e-01 1.49523902e+00 5.29752135e-01 2.40405068e-01 -2.45937139e-01 1.83514189e-02 5.98167479e-01 3.55598181e-01 1.33107781e+00 -3.16364288e-01 -1.30931456e-02 -6.82464391e-02 -7.43214130e-01 -1.36981797e+00 -1.19602275e+00 2.29530156e-01 1.62908578e+00 2.11377032e-02 1.67033777e-01 -5.10672569e-01 -6.76943421e-01 4.27608490e-01 5.68513155e-01 -6.65379465e-01 -3.88561398e-01 -9.50114965e-01 -3.79165292e-01 3.76047969e-01 6.38566256e-01 3.03731322e-01 -1.92390275e+00 -1.10235548e+00 5.65050602e-01 5.48837483e-01 -8.79868507e-01 -7.61995907e-04 7.37548828e-01 -9.74372983e-01 -1.05740333e+00 -5.31527162e-01 -1.10171402e+00 7.58087754e-01 2.72604227e-01 9.48734641e-01 1.64741918e-01 -2.65314937e-01 1.10478759e+00 -3.94766718e-01 -3.62472177e-01 -8.23336661e-01 2.45873883e-01 6.64167106e-01 -8.89592230e-01 6.28435165e-02 -9.64589536e-01 -2.99047083e-01 2.33403474e-01 -7.18468487e-01 -4.19912308e-01 1.16884506e+00 1.22533059e+00 3.73210311e-01 6.14172034e-02 8.35491955e-01 -7.09861875e-01 9.27016556e-01 -5.50075352e-01 -5.30740678e-01 3.50333303e-01 -6.42280579e-01 4.22702461e-01 8.31360877e-01 -9.52997267e-01 -9.88031685e-01 2.22278729e-01 1.98684663e-01 -4.18621212e-01 -5.36581278e-01 2.12245882e-01 4.19796437e-01 -2.64594942e-01 1.10707462e+00 9.83380973e-02 1.15478739e-01 -3.49052727e-01 7.00061679e-01 2.84115225e-01 7.63663352e-01 -1.32777071e+00 8.40038896e-01 2.34752849e-01 -3.08752060e-01 -6.91020310e-01 -4.90089059e-01 -9.83905494e-02 -8.03988159e-01 3.43549028e-02 4.54601288e-01 -5.44158638e-01 -9.12158906e-01 5.19514501e-01 -8.15243661e-01 -1.29060829e+00 -7.40907550e-01 3.55261475e-01 -1.15460908e+00 1.67591736e-01 -6.20737195e-01 -3.84321064e-01 -4.08598110e-02 -1.05476499e+00 6.88738525e-01 2.49350682e-01 -1.07738040e-01 -7.92516947e-01 3.49473774e-01 -3.93461227e-01 7.39493012e-01 -1.44813597e-01 9.81362164e-01 -4.56197858e-01 -3.57824564e-01 9.10943076e-02 9.85370949e-02 4.62035626e-01 5.48801243e-01 -2.53223896e-01 -5.54044604e-01 -8.08987916e-01 -1.65786922e-01 -1.08020937e+00 7.58195341e-01 3.31682488e-02 1.07784915e+00 -3.02541494e-01 -4.57899332e-01 3.49775553e-01 1.34434676e+00 3.60978067e-01 1.64095312e-01 6.53396070e-01 5.04550517e-01 3.67556691e-01 5.89958251e-01 3.24556857e-01 1.40369162e-01 3.28133851e-01 5.90267360e-01 5.16790152e-01 -6.76587895e-02 -3.14186066e-01 4.67199773e-01 5.01787603e-01 -1.64302319e-01 1.64522111e-01 -8.76084447e-01 7.57346809e-01 -1.98485553e+00 -8.85917783e-01 8.83523643e-01 1.91488242e+00 1.09155309e+00 4.27276671e-01 1.14163183e-01 -4.73711342e-01 2.91935205e-01 -4.17345852e-01 -1.20042956e+00 -3.88237625e-01 2.30921134e-01 3.88448179e-01 7.20793247e-01 4.47678268e-01 -9.76776659e-01 1.47584867e+00 7.52346754e+00 4.21495825e-01 -1.38745308e+00 -5.75216934e-02 7.19767669e-03 -3.13866436e-02 4.12109643e-02 -6.09128550e-02 -4.44453508e-01 1.98802233e-01 7.44423389e-01 3.19972150e-02 1.21529508e+00 1.19144309e+00 -1.58582211e-01 -1.84346035e-01 -1.24553144e+00 5.90838552e-01 -2.58926123e-01 -1.10986042e+00 -8.80727917e-02 -1.75793737e-01 8.66132855e-01 6.72520578e-01 -1.42022058e-01 8.89314771e-01 1.26049447e+00 -8.74821544e-01 7.19102442e-01 4.67737168e-01 6.08520508e-01 -3.33104193e-01 3.40926707e-01 7.22895086e-01 -5.39479733e-01 -6.92940891e-01 -7.69797564e-01 -2.12271705e-01 -9.99268070e-02 -1.40511096e-01 -1.35014248e+00 -4.90833819e-02 9.33805048e-01 5.86775661e-01 -3.93953592e-01 9.36816037e-01 -2.64008135e-01 5.32432199e-01 -4.70521063e-01 -2.79869944e-01 3.21595699e-01 2.82644451e-01 3.87178570e-01 9.94983792e-01 1.46658003e-01 3.36659513e-02 4.43818659e-01 4.33546782e-01 1.25023097e-01 -4.66226459e-01 -8.64551663e-01 -7.19133466e-02 5.28969288e-01 1.01024663e+00 -5.93351603e-01 -3.05356771e-01 -1.15620971e-01 7.99439788e-01 8.00968647e-01 5.27136028e-01 -3.91218275e-01 -1.88386783e-01 5.88375270e-01 -2.85794050e-01 7.59996176e-01 -6.73954427e-01 1.01828620e-01 -9.44109499e-01 -2.50925630e-01 -1.07585180e+00 -8.61350149e-02 -7.63660669e-01 -1.63541722e+00 3.60392362e-01 6.17827848e-02 -7.98839688e-01 -5.96560895e-01 -9.73164082e-01 -1.77930236e-01 3.01797777e-01 -1.37876976e+00 -8.43650281e-01 -1.99026987e-01 5.86474597e-01 7.49732375e-01 -4.49314415e-01 9.36881185e-01 -2.40842521e-01 1.16983183e-01 3.84509951e-01 3.64346832e-01 -1.11674756e-01 7.76539624e-01 -1.20787442e+00 2.52975345e-01 2.32064530e-01 -2.71508321e-02 7.14355469e-01 8.67291331e-01 -5.14482558e-01 -1.58304763e+00 -7.87736952e-01 -2.71888435e-01 -5.71275592e-01 9.60814178e-01 -1.43092617e-01 -9.82243240e-01 1.01591861e+00 9.85016525e-02 8.09711218e-02 -6.36786222e-02 1.85459480e-01 -2.41940469e-01 -2.29910418e-01 -1.16225171e+00 5.43531060e-01 1.34049964e+00 -2.09452540e-01 -9.80172873e-01 2.71965742e-01 9.43540931e-01 -3.73855829e-01 -8.72316122e-01 6.03646517e-01 7.47412503e-01 -4.85086977e-01 9.24056590e-01 -1.08264780e+00 2.21891373e-01 -1.52744129e-01 -3.18783581e-01 -1.80700445e+00 -6.44132137e-01 -5.83493412e-01 -3.56979668e-02 7.01308310e-01 2.55556971e-01 -7.58904517e-01 5.78069627e-01 2.05044955e-01 -3.77439260e-01 -5.85127771e-01 -5.56865573e-01 -1.02051580e+00 5.12777984e-01 -9.33775678e-02 4.11729246e-01 7.91719198e-01 1.05945133e-01 1.97561070e-01 -3.01095862e-02 -1.64944027e-02 4.66158152e-01 -1.10341839e-01 1.18457377e+00 -1.02187014e+00 -7.17711568e-01 -4.52123165e-01 -1.00187294e-01 -1.10410893e+00 2.76599139e-01 -5.24502993e-01 7.38783300e-01 -1.36884999e+00 1.57195121e-01 -9.59925234e-01 -4.87149656e-01 9.58439410e-01 5.24202399e-02 -2.57090092e-01 2.30484575e-01 5.36490679e-01 -7.04377532e-01 6.55340016e-01 1.52803850e+00 -2.31952175e-01 -3.27869415e-01 -1.04054585e-01 -5.93157470e-01 9.52260137e-01 9.65820312e-01 -4.86970127e-01 -5.45496404e-01 -8.34107220e-01 9.48440582e-02 -2.87255377e-01 2.22750515e-01 -1.21497226e+00 1.84594706e-01 -6.29636765e-01 4.18653160e-01 1.77560732e-01 1.80032089e-01 -9.58094120e-01 -2.75206149e-01 7.28841960e-01 -3.64706635e-01 9.05179381e-02 5.41324496e-01 7.91579306e-01 4.17036206e-01 -3.71051222e-01 6.83912814e-01 -6.54134154e-01 -1.37631845e+00 -4.98942914e-04 -5.28070748e-01 2.41535157e-02 1.08621681e+00 -1.64183125e-01 -3.42220098e-01 -1.23123989e-01 -9.25214052e-01 3.62610102e-01 7.51577914e-01 6.86148465e-01 3.72846037e-01 -9.19813395e-01 -4.06269372e-01 1.33715302e-01 2.47528236e-02 4.10937853e-02 -2.99906373e-01 1.39570698e-01 -4.80526388e-01 3.08050394e-01 -8.18350434e-01 -6.65103555e-01 -5.12665629e-01 7.78552413e-01 3.48078370e-01 -4.53318805e-02 -6.66515470e-01 7.92292297e-01 8.74156132e-02 -7.69533515e-01 3.73716444e-01 -5.40674448e-01 2.17826426e-01 -7.17187583e-01 -5.04602939e-02 -1.40819736e-02 -3.16812396e-01 8.04876089e-02 -9.95729789e-02 4.92412806e-01 -2.34320208e-01 -2.73797303e-01 1.68320215e+00 4.59966343e-03 4.27216031e-02 5.24592698e-01 6.80710018e-01 -4.03616190e-01 -2.06505036e+00 -3.40026557e-01 2.44709253e-02 -2.28874594e-01 -4.64421928e-01 -1.10277617e+00 -6.74363077e-01 5.70975304e-01 4.67668593e-01 1.29358908e-02 7.99977064e-01 1.73016965e-01 5.54975986e-01 1.57302487e+00 1.20774221e+00 -1.28331578e+00 7.14293838e-01 9.39488351e-01 1.08784306e+00 -1.25888014e+00 -1.00826688e-01 4.57672864e-01 -5.87544560e-01 1.17536676e+00 9.35811937e-01 -6.66096091e-01 4.90321577e-01 2.11204588e-01 -1.12019159e-01 -6.37349710e-02 -8.62584412e-01 -6.68062828e-03 -3.41576964e-01 1.08003616e+00 -2.47490659e-01 -6.48756027e-02 2.32622534e-01 1.58859998e-01 -2.79302508e-01 2.27838442e-01 4.81788754e-01 1.34361708e+00 -1.06960642e+00 -1.06116223e+00 -1.30127728e-01 3.79470050e-01 1.22517914e-01 4.28515583e-01 -2.22825959e-01 1.05651498e+00 1.47211030e-01 5.12213886e-01 -1.10829622e-01 -3.25392187e-01 3.25278729e-01 5.81604354e-02 1.08649218e+00 -9.05323505e-01 -3.29168737e-01 -3.22869152e-01 -1.56822264e-01 -6.52993083e-01 -3.65619719e-01 -7.23572016e-01 -1.55776024e+00 -6.50798976e-02 8.38199481e-02 8.79668805e-04 6.03692830e-01 1.01577592e+00 3.05538058e-01 5.72778761e-01 5.18308461e-01 -1.27122819e+00 -1.28365111e+00 -1.02078021e+00 -2.25669473e-01 3.13101530e-01 5.31091511e-01 -9.46104407e-01 -1.77473351e-01 2.91677475e-01]
[4.394259452819824, 1.0371134281158447]
642d2f6a-b44c-4e6a-acf7-4fd210d9e66e
egocentric-audio-visual-object-localization
2303.13471
null
https://arxiv.org/abs/2303.13471v1
https://arxiv.org/pdf/2303.13471v1.pdf
Egocentric Audio-Visual Object Localization
Humans naturally perceive surrounding scenes by unifying sound and sight in a first-person view. Likewise, machines are advanced to approach human intelligence by learning with multisensory inputs from an egocentric perspective. In this paper, we explore the challenging egocentric audio-visual object localization task and observe that 1) egomotion commonly exists in first-person recordings, even within a short duration; 2) The out-of-view sound components can be created while wearers shift their attention. To address the first problem, we propose a geometry-aware temporal aggregation module to handle the egomotion explicitly. The effect of egomotion is mitigated by estimating the temporal geometry transformation and exploiting it to update visual representations. Moreover, we propose a cascaded feature enhancement module to tackle the second issue. It improves cross-modal localization robustness by disentangling visually-indicated audio representation. During training, we take advantage of the naturally available audio-visual temporal synchronization as the ``free'' self-supervision to avoid costly labeling. We also annotate and create the Epic Sounding Object dataset for evaluation purposes. Extensive experiments show that our method achieves state-of-the-art localization performance in egocentric videos and can be generalized to diverse audio-visual scenes.
['Chenliang Xu', 'Anurag Kumar', 'Yapeng Tian', 'Chao Huang']
2023-03-23
null
http://openaccess.thecvf.com//content/CVPR2023/html/Huang_Egocentric_Audio-Visual_Object_Localization_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Huang_Egocentric_Audio-Visual_Object_Localization_CVPR_2023_paper.pdf
cvpr-2023-1
['object-localization']
['computer-vision']
[-2.49400940e-02 -3.43069553e-01 6.07085116e-02 -2.63391227e-01 -7.25019336e-01 -6.26897931e-01 4.21813637e-01 -1.99784294e-01 -2.69577146e-01 2.44650498e-01 5.69826961e-01 6.66964412e-01 -1.20863318e-02 -1.50065497e-01 -7.85078943e-01 -5.97120821e-01 3.41318250e-02 -2.28789419e-01 1.00680172e-01 8.71429592e-02 9.78828743e-02 2.79519588e-01 -1.97254682e+00 3.60324264e-01 5.36704481e-01 1.05078125e+00 4.20860201e-01 6.41953528e-01 2.15249807e-01 8.24613690e-01 -4.81359005e-01 -9.65857729e-02 2.32836142e-01 -2.55271137e-01 -5.32702148e-01 2.80199885e-01 9.34663057e-01 -3.48772109e-01 -3.34302872e-01 1.12563789e+00 7.37775981e-01 5.25843620e-01 2.34556139e-01 -1.29013455e+00 -6.03069127e-01 3.38885725e-01 -8.22909653e-01 2.74057895e-01 7.78719246e-01 1.27905861e-01 9.66372907e-01 -1.10585749e+00 4.80998605e-01 1.15192509e+00 5.95977962e-01 4.58152801e-01 -1.26076031e+00 -5.80266237e-01 3.81227344e-01 5.68432748e-01 -1.54895031e+00 -8.27397346e-01 1.30427241e+00 -5.79693317e-01 7.22606778e-01 2.57290035e-01 7.42740989e-01 1.32288063e+00 -9.87163037e-02 8.67279112e-01 7.26410747e-01 -3.42940331e-01 2.47160241e-01 4.26422544e-02 -3.41000855e-01 3.91806453e-01 -3.25587749e-01 1.30563984e-02 -1.05197465e+00 6.18104786e-02 6.62170649e-01 1.16359457e-01 -4.96429741e-01 -7.38856494e-01 -1.47773540e+00 4.81712610e-01 5.04090130e-01 3.09997380e-01 -3.15942258e-01 3.50596756e-01 5.21169126e-01 9.01018754e-02 4.44524854e-01 4.30010736e-01 -2.53615975e-01 -2.65451282e-01 -7.04232335e-01 8.57659280e-02 2.44473189e-01 1.05153644e+00 6.56578481e-01 1.63105294e-01 -1.30421638e-01 9.24591660e-01 2.08401177e-02 4.05050218e-01 6.61767006e-01 -1.16902542e+00 4.72124398e-01 1.64166704e-01 1.96345225e-01 -1.13244271e+00 -5.01833379e-01 -5.63877106e-01 -5.11684597e-01 -9.25144702e-02 2.61850238e-01 1.76908731e-01 -3.77507657e-01 2.09480023e+00 5.50096333e-01 5.07670879e-01 -2.74841607e-01 1.32982075e+00 6.56458855e-01 1.57733485e-01 2.90447380e-02 -2.47335583e-01 1.62745380e+00 -1.08114934e+00 -9.15606678e-01 -1.73085645e-01 3.05164725e-01 -7.48251736e-01 1.20707357e+00 4.97692168e-01 -9.53428507e-01 -8.83517385e-01 -8.70544493e-01 -5.55110350e-02 -1.33472010e-01 4.10142183e-01 5.21283567e-01 3.07697773e-01 -7.55829751e-01 3.90812784e-01 -1.03370380e+00 -4.09888774e-01 1.21993586e-01 4.62777317e-02 -6.42943859e-01 3.26792873e-03 -9.94399309e-01 4.47268993e-01 2.67972410e-01 1.30705953e-01 -9.91438329e-01 -6.64476514e-01 -1.08722770e+00 -1.55954883e-01 5.64850211e-01 -9.20471072e-01 1.20092916e+00 -1.14782333e+00 -1.46254694e+00 5.62251508e-01 -4.36870962e-01 -6.82568997e-02 3.74599427e-01 -6.31376147e-01 -5.35602808e-01 6.35315478e-01 3.02094758e-01 6.25250578e-01 1.20118821e+00 -1.22161651e+00 -3.96783501e-01 -6.48707688e-01 1.36686832e-01 5.28926194e-01 -6.03014529e-01 -4.21205088e-02 -4.64673221e-01 -1.04756677e+00 2.35160053e-01 -8.45920205e-01 2.06660435e-01 7.22051561e-02 -2.56530702e-01 -9.13585573e-02 7.92265654e-01 -5.22199571e-01 9.49072361e-01 -2.30224228e+00 4.12399054e-01 -2.48580590e-01 1.62210569e-01 -2.38126412e-01 -2.00567931e-01 1.58527985e-01 -2.02550158e-01 -5.78030527e-01 2.10785910e-01 -7.92299569e-01 -1.18437119e-01 -1.77551433e-01 -3.46342027e-01 7.79130876e-01 -1.86593384e-02 7.15627253e-01 -1.19905329e+00 -3.66480857e-01 3.54435414e-01 6.50228202e-01 -7.08536923e-01 2.59843230e-01 1.18304998e-01 8.78860593e-01 -2.50983208e-01 7.78700173e-01 5.02532244e-01 -6.14731722e-02 -1.98724210e-01 -5.65012634e-01 -5.56255430e-02 4.87469248e-02 -1.28281832e+00 2.63413954e+00 -5.80276608e-01 5.87347627e-01 1.50654659e-01 -8.17682028e-01 6.21238172e-01 4.46346402e-01 6.15646899e-01 -5.80020905e-01 5.91177605e-02 -6.53760582e-02 -3.85682017e-01 -8.50596964e-01 5.18712401e-01 7.88586736e-02 -7.45243728e-02 2.72162110e-01 3.21753770e-01 5.50988615e-02 -1.92780718e-01 5.40805534e-02 8.62324119e-01 3.40561688e-01 2.92892694e-01 1.29373625e-01 6.33351028e-01 -5.14511168e-01 5.71572244e-01 5.44102907e-01 -4.25639540e-01 7.92921960e-01 3.15153487e-02 -1.87052220e-01 -5.96823275e-01 -1.27420604e+00 8.66623297e-02 1.50210941e+00 2.68652290e-01 -6.07856333e-01 -6.57537639e-01 -5.33048630e-01 -2.16857389e-01 5.82174122e-01 -6.24525011e-01 -2.30162278e-01 -4.47610229e-01 -1.21557675e-01 3.55417550e-01 7.56099403e-01 3.67546767e-01 -9.06424582e-01 -9.00159895e-01 5.54842092e-02 -6.56243920e-01 -1.39689004e+00 -8.79820287e-01 -1.81976035e-01 -3.72736424e-01 -8.89925003e-01 -9.06940162e-01 -5.17572999e-01 4.69865024e-01 7.62090445e-01 6.31572306e-01 -6.87912226e-01 -4.59227026e-01 1.12765372e+00 -4.11702484e-01 -2.06879780e-01 3.22693259e-01 -3.90894622e-01 5.18224597e-01 5.18958688e-01 3.68419081e-01 -9.79408145e-01 -8.74101043e-01 3.19152117e-01 -4.99491006e-01 -1.43240948e-04 1.46730155e-01 7.59856701e-01 5.65232575e-01 -6.57315850e-02 5.37465274e-01 -1.59654081e-01 2.55604863e-01 -3.50430548e-01 -3.35636735e-01 1.14514353e-02 1.17711343e-01 -3.49875122e-01 5.55921376e-01 -9.21640217e-01 -1.06045520e+00 4.21210319e-01 3.18955541e-01 -1.21869516e+00 -3.60632509e-01 1.51493207e-01 -4.28452343e-01 -1.98821514e-03 4.96051222e-01 1.31498843e-01 -2.29614601e-01 -6.19077802e-01 7.61059225e-01 5.24606526e-01 1.07543898e+00 -5.00890255e-01 6.28672779e-01 9.47486758e-01 -1.15843467e-01 -9.58783090e-01 -9.79540169e-01 -8.43113244e-01 -7.24368930e-01 -3.98273826e-01 9.97807503e-01 -1.38556194e+00 -9.22411501e-01 3.06285650e-01 -1.26576269e+00 2.88729556e-02 -4.05119836e-01 8.61696482e-01 -8.63572657e-01 6.48992181e-01 -2.65530229e-01 -9.25481319e-01 -5.07768430e-02 -1.07109261e+00 1.35027921e+00 6.78973123e-02 -3.89993340e-01 -6.53834343e-01 2.27707446e-01 4.04235065e-01 1.44264847e-02 1.20929338e-01 1.61623344e-01 -3.23834211e-01 -4.11186844e-01 -7.01326132e-02 -3.44925039e-02 1.50485158e-01 2.98639834e-01 -6.89621150e-01 -1.61102962e+00 -4.25606698e-01 3.25815558e-01 -2.60864854e-01 8.49483967e-01 3.85375202e-01 1.24031723e+00 -2.98499882e-01 -7.71783739e-02 8.09222460e-01 8.94884944e-01 -1.20144032e-01 1.10240862e-01 7.11234137e-02 1.06546175e+00 9.95547831e-01 7.27527916e-01 6.66977525e-01 5.33380806e-01 1.16214669e+00 5.30622959e-01 1.79611564e-01 -2.49134198e-01 -4.41852778e-01 4.20796990e-01 8.17295074e-01 -6.46940842e-02 7.99558908e-02 -4.69321340e-01 5.92510343e-01 -1.88783491e+00 -1.01567841e+00 2.98181385e-01 2.14076042e+00 6.10491574e-01 -2.99023569e-01 2.42854282e-01 1.63141131e-01 6.47049248e-01 2.39114717e-01 -6.41392946e-01 1.45513251e-01 -1.35210350e-01 -1.99030265e-01 1.21681541e-01 3.00244391e-01 -1.34806871e+00 8.53912950e-01 5.33645344e+00 6.73287988e-01 -1.22816575e+00 4.09229785e-01 -1.76519170e-01 -6.13144934e-01 -4.58257124e-02 -2.04896733e-01 -3.93081814e-01 3.65560114e-01 4.65749264e-01 1.34366468e-01 5.56726456e-01 9.02488112e-01 2.87602901e-01 2.63902545e-02 -1.26680422e+00 1.51139510e+00 5.47689915e-01 -8.49753201e-01 -2.63069898e-01 -1.63025230e-01 5.04898369e-01 -1.59805074e-01 1.55748636e-01 -1.07392795e-01 -3.31179917e-01 -6.71706498e-01 1.17652237e+00 7.12952554e-01 7.66593039e-01 -7.74647772e-01 1.61448941e-01 2.72001565e-01 -1.67605102e+00 -3.54597121e-01 -2.41679966e-01 -1.50464103e-02 2.36136377e-01 3.50212842e-01 -4.23098177e-01 6.50344312e-01 1.13597047e+00 9.88917470e-01 -6.91739917e-01 1.00524688e+00 -2.97789484e-01 9.51898694e-02 -2.39540279e-01 5.19629836e-01 -2.45463520e-01 6.25962242e-02 1.11450768e+00 1.00296164e+00 3.78174663e-01 -1.22395437e-02 1.98040649e-01 1.05665648e+00 2.94193625e-01 1.13662042e-01 -6.98384106e-01 1.66421562e-01 4.37257648e-01 1.23790205e+00 -4.16358203e-01 -6.26553968e-03 -4.13391083e-01 1.37072730e+00 2.51488477e-01 6.06660843e-01 -9.29061294e-01 -4.19866115e-01 7.08903611e-01 -1.02704175e-01 4.79443341e-01 -3.03485632e-01 1.13449380e-01 -1.50336206e+00 2.80384421e-01 -5.50334692e-01 3.70879501e-01 -1.14644146e+00 -1.14032030e+00 4.48432118e-01 -2.42128503e-02 -1.50224161e+00 -4.05238807e-01 -2.78361589e-01 -4.79159057e-01 4.41893518e-01 -1.25371087e+00 -1.42628074e+00 -6.09232605e-01 9.16763484e-01 6.63409531e-01 -8.14435855e-02 6.86710715e-01 4.95585412e-01 -4.65540975e-01 8.41491878e-01 -3.39903355e-01 -9.48646069e-02 1.12728035e+00 -1.10134327e+00 -6.05043955e-02 8.86813998e-01 4.78752822e-01 7.70935893e-01 7.66226709e-01 -4.16630447e-01 -1.47153616e+00 -1.00672925e+00 5.21075904e-01 -6.17336988e-01 7.10284114e-01 -6.53343678e-01 -7.74303794e-01 6.68219388e-01 -1.08585894e-01 2.44551793e-01 6.74848855e-01 1.02873877e-01 -7.55411446e-01 -2.97612488e-01 -7.32452393e-01 5.47474921e-01 1.26887918e+00 -1.05320168e+00 -6.27904773e-01 1.83116451e-01 8.77388656e-01 -3.35456640e-01 -6.64455175e-01 2.76239723e-01 7.85761297e-01 -1.00775635e+00 1.23261976e+00 -3.13460231e-01 6.24534450e-02 -5.42666912e-01 -4.49147254e-01 -1.21216321e+00 -2.20263556e-01 -8.70609403e-01 -4.62952614e-01 1.48168170e+00 -4.41287339e-01 -4.06430930e-01 3.36895049e-01 7.16433227e-02 -1.25177965e-01 -2.16303155e-01 -1.22865653e+00 -7.95175016e-01 -6.18150473e-01 -7.08218813e-01 4.50404942e-01 1.00460565e+00 2.78015316e-01 3.52065235e-01 -7.47633159e-01 3.01492125e-01 7.66749859e-01 1.54139295e-01 9.66907799e-01 -1.08564794e+00 -4.80857909e-01 -1.06122620e-01 -6.36710882e-01 -1.27907157e+00 3.75600427e-01 -5.81409752e-01 7.19934180e-02 -7.82512128e-01 6.12836480e-02 1.83014691e-01 -4.13940519e-01 2.40334839e-01 7.43903518e-02 4.66192871e-01 2.81229377e-01 2.63196915e-01 -8.13404858e-01 8.85998309e-01 1.16455269e+00 1.89317968e-02 -3.55449796e-01 -2.09314913e-01 -5.97312450e-01 9.44414794e-01 3.06470335e-01 -2.64949232e-01 -6.24380589e-01 -4.84947115e-01 1.86074391e-01 3.55262458e-02 8.96305084e-01 -1.09428024e+00 4.78075773e-01 -5.91790341e-02 3.25035423e-01 -6.91414595e-01 8.97074282e-01 -8.69782090e-01 -1.05081219e-03 -1.96543001e-02 -3.15139264e-01 -4.00342457e-02 1.55379489e-01 1.04421234e+00 -2.55406380e-01 1.50937244e-01 6.50126576e-01 -1.16093472e-01 -6.73995554e-01 1.34018138e-01 -1.05199322e-01 2.91184783e-02 9.35838580e-01 -1.73628300e-01 -1.86061740e-01 -6.34542108e-01 -8.85219097e-01 1.23157956e-01 4.93658006e-01 6.84178889e-01 5.83580911e-01 -1.60285163e+00 -2.34458789e-01 3.13631594e-01 4.32010859e-01 -2.38487348e-01 8.44035506e-01 1.20944333e+00 7.88627043e-02 3.13752323e-01 -1.60895348e-01 -1.02533138e+00 -1.13564932e+00 8.02008569e-01 2.81123728e-01 5.03488004e-01 -8.68228555e-01 9.91323948e-01 8.90017152e-01 -2.39463095e-02 5.55339754e-01 -2.41548821e-01 -3.00329089e-01 3.22665632e-01 7.96147287e-01 5.27743101e-01 -1.62920162e-01 -1.02872372e+00 -4.69368190e-01 8.73369396e-01 2.14061335e-01 -2.76781201e-01 1.16187334e+00 -7.22216904e-01 6.52414560e-02 9.02360976e-01 1.30409193e+00 3.22534084e-01 -1.48127317e+00 -3.75669688e-01 -3.83768976e-01 -7.13275135e-01 2.24882290e-02 -3.47280473e-01 -9.52462375e-01 1.13706708e+00 7.68558860e-01 -5.77488169e-02 1.31090331e+00 1.48126766e-01 5.94949961e-01 3.12984318e-01 2.87570804e-01 -1.06065571e+00 6.70409977e-01 2.08931610e-01 1.01522994e+00 -1.02344513e+00 -4.46929485e-02 -3.74087155e-01 -8.27352703e-01 9.99010444e-01 6.03831589e-01 2.59255003e-02 5.40460706e-01 -1.75893843e-01 -7.42688263e-03 -1.63675860e-01 -5.78008652e-01 -3.48190874e-01 6.20141149e-01 8.27323198e-01 6.37212694e-02 -1.16047844e-01 4.75135416e-01 8.24414551e-01 -4.12789255e-01 -3.43086898e-01 2.44201943e-01 7.12977350e-01 -1.69861913e-01 -4.28122133e-01 -4.99077111e-01 -4.50913876e-01 -3.53691429e-01 1.25323951e-01 -2.78796732e-01 4.74638462e-01 2.68600404e-01 9.94869828e-01 1.25143096e-01 -4.10385609e-01 4.41093296e-01 1.29255772e-01 7.67941117e-01 -5.04750550e-01 -1.99379846e-01 6.63651764e-01 -2.77688831e-01 -9.61420298e-01 -7.00979769e-01 -7.61460483e-01 -8.49359512e-01 2.88450569e-01 -3.69485408e-01 -6.40816940e-03 5.93146205e-01 7.72853017e-01 5.00968456e-01 7.08768427e-01 5.91105878e-01 -1.30037355e+00 -4.28588569e-01 -8.48531783e-01 -7.90036917e-01 5.01051605e-01 7.50978947e-01 -1.07442808e+00 -6.15788221e-01 2.23527744e-01]
[14.762530326843262, 4.944378852844238]
cdf38deb-0ee5-4125-94e4-d3153c69c9a0
document-image-cleaning-using-budget-aware
2306.13236
null
https://arxiv.org/abs/2306.13236v1
https://arxiv.org/pdf/2306.13236v1.pdf
Document Image Cleaning using Budget-Aware Black-Box Approximation
Recent work has shown that by approximating the behaviour of a non-differentiable black-box function using a neural network, the black-box can be integrated into a differentiable training pipeline for end-to-end training. This methodology is termed "differentiable bypass,'' and a successful application of this method involves training a document preprocessor to improve the performance of a black-box OCR engine. However, a good approximation of an OCR engine requires querying it for all samples throughout the training process, which can be computationally and financially expensive. Several zeroth-order optimization (ZO) algorithms have been proposed in black-box attack literature to find adversarial examples for a black-box model by computing its gradient in a query-efficient manner. However, the query complexity and convergence rate of such algorithms makes them infeasible for our problem. In this work, we propose two sample selection algorithms to train an OCR preprocessor with less than 10% of the original system's OCR engine queries, resulting in more than 60% reduction of the total training time without significant loss of accuracy. We also show an improvement of 4% in the word-level accuracy of a commercial OCR engine with only 2.5% of the total queries and a 32x reduction in monetary cost. Further, we propose a simple ranking technique to prune 30% of the document images from the training dataset without affecting the system's performance.
['Nilanjan Ray', 'Eric Van Oeveren', 'Katyani Singh', 'Ganesh Tata']
2023-06-22
null
null
null
null
['optical-character-recognition']
['computer-vision']
[ 3.40815216e-01 -4.38552059e-04 -1.70018837e-01 -2.65597284e-01 -1.17310202e+00 -1.02850068e+00 3.78323793e-01 1.30697027e-01 -5.75780809e-01 3.15021843e-01 -5.63799322e-01 -8.42755437e-01 2.20756024e-01 -6.62826002e-01 -1.12353194e+00 -3.28082323e-01 2.77032822e-01 2.38618150e-01 1.82556763e-01 -2.48161316e-01 4.89345253e-01 9.22662795e-01 -1.14792895e+00 3.57896328e-01 9.86511528e-01 9.36588466e-01 8.04424379e-03 1.22943389e+00 6.64715692e-02 5.00865459e-01 -1.46892619e+00 -1.06925011e+00 5.68079054e-01 -1.20663978e-01 -5.91832161e-01 -2.38857448e-01 6.85013652e-01 -6.89461291e-01 -5.84641635e-01 1.10230494e+00 5.22121489e-01 -6.78229928e-02 5.13356626e-01 -9.50086057e-01 -5.15687525e-01 3.34039241e-01 -2.81732023e-01 -1.51996732e-01 2.48428345e-01 2.60308176e-01 9.72538054e-01 -7.53569722e-01 3.09178680e-01 9.84107196e-01 4.23622489e-01 6.59739137e-01 -1.03480268e+00 -7.50969589e-01 -2.94997633e-01 -3.51353735e-02 -1.24011350e+00 -3.43771458e-01 4.89345729e-01 6.39014915e-02 1.05253553e+00 8.33278179e-01 3.57558250e-01 6.52982950e-01 1.89065903e-01 1.06992626e+00 8.59463334e-01 -4.57489222e-01 2.38864720e-01 3.74714345e-01 -5.37148230e-02 4.93072510e-01 3.39703321e-01 2.05907226e-01 -2.70104676e-01 -1.83231860e-01 4.15046334e-01 -5.50282151e-02 -9.23146605e-02 2.42876932e-01 -3.69562924e-01 6.61111355e-01 4.31784481e-01 7.19701918e-03 1.73646972e-01 4.54853863e-01 2.95134932e-01 4.83483791e-01 3.81400585e-01 9.68682230e-01 -2.26609826e-01 -1.83504447e-01 -1.15686584e+00 1.43692225e-01 9.61459041e-01 8.84637058e-01 7.54877210e-01 1.85833007e-01 -8.69753882e-02 6.13555729e-01 8.81255046e-02 6.83632016e-01 3.36652994e-01 -6.89511120e-01 6.48078144e-01 6.90470278e-01 9.80845690e-02 -8.34293842e-01 2.26111680e-01 -4.35007125e-01 -3.46705049e-01 6.31179094e-01 6.81957304e-01 -1.03377692e-01 -1.10884678e+00 1.00364590e+00 1.85979337e-01 -2.30073079e-01 -1.04288012e-02 7.56640017e-01 1.58915445e-02 8.44679117e-01 -2.71176457e-01 2.58357406e-01 1.33234489e+00 -9.82422948e-01 -4.48120654e-01 -3.67001176e-01 6.48829103e-01 -1.09550536e+00 1.43714893e+00 7.76794553e-01 -1.11354816e+00 -3.75237823e-01 -1.63200152e+00 -1.92729190e-01 -6.01755798e-01 2.68194169e-01 4.65046227e-01 1.29234338e+00 -6.72567964e-01 8.70027006e-01 -5.34435868e-01 4.14099216e-01 4.21775043e-01 6.41870558e-01 -7.12351799e-02 -1.22465566e-01 -9.93059456e-01 8.17523062e-01 2.52086133e-01 -3.30687091e-02 -1.09702206e+00 -9.52032566e-01 -4.77380961e-01 3.02827835e-01 4.37483847e-01 -2.09423993e-02 1.29380870e+00 -9.08136189e-01 -1.79197323e+00 5.07098913e-01 -7.04134032e-02 -4.92273688e-01 9.04187024e-01 -7.77811885e-01 -3.62717927e-01 3.41485202e-01 -5.56795895e-01 2.61296093e-01 1.43649280e+00 -1.02848971e+00 -5.40558159e-01 -1.62028730e-01 2.47641250e-01 -3.28149386e-02 -5.07265508e-01 1.58033803e-01 -7.21865356e-01 -8.74389768e-01 -4.14499849e-01 -9.22303081e-01 -1.84837386e-01 5.49785001e-03 -5.66134334e-01 6.69685528e-02 1.07500291e+00 -9.44186687e-01 1.34612906e+00 -2.38637447e+00 -4.42763537e-01 3.41989934e-01 -1.32574245e-01 6.78301811e-01 -1.56963974e-01 3.48360002e-01 2.02695280e-02 6.15519106e-01 -3.03560555e-01 -5.35414279e-01 9.18644667e-03 -1.65480942e-01 -8.76777172e-01 6.07132852e-01 4.11106229e-01 9.45472062e-01 -8.57681990e-01 -1.03363864e-01 4.84583080e-02 4.55065459e-01 -6.20769799e-01 3.93924534e-01 -4.70463246e-01 -1.96588904e-01 4.89376253e-03 8.79178762e-01 5.58446527e-01 2.18760788e-01 -2.68610138e-02 4.80582081e-02 1.81566119e-01 3.53860557e-01 -9.53742206e-01 1.14792633e+00 -6.72413707e-01 1.06388021e+00 -2.41037980e-01 -5.91501772e-01 9.45119619e-01 1.04597628e-01 -1.25317290e-01 -6.67384624e-01 4.26958539e-02 5.20328581e-01 -1.62927091e-01 2.32723001e-02 9.20353830e-01 3.61279517e-01 -1.51662096e-01 5.25097132e-01 -3.25524360e-01 -4.20513272e-01 1.00117691e-01 4.71006669e-02 1.14553452e+00 -4.71456349e-02 -2.67137527e-01 2.88037747e-01 5.45859516e-01 9.51962173e-02 9.27381665e-02 9.54599857e-01 1.15223534e-01 5.96890807e-01 5.15860021e-01 -2.46820092e-01 -1.38660789e+00 -7.72435665e-01 7.48701766e-02 1.03945267e+00 -7.48503730e-02 -4.09647226e-01 -1.03605223e+00 -8.56334090e-01 8.12431052e-02 9.67465758e-01 -2.93331981e-01 -3.82209033e-01 -9.35216129e-01 -5.84336162e-01 1.12400508e+00 2.71435022e-01 5.20437777e-01 -8.26508403e-01 -5.09396911e-01 -9.01412442e-02 3.64630371e-01 -9.11974907e-01 -7.95094252e-01 8.05646405e-02 -8.19818318e-01 -9.95695889e-01 -7.95128226e-01 -4.42441106e-01 8.94590914e-01 5.69170080e-02 7.62959421e-01 4.10487831e-01 -4.98243928e-01 -3.20934542e-02 -2.98087507e-01 -6.17365956e-01 -8.76818895e-01 1.25838727e-01 -2.35151857e-01 -4.93049622e-02 2.63751507e-01 -8.08143429e-03 -6.89507723e-01 3.97021383e-01 -1.21871400e+00 -5.05599976e-01 6.98088229e-01 7.63452470e-01 4.98931885e-01 3.45188707e-01 2.05836803e-01 -8.38121295e-01 7.37401426e-01 -6.67769648e-03 -1.18509734e+00 3.48875135e-01 -8.95183504e-01 2.59968132e-01 1.11446476e+00 -8.43816698e-01 -7.78191209e-01 1.44707277e-01 -1.10381478e-02 -6.43228412e-01 3.43527585e-01 9.02655497e-02 -2.24587008e-01 -3.78091037e-01 7.57839739e-01 4.18401629e-01 1.52412280e-02 -4.16859716e-01 4.49532777e-01 8.65415931e-01 6.77472830e-01 -5.20050347e-01 1.41387630e+00 3.52271408e-01 -8.34198296e-02 -6.57510400e-01 -5.53018212e-01 -3.13222319e-01 -9.77171063e-02 -1.96637809e-02 4.41079706e-01 -4.41120654e-01 -1.03327930e+00 6.39775217e-01 -1.07765651e+00 -4.37937438e-01 -3.59619707e-01 1.15694217e-01 -2.58056968e-01 5.44819593e-01 -4.91397619e-01 -9.86697257e-01 -7.13064551e-01 -1.39769888e+00 9.89026070e-01 2.93493003e-01 3.80183011e-02 -5.97212136e-01 -2.40235344e-01 3.65416855e-01 4.04102504e-01 -9.52201895e-03 8.78377020e-01 -9.58935738e-01 -8.05660486e-01 -1.01268196e+00 -1.34338990e-01 8.88508737e-01 -2.03844115e-01 2.04226837e-01 -1.05352902e+00 -5.51690996e-01 1.12819880e-01 -1.45843685e-01 4.93913323e-01 -2.49733657e-01 1.33179259e+00 -5.21635830e-01 -8.70200992e-02 7.23763645e-01 1.51431739e+00 4.31479782e-01 9.48661447e-01 4.71070915e-01 5.70848346e-01 3.30244988e-01 8.74621272e-01 1.52807012e-01 -5.19624293e-01 4.63337034e-01 5.05757630e-01 1.18734175e-02 -5.18842451e-02 -4.63591576e-01 8.47386003e-01 3.12916726e-01 3.24056745e-01 -3.10164839e-01 -7.08528697e-01 2.40825266e-01 -1.16522098e+00 -8.24615180e-01 2.55990833e-01 2.67193055e+00 1.01815605e+00 4.63274628e-01 -1.36123300e-02 3.65655869e-01 7.02933133e-01 6.96921945e-02 -7.61289477e-01 -8.29156160e-01 1.26268774e-01 5.27901292e-01 1.05182290e+00 5.92932522e-01 -9.56872582e-01 1.05801773e+00 6.78435564e+00 1.14243877e+00 -1.22778845e+00 -2.59789377e-01 7.62016952e-01 -2.24739447e-01 -2.92138189e-01 2.10401922e-01 -1.07974470e+00 5.30870974e-01 1.39622056e+00 -2.04224184e-01 8.40564430e-01 1.03540349e+00 -1.36681780e-01 -3.48088183e-02 -1.15252876e+00 8.56072068e-01 8.29975009e-02 -1.20727623e+00 5.07774018e-02 2.89759994e-01 4.04924542e-01 -4.20710742e-01 3.48913282e-01 3.93846035e-01 2.85511762e-01 -1.19623339e+00 6.61648452e-01 2.10602641e-01 1.08506417e+00 -1.14540172e+00 4.70915735e-01 2.70075262e-01 -7.66132057e-01 -2.58600801e-01 -4.02758151e-01 4.18748677e-01 -1.13232434e-01 3.89962852e-01 -1.18261909e+00 3.31041783e-01 5.13014495e-01 -2.21540779e-01 -8.25505018e-01 1.05182791e+00 -3.89926881e-01 8.24904740e-01 -4.81533706e-01 -4.88704920e-01 2.17493907e-01 -1.35035306e-01 6.22929037e-01 1.16358519e+00 1.63628191e-01 -2.99983293e-01 -4.27904069e-01 7.09455907e-01 -4.98065293e-01 -1.86757371e-02 -3.01514149e-01 -4.23688799e-01 5.30683875e-01 1.07682920e+00 -2.16067299e-01 -3.44607234e-01 -5.92921153e-02 1.16739655e+00 7.46939331e-02 2.65522420e-01 -9.65988696e-01 -1.19030941e+00 5.05559802e-01 5.36564514e-02 1.76759049e-01 -2.02065006e-01 -4.32808697e-01 -7.97691584e-01 2.49711871e-01 -1.33209503e+00 1.45069227e-01 -4.86533701e-01 -9.37357843e-01 4.22005683e-01 -2.99504191e-01 -1.19718480e+00 -3.31158906e-01 -8.48039269e-01 -5.91451585e-01 1.08766651e+00 -1.33418810e+00 -6.52730405e-01 -5.57311438e-02 4.08950597e-01 5.66276252e-01 -4.18017089e-01 6.82813466e-01 3.06733191e-01 -6.52934253e-01 1.40577066e+00 4.25728858e-01 4.36545104e-01 6.73481405e-01 -1.45214987e+00 8.17101181e-01 1.12552083e+00 1.69134066e-01 7.66582251e-01 6.56180322e-01 -6.48981869e-01 -1.84335279e+00 -9.69954312e-01 6.03906870e-01 -4.07610834e-01 6.53582454e-01 -6.15971923e-01 -8.31520438e-01 2.68712819e-01 -4.37355004e-02 -2.04434738e-01 5.14775455e-01 -3.30207080e-01 -5.61716318e-01 -3.55808288e-01 -1.34436643e+00 9.98237252e-01 3.46701145e-01 -8.42902303e-01 -3.89450341e-01 5.34284711e-01 8.26040328e-01 -5.90912282e-01 -7.48023748e-01 1.27441838e-01 6.29117608e-01 -5.47705054e-01 9.70948517e-01 -6.03189170e-01 3.88806731e-01 -2.08769903e-01 -6.98178485e-02 -9.30979550e-01 3.02627146e-01 -1.19955063e+00 -6.75003469e-01 1.03355229e+00 4.48561996e-01 -7.19757915e-01 1.09163105e+00 8.63004088e-01 -2.77753100e-02 -9.72527146e-01 -8.40336144e-01 -9.77540016e-01 9.97580364e-02 -6.22905254e-01 5.09782255e-01 2.82554418e-01 -3.63572627e-01 -2.55959239e-02 -2.90657580e-01 4.75663275e-01 5.23556530e-01 -2.32332826e-01 9.69987631e-01 -4.90895808e-01 -7.26359367e-01 -4.16049212e-01 -2.19829410e-01 -1.19728220e+00 -2.63254307e-02 -7.92592108e-01 -4.20679860e-02 -9.00844812e-01 -2.36591518e-01 -3.54193181e-01 -7.07186088e-02 3.52182835e-01 -2.96954870e-01 1.79528311e-01 4.48554724e-01 3.43595892e-01 1.73486978e-01 6.46839887e-02 9.65784550e-01 -5.61990559e-01 -2.26635426e-01 2.44373873e-01 -4.54002947e-01 2.47808412e-01 8.30626011e-01 -6.38852537e-01 -4.46575403e-01 -4.55192238e-01 2.33075529e-01 -1.68012470e-01 1.78073928e-01 -9.53610241e-01 3.07400674e-01 1.29549026e-01 4.79067475e-01 -6.74596608e-01 4.66732800e-01 -9.45340395e-01 -1.33473173e-01 7.74235666e-01 -2.86936402e-01 9.27827656e-02 2.90162593e-01 4.68508065e-01 -6.95543513e-02 -9.67910230e-01 8.47377658e-01 2.60997534e-01 -1.14839964e-01 1.14116311e-01 -1.91229358e-01 -2.04971209e-01 1.07521844e+00 -3.08134884e-01 -5.06874025e-01 -2.65708953e-01 -1.56992465e-01 -2.21196458e-01 4.48044389e-01 2.41738573e-01 6.04911506e-01 -8.48339975e-01 -3.95594060e-01 9.01724026e-02 -3.84420574e-01 -3.41387093e-02 -2.12020263e-01 1.90350622e-01 -1.13970411e+00 5.08482516e-01 3.05631071e-01 -2.12454170e-01 -1.32187390e+00 6.98437333e-01 5.47712803e-01 -5.19236445e-01 -3.05538118e-01 7.39198387e-01 -4.03525978e-01 -1.14821546e-01 4.44957852e-01 8.62407982e-02 3.68931472e-01 -1.98147029e-01 7.54377306e-01 4.66119796e-01 2.62387961e-01 5.06568849e-02 -9.50070247e-02 1.97352812e-01 -4.86110598e-01 -3.97016943e-01 1.00932014e+00 5.70432544e-01 8.93007070e-02 -5.61586395e-03 1.49293065e+00 4.94424820e-01 -1.18984795e+00 1.56068742e-01 -5.85933812e-02 -8.17677081e-01 4.01798822e-02 -8.70606124e-01 -9.40904260e-01 1.07736230e+00 5.86360812e-01 1.96896732e-01 1.24758816e+00 -5.60072601e-01 1.02043247e+00 6.85384035e-01 1.52600884e-01 -1.16259241e+00 3.55250299e-01 3.00561458e-01 8.50538850e-01 -9.00312960e-01 2.44639844e-01 -2.75573999e-01 -4.00462985e-01 1.37275934e+00 4.65938807e-01 -3.95283878e-01 2.21613765e-01 2.54071653e-01 -8.38840455e-02 2.22587481e-01 -3.77515882e-01 4.20767128e-01 3.50485355e-01 2.86296725e-01 -1.30709052e-01 -2.54336447e-01 -7.84168914e-02 3.00841331e-01 -3.49731237e-01 -3.64902705e-01 4.21875805e-01 1.02814841e+00 -2.67281413e-01 -1.26523018e+00 -5.67099512e-01 3.52390856e-01 -9.39046323e-01 -3.74483883e-01 -4.94586378e-01 7.48015881e-01 -5.03187656e-01 9.65789557e-01 -5.49118817e-02 -5.71006536e-01 4.42273468e-01 1.64061517e-01 1.98208481e-01 -3.59725386e-01 -1.00699353e+00 -7.64535964e-02 3.03967800e-02 -5.24596751e-01 4.97909904e-01 -1.80356249e-01 -9.91186917e-01 -4.96085852e-01 -5.07763326e-01 -2.78090574e-02 1.12491477e+00 6.54241204e-01 3.39269757e-01 3.53528708e-01 1.06782293e+00 -5.56374192e-01 -1.29679585e+00 -6.95229769e-01 -1.14924282e-01 1.74447715e-01 2.33187869e-01 6.64255023e-02 -5.20873547e-01 -1.27054796e-01]
[5.894392490386963, 8.029037475585938]
9c6a7c25-a721-4e44-9530-8f75806d6194
deeply-interleaved-two-stream-encoder-for
2203.15969
null
https://arxiv.org/abs/2203.15969v1
https://arxiv.org/pdf/2203.15969v1.pdf
Deeply Interleaved Two-Stream Encoder for Referring Video Segmentation
Referring video segmentation aims to segment the corresponding video object described by the language expression. To address this task, we first design a two-stream encoder to extract CNN-based visual features and transformer-based linguistic features hierarchically, and a vision-language mutual guidance (VLMG) module is inserted into the encoder multiple times to promote the hierarchical and progressive fusion of multi-modal features. Compared with the existing multi-modal fusion methods, this two-stream encoder takes into account the multi-granularity linguistic context, and realizes the deep interleaving between modalities with the help of VLGM. In order to promote the temporal alignment between frames, we further propose a language-guided multi-scale dynamic filtering (LMDF) module to strengthen the temporal coherence, which uses the language-guided spatial-temporal features to generate a set of position-specific dynamic filters to more flexibly and effectively update the feature of current frame. Extensive experiments on four datasets verify the effectiveness of the proposed model.
['Huchuan Lu', 'Zhiwei Hu', 'Lihe Zhang', 'Guang Feng']
2022-03-30
null
null
null
null
['referring-expression-segmentation']
['computer-vision']
[ 2.14317609e-02 -2.42464855e-01 -2.68515646e-01 -5.02619624e-01 -5.66662669e-01 -1.01482503e-01 6.40358567e-01 -7.06065744e-02 -5.51464736e-01 3.13543439e-01 5.19790590e-01 2.15467662e-02 -3.31138335e-02 -9.25903320e-01 -5.32251239e-01 -7.09091485e-01 3.86623174e-01 -2.17800051e-01 7.80076206e-01 -1.56166881e-01 2.75560856e-01 2.02792376e-01 -1.60610199e+00 6.16563201e-01 8.73354375e-01 1.04930735e+00 7.75967300e-01 3.56112838e-01 -4.02345121e-01 1.14792907e+00 -9.86657292e-02 -9.70589519e-02 -1.08142737e-02 -5.21800160e-01 -7.75623441e-01 5.75646639e-01 1.29752308e-01 -5.93269587e-01 -5.12585402e-01 1.17180908e+00 3.66121799e-01 4.38477606e-01 1.02996014e-01 -1.08982170e+00 -5.89686573e-01 6.22328460e-01 -6.18487895e-01 4.34303790e-01 4.69458431e-01 3.22951078e-01 8.10002148e-01 -6.60829186e-01 5.56395352e-01 1.59876788e+00 2.93993592e-01 2.71944284e-01 -7.02200711e-01 -5.99733114e-01 6.35189831e-01 5.43221354e-01 -1.24247229e+00 -4.21222866e-01 1.00396252e+00 -2.83674151e-01 5.69850087e-01 -2.39685318e-03 7.66915202e-01 6.52414978e-01 6.00183830e-02 1.01254058e+00 8.39748502e-01 -2.47406557e-01 -2.87006199e-01 -2.11819828e-01 -6.25854954e-02 9.90205824e-01 -3.61591429e-01 -2.50563794e-03 -4.92598534e-01 2.95488626e-01 1.05085468e+00 2.63717085e-01 -4.04942125e-01 7.38435760e-02 -1.56972885e+00 5.82717836e-01 3.56761903e-01 7.59468019e-01 -4.97748137e-01 1.17902689e-01 5.31409264e-01 -3.36264186e-02 2.14546323e-01 -1.44661486e-01 -3.35237056e-01 3.29436250e-02 -1.13295317e+00 -1.73632056e-01 2.33392239e-01 9.47481692e-01 1.15019715e+00 6.14247061e-02 -5.75448155e-01 6.50499761e-01 7.58319855e-01 3.15638989e-01 6.73000395e-01 -1.28557849e+00 4.01470542e-01 7.92152107e-01 -9.34655815e-02 -1.16653514e+00 -2.83281326e-01 -1.52000606e-01 -8.03981841e-01 -2.01214105e-01 -2.87401732e-02 2.49269605e-02 -9.29782867e-01 1.77548575e+00 4.88281339e-01 5.83802819e-01 3.49742621e-02 1.21944523e+00 9.40157652e-01 9.16273892e-01 2.98213661e-01 -7.17282414e-01 1.57412040e+00 -9.88846302e-01 -9.64026392e-01 7.48924315e-02 4.24691349e-01 -9.09074664e-01 8.26047838e-01 -7.53943548e-02 -1.17913556e+00 -9.75804150e-01 -9.29757833e-01 -4.18411106e-01 -7.70958364e-02 1.85284823e-01 3.36578012e-01 -1.26631930e-01 -8.59208465e-01 1.82317361e-01 -7.82931745e-01 -1.27034724e-01 2.86172152e-01 2.04090968e-01 -2.92821705e-01 -3.20954174e-02 -1.58947039e+00 6.71873033e-01 8.05620790e-01 3.95971090e-01 -7.04496741e-01 -2.86392063e-01 -1.12449241e+00 -1.26696900e-01 5.35467148e-01 -7.16750622e-01 9.68682170e-01 -1.37242460e+00 -1.48144388e+00 6.23385549e-01 -4.49802428e-01 -2.55520582e-01 3.59994173e-01 -6.24953862e-03 -3.84709686e-01 5.31877697e-01 2.08119050e-01 9.85076308e-01 8.10918629e-01 -1.20813560e+00 -1.27907217e+00 -1.98410347e-01 4.45048571e-01 7.31296241e-01 -3.92804325e-01 1.51952401e-01 -1.17694032e+00 -7.08151639e-01 2.30035350e-01 -4.50535864e-01 -1.52967572e-01 -1.82793006e-01 -1.91348240e-01 -1.99549079e-01 1.03895390e+00 -8.39769363e-01 1.54538381e+00 -2.17094731e+00 5.16098380e-01 -6.28642961e-02 1.82525218e-01 2.38309652e-01 -7.41432235e-02 -9.02997479e-02 1.82794541e-01 -2.68790871e-01 -2.42332652e-01 -2.53380239e-01 -3.55194092e-01 4.36334193e-01 -2.02461723e-02 2.67944485e-01 1.15909003e-01 9.35388088e-01 -9.85104680e-01 -1.06672645e+00 6.49902821e-01 5.21093786e-01 -4.92655456e-01 2.59612501e-01 -3.86089802e-01 5.08336127e-01 -7.52983809e-01 4.93137389e-01 5.40960312e-01 -2.26564452e-01 -5.35795428e-02 -8.50121558e-01 -3.90738547e-01 -1.67316198e-02 -1.17358494e+00 2.07412195e+00 -3.89775634e-01 2.55890071e-01 5.86319789e-02 -9.44881439e-01 7.95412779e-01 2.79833794e-01 7.53799677e-01 -7.99906313e-01 3.89656752e-01 -5.93747050e-02 -2.06858769e-01 -7.63479471e-01 4.72470075e-01 1.04489006e-01 3.65454033e-02 8.68939236e-02 2.23290011e-01 2.65133053e-01 2.69062072e-01 2.21003622e-01 5.21028399e-01 3.61336261e-01 1.22856252e-01 -2.03561947e-01 1.34442997e+00 -2.80441701e-01 9.49873567e-01 6.19577095e-02 -3.23297381e-01 3.50471914e-01 7.90500715e-02 -3.69616777e-01 -7.63005078e-01 -6.02147698e-01 2.79964238e-01 1.08135843e+00 7.53323257e-01 -4.39620405e-01 -6.62323296e-01 -4.93548512e-01 -4.03722495e-01 4.55142677e-01 -5.24060786e-01 -3.01689804e-01 -6.15199506e-01 -5.20970941e-01 2.72507161e-01 5.22797883e-01 1.06459296e+00 -1.04404068e+00 -5.97664773e-01 4.43116069e-01 -6.15252018e-01 -1.27929807e+00 -7.15869844e-01 -1.80089533e-01 -5.11816680e-01 -9.67827976e-01 -6.34635508e-01 -1.21646369e+00 4.18194503e-01 3.79451513e-01 5.78676701e-01 2.47097537e-01 2.12344244e-01 3.05502057e-01 -4.88792390e-01 3.03574681e-01 -1.68293521e-01 -1.09719425e-01 -2.24460855e-01 3.99874270e-01 2.87658662e-01 -5.34630537e-01 -5.63765883e-01 2.08791226e-01 -1.14444470e+00 6.61930382e-01 5.09662390e-01 8.24777901e-01 7.42974162e-01 2.11819142e-01 2.87978470e-01 -3.52188110e-01 3.34306598e-01 -1.53743625e-01 -5.40930450e-01 4.98711973e-01 -1.39247373e-01 5.23973294e-02 4.86726850e-01 -5.00991344e-01 -1.36999011e+00 1.16604947e-01 -1.22742578e-01 -7.91454554e-01 -1.50953084e-01 5.42378485e-01 -5.30488372e-01 1.12857744e-01 -1.78338706e-01 6.83191657e-01 -9.51654091e-02 -2.73208708e-01 7.94404626e-01 6.13423347e-01 8.51884842e-01 -5.63171983e-01 3.74503285e-01 4.74988878e-01 -1.98896125e-01 -4.41756815e-01 -8.79691720e-01 -3.03303450e-01 -8.22913408e-01 -4.28451449e-01 1.53685284e+00 -1.14235151e+00 -7.45050192e-01 5.82270622e-01 -1.34651005e+00 -9.46883932e-02 -1.02821864e-01 5.62767267e-01 -7.05437601e-01 6.82380557e-01 -9.07373846e-01 -4.68203276e-01 -2.67142832e-01 -1.61411238e+00 1.20504081e+00 6.26618505e-01 3.06298226e-01 -8.53749514e-01 -3.84778976e-01 2.94626862e-01 9.90499649e-03 3.68735529e-02 6.33128047e-01 -2.36414388e-01 -7.45361328e-01 4.41098303e-01 -4.88227338e-01 2.98076183e-01 1.83494717e-01 2.22124815e-01 -5.01342237e-01 -3.84107605e-02 1.31090805e-01 1.97121128e-03 9.79272723e-01 4.42414254e-01 1.11476743e+00 -4.16382290e-02 -2.13648796e-01 6.62433743e-01 1.26266468e+00 4.38913822e-01 5.27218878e-01 3.60109419e-01 1.04340470e+00 5.22042632e-01 9.26041961e-01 3.99347961e-01 8.46898198e-01 5.61946869e-01 3.92155170e-01 -1.16869472e-01 -1.10882767e-01 -2.35357702e-01 3.74727935e-01 1.22638690e+00 -1.62098110e-01 -3.15735936e-02 -4.21282351e-01 3.81790727e-01 -2.11983681e+00 -1.18717551e+00 3.23389890e-03 1.63837469e+00 8.66181791e-01 8.56167749e-02 -1.76976286e-02 -7.59179518e-02 9.77213204e-01 3.83239120e-01 -3.95556122e-01 2.04938456e-01 -3.57542336e-01 -4.96516436e-01 2.55333155e-01 5.56465983e-01 -1.16077638e+00 1.11999929e+00 5.25660944e+00 1.25083756e+00 -1.36743307e+00 2.73709059e-01 5.97612023e-01 1.02500454e-01 -4.46959227e-01 1.89031854e-01 -6.34893239e-01 6.49354398e-01 3.87088865e-01 -1.63907424e-01 3.90786320e-01 4.27143663e-01 5.10558665e-01 -9.04998854e-02 -7.22923875e-01 1.08814669e+00 3.02084718e-05 -1.55638111e+00 2.55428761e-01 -1.42352834e-01 6.09650671e-01 -9.82160345e-02 -1.76804721e-01 6.60172403e-02 -3.18538994e-02 -4.48285818e-01 1.10998333e+00 9.72862363e-01 6.23900890e-01 -9.51993346e-01 5.52025378e-01 2.97440469e-01 -1.81264842e+00 -2.03037813e-01 -2.65469551e-01 2.62750924e-01 5.58842480e-01 5.03149152e-01 1.71033870e-02 9.29304242e-01 7.25123525e-01 1.17127550e+00 -4.71807957e-01 6.00694537e-01 -6.70718551e-02 1.81575060e-01 -2.25852773e-01 4.27085340e-01 4.67473328e-01 -2.70523548e-01 4.69657481e-01 1.22234404e+00 2.09836856e-01 3.56998473e-01 5.91492057e-01 6.46102130e-01 2.38886252e-01 7.41574541e-02 -5.60988374e-02 3.25536951e-02 4.89329278e-01 1.35087705e+00 -6.67950332e-01 -7.46489465e-01 -7.48853445e-01 1.13518500e+00 2.57954672e-02 3.52427661e-01 -1.10806751e+00 -2.40314573e-01 3.18735033e-01 -1.74924821e-01 3.95772815e-01 -2.43298531e-01 1.66051969e-01 -1.37677276e+00 -3.46464850e-03 -7.95339286e-01 4.32629526e-01 -1.01245439e+00 -9.30449545e-01 6.04870617e-01 -8.99091139e-02 -1.43623519e+00 -1.36728168e-01 -1.36466831e-01 -4.32161003e-01 8.83799016e-01 -1.64078927e+00 -1.59260869e+00 -3.29052925e-01 1.12599742e+00 7.65714228e-01 -8.03665593e-02 2.70882875e-01 6.01792514e-01 -6.20203018e-01 1.18475251e-01 -4.34930086e-01 2.03798249e-01 5.68561554e-01 -6.92444146e-01 -2.89055020e-01 1.11130261e+00 -9.68843028e-02 5.81883788e-01 3.86435926e-01 -6.97805107e-01 -1.28029358e+00 -1.21560669e+00 6.63403392e-01 1.64693013e-01 6.25564396e-01 1.86478108e-01 -1.03243804e+00 6.38206601e-01 3.10193270e-01 -1.49238529e-02 2.90599257e-01 -6.02299213e-01 -1.21673472e-01 -2.35140011e-01 -7.72854209e-01 5.18772960e-01 9.95591640e-01 -7.45760798e-01 -8.40841770e-01 6.04989044e-02 1.28539252e+00 -5.02500534e-01 -9.37951684e-01 7.57536471e-01 4.00882632e-01 -1.04389572e+00 8.30965042e-01 -1.26747787e-01 4.92167324e-01 -9.59959388e-01 -3.37206572e-01 -7.09606707e-01 -5.09312451e-01 -4.55147624e-01 2.16457788e-02 1.61587596e+00 -2.57544309e-01 -2.47714818e-01 2.77545094e-01 2.10991666e-01 -2.24731520e-01 -8.15222502e-01 -8.33571076e-01 -1.49457246e-01 -4.80925292e-01 -4.89448279e-01 6.15285754e-01 9.98068929e-01 -3.89873795e-02 2.80553877e-01 -5.16245306e-01 2.42837533e-01 2.58664399e-01 3.93044561e-01 3.50587636e-01 -7.26366222e-01 -1.22511134e-01 -6.32610857e-01 -5.61177611e-01 -1.49306107e+00 3.37887913e-01 -6.01761997e-01 2.57902425e-02 -1.53674960e+00 2.30653614e-01 -1.02962665e-01 -5.23732722e-01 4.40231025e-01 -3.34176153e-01 6.58779545e-03 2.76469290e-01 1.81717947e-01 -1.02791286e+00 7.63637662e-01 1.66609907e+00 -1.42635494e-01 -1.59449250e-01 -3.84512961e-01 -5.32634974e-01 7.64089763e-01 2.55985498e-01 9.69177112e-02 -5.24918199e-01 -5.88364720e-01 -2.19463408e-01 4.48176920e-01 3.49528164e-01 -9.45713162e-01 5.44206083e-01 -5.12437344e-01 3.78469676e-01 -8.74378741e-01 3.12588125e-01 -8.84574711e-01 3.33416574e-02 3.40252876e-01 -3.52981120e-01 3.59075032e-02 2.99439933e-02 5.37078381e-01 -6.40953898e-01 2.44799092e-01 7.23213911e-01 -1.89521402e-01 -1.33121264e+00 6.07931852e-01 -3.83599252e-01 -2.57033974e-01 1.11682796e+00 -1.08935907e-01 -1.00109428e-01 -1.36080578e-01 -7.54541278e-01 6.67749286e-01 4.47430104e-01 6.01874411e-01 7.63185203e-01 -1.54651344e+00 -3.82985801e-01 2.79430687e-01 -5.16651943e-03 1.31495729e-01 7.09886074e-01 1.06439626e+00 -3.22298676e-01 1.56623349e-01 -3.33838344e-01 -7.55961299e-01 -1.12313139e+00 6.34611607e-01 4.97425258e-01 -9.74366441e-02 -5.69272161e-01 7.67473817e-01 4.01476383e-01 1.05318166e-01 -4.73930240e-02 -3.75205606e-01 -5.78524590e-01 2.26392955e-01 6.53670371e-01 6.80857748e-02 -4.01308447e-01 -1.40473211e+00 -3.37335318e-01 7.18258858e-01 5.39568849e-02 -2.24549621e-01 9.95095730e-01 -7.88535714e-01 -5.71684837e-01 4.49006647e-01 1.26390326e+00 -1.97601363e-01 -1.50058913e+00 -5.75021505e-01 -2.30539471e-01 -3.26899707e-01 2.29602501e-01 -2.54079193e-01 -1.43819606e+00 8.14796567e-01 3.56686085e-01 -7.75367245e-02 1.65837824e+00 -1.32766977e-01 1.06172037e+00 -1.88351005e-01 1.97156817e-01 -9.96012628e-01 1.61639690e-01 5.51779807e-01 5.09033501e-01 -9.25298572e-01 -8.41973051e-02 -5.01015961e-01 -7.22831428e-01 1.28133929e+00 8.52851748e-01 -2.85343975e-02 6.36092067e-01 1.78124234e-01 -9.81516019e-03 4.57987562e-02 -6.52529776e-01 -6.08843267e-01 4.32199538e-01 3.41138363e-01 2.32130423e-01 -1.88120812e-01 -5.02398014e-01 6.89961791e-01 2.85634369e-01 2.51866400e-01 5.57983257e-02 8.30639005e-01 -7.07303464e-01 -1.06313944e+00 -1.74631298e-01 1.68037876e-01 -4.47597988e-02 -5.27962018e-03 3.20729166e-01 3.27236325e-01 8.10665905e-01 9.56502318e-01 2.13341102e-01 -7.15298772e-01 2.62076296e-02 -1.10794410e-01 3.01490545e-01 -3.63343596e-01 -4.50121462e-01 7.60639429e-01 -3.07362169e-01 -8.16851914e-01 -1.12173426e+00 -7.05564618e-01 -1.76130450e+00 2.16973783e-03 -2.56353021e-01 -1.27431415e-02 8.44687968e-02 1.43082905e+00 3.09451908e-01 9.32496011e-01 5.25111318e-01 -9.54735696e-01 1.62633121e-01 -7.28397250e-01 -3.47865105e-01 5.21253169e-01 3.64614516e-01 -7.06836045e-01 -3.24782133e-02 4.15428847e-01]
[9.524349212646484, 0.06990666687488556]
9d27506c-26f5-42ca-9364-34c27112ed7f
text-to-sql-error-correction-with-language
2305.13073
null
https://arxiv.org/abs/2305.13073v2
https://arxiv.org/pdf/2305.13073v2.pdf
Text-to-SQL Error Correction with Language Models of Code
Despite recent progress in text-to-SQL parsing, current semantic parsers are still not accurate enough for practical use. In this paper, we investigate how to build automatic text-to-SQL error correction models. Noticing that token-level edits are out of context and sometimes ambiguous, we propose building clause-level edit models instead. Besides, while most language models of code are not specifically pre-trained for SQL, they know common data structures and their operations in programming languages such as Python. Thus, we propose a novel representation for SQL queries and their edits that adheres more closely to the pre-training corpora of language models of code. Our error correction model improves the exact set match accuracy of different parsers by 2.4-6.5 and obtains up to 4.3 point absolute improvement over two strong baselines. Our code and data are available at https://github.com/OSU-NLP-Group/Auto-SQL-Correction.
['Huan Sun', 'Yu Su', 'Jayanth Srinivasa', 'Ali Payani', 'Raymond Mooney', 'Michael White', 'Shijie Chen', 'Ziru Chen']
2023-05-22
null
null
null
null
['text-to-sql']
['computer-code']
[ 2.98994817e-02 1.23373955e-01 -1.63285583e-01 -9.47629154e-01 -1.23455119e+00 -7.56397188e-01 1.93855554e-01 8.53279471e-01 -2.66574353e-01 3.03497314e-01 2.71556079e-01 -6.51635826e-01 4.44789261e-01 -9.32921767e-01 -1.19573820e+00 4.60185498e-01 3.36912751e-01 3.82020354e-01 4.76788521e-01 -2.30950847e-01 5.18924117e-01 -6.14210889e-02 -1.06129885e+00 6.72824562e-01 1.16424549e+00 4.38234001e-01 1.20478764e-01 7.74201334e-01 -8.04018676e-01 9.25271630e-01 -4.26194727e-01 -1.03082502e+00 9.56652127e-03 -2.76908398e-01 -1.13564193e+00 -5.95257938e-01 5.82540870e-01 -1.58097699e-01 -1.07892916e-01 1.19366980e+00 1.93837270e-01 -2.75348932e-01 9.61650759e-02 -9.25775409e-01 -6.97582424e-01 9.64406729e-01 -3.70891362e-01 6.24951012e-02 4.54816520e-01 -2.56316692e-01 1.27845109e+00 -8.96603525e-01 8.47702503e-01 9.88441586e-01 8.74681473e-01 6.18004322e-01 -1.22541928e+00 -3.37234259e-01 -1.83005229e-01 -5.60102575e-02 -1.05507863e+00 -5.37014306e-01 3.44713032e-01 -4.38669056e-01 1.58488572e+00 3.89957905e-01 6.64437190e-03 7.98717618e-01 2.03053296e-01 6.42687917e-01 6.26967788e-01 -6.17832363e-01 -9.29457918e-02 2.74188280e-01 3.77076238e-01 9.87121403e-01 4.51258838e-01 -3.53795052e-01 -4.42430913e-01 -3.43406528e-01 2.43851691e-01 2.30341889e-02 1.45318463e-01 -3.66465211e-01 -8.84822428e-01 8.02612126e-01 2.31447607e-01 3.10443908e-01 1.04031339e-01 5.62770009e-01 8.65996480e-01 4.16180044e-01 2.88889855e-01 7.15611219e-01 -8.80361915e-01 -5.77708781e-01 -9.74412560e-01 3.55264157e-01 9.20814335e-01 1.46877885e+00 8.79800498e-01 -3.11200678e-01 1.73335239e-01 8.21636558e-01 1.76284939e-01 2.14077920e-01 3.01290423e-01 -1.02659297e+00 9.49010491e-01 8.57452571e-01 -3.66070792e-02 -8.52468729e-01 -1.28561497e-01 -1.97284564e-01 -1.47628202e-03 -1.01827160e-01 4.50407505e-01 7.59382397e-02 -5.03631890e-01 1.48489249e+00 6.45563900e-02 -3.54995638e-01 -1.13865718e-01 2.92687744e-01 5.29547989e-01 3.10562670e-01 2.33541816e-01 2.56364256e-01 1.34683454e+00 -9.77304697e-01 -4.32805657e-01 -6.52589977e-01 1.30630243e+00 -1.09645343e+00 1.37859237e+00 1.83755830e-02 -1.25649858e+00 -4.04886484e-01 -7.42276847e-01 -4.93373841e-01 -3.17933381e-01 4.81552407e-02 6.77084684e-01 5.53911269e-01 -8.72337401e-01 7.34366000e-01 -1.30869901e+00 -6.34243965e-01 2.83346832e-01 -6.20902479e-02 -3.41094375e-01 -1.29986167e-01 -6.36856019e-01 8.54601026e-01 4.38217640e-01 -4.87697423e-01 -4.89975691e-01 -1.17504299e+00 -9.69368219e-01 -3.52275297e-02 5.29366314e-01 -5.97473502e-01 1.80315745e+00 -7.76891410e-01 -8.88411283e-01 1.10284877e+00 -6.96319401e-01 -4.63730842e-01 4.51885670e-01 -6.41453207e-01 -3.49087000e-01 -2.64954448e-01 3.09633523e-01 2.74320245e-01 1.97404802e-01 -1.10667026e+00 -5.18059194e-01 -3.56691360e-01 3.29287380e-01 -2.54874170e-01 -1.25454590e-01 5.93857765e-01 -7.18088090e-01 -7.68783808e-01 1.37790851e-02 -6.46494627e-01 -1.96391359e-01 8.17397013e-02 -4.25025165e-01 3.70404013e-02 3.02845329e-01 -1.04675698e+00 1.58256710e+00 -2.10162735e+00 -1.92300886e-01 -1.91711783e-02 1.81246266e-01 1.84092134e-01 -1.45007610e-01 8.18483829e-01 -2.19431832e-01 6.62264228e-01 -6.43052280e-01 -4.16682273e-01 1.72223449e-01 3.06605339e-01 -4.14612651e-01 -6.38646958e-03 3.99672478e-01 1.03357923e+00 -8.55941713e-01 -6.21957362e-01 -1.18878648e-01 5.82288019e-02 -1.16169298e+00 3.54972929e-01 -6.21116102e-01 -4.43754531e-02 -4.68424082e-01 7.15590477e-01 5.70014834e-01 -1.26039207e-01 3.33494037e-01 -2.18553152e-02 -1.22527130e-01 1.06069970e+00 -8.46317232e-01 2.10557270e+00 -7.46422708e-01 4.81659770e-01 -2.34972805e-01 -8.84495020e-01 8.98254633e-01 -9.64226723e-02 2.07292065e-01 -7.00133681e-01 -3.19080979e-01 5.42380810e-01 -2.16814607e-01 -7.04939961e-01 7.63704062e-01 2.10001394e-01 -3.91827822e-01 3.87133956e-01 6.66349530e-02 -3.65694880e-01 2.94808179e-01 4.91123080e-01 1.47854042e+00 4.60770547e-01 3.09123307e-01 -1.33698106e-01 3.42507303e-01 3.71001214e-01 6.94952309e-01 7.44770288e-01 1.21698603e-01 6.80794537e-01 7.08632290e-01 -1.85522899e-01 -1.15298271e+00 -9.30266440e-01 -1.85549915e-01 1.28149664e+00 -8.06022063e-02 -1.18936396e+00 -9.13889468e-01 -9.44191217e-01 8.14645737e-02 1.21876550e+00 -3.29332709e-01 -8.60531405e-02 -9.24943507e-01 -3.73724580e-01 7.88675487e-01 8.47937346e-01 1.84270814e-01 -6.75195992e-01 -4.54761118e-01 3.52883637e-01 -1.94518045e-01 -1.11396742e+00 -6.47404850e-01 2.63319552e-01 -9.92817461e-01 -1.28403747e+00 2.04387173e-01 -7.12594330e-01 6.08572304e-01 -1.23981610e-01 1.71785998e+00 5.79911470e-01 -3.02264392e-01 2.22023278e-01 -5.05240440e-01 -5.39409757e-01 -9.87364292e-01 1.82780787e-01 -6.78057909e-01 -8.52787852e-01 8.27090800e-01 -4.13356096e-01 -9.22572687e-02 -1.00591080e-02 -8.43313158e-01 -1.32858217e-01 3.69318992e-01 6.44663572e-01 4.63099062e-01 -4.01356667e-01 3.13170031e-02 -1.47185600e+00 3.92388374e-01 -3.95960242e-01 -7.49297857e-01 4.91409540e-01 -7.26189077e-01 4.74431992e-01 6.26440048e-01 3.19840312e-01 -1.16243458e+00 1.24262869e-02 -4.60961401e-01 8.07419941e-02 -1.75996944e-01 7.53754914e-01 -3.75744165e-03 9.28588659e-02 7.76130259e-01 -1.06084704e-01 -3.10762912e-01 -7.00176954e-01 2.75211066e-01 5.03504515e-01 5.73410928e-01 -9.07286525e-01 8.65173340e-01 1.19553007e-01 -5.85698843e-01 -2.64125854e-01 -6.38498187e-01 -3.69740158e-01 -5.17300606e-01 5.01933634e-01 8.19063962e-01 -8.53130758e-01 -2.47474000e-01 2.73567468e-01 -1.49172235e+00 -3.19583982e-01 -2.40689844e-01 9.09639671e-02 -4.10720795e-01 7.13804722e-01 -9.82253969e-01 -4.15701330e-01 -3.00343663e-01 -1.12742913e+00 1.00399482e+00 -9.73725542e-02 -4.74151164e-01 -9.15473402e-01 1.50362730e-01 4.83121812e-01 5.47137916e-01 3.05199320e-03 1.31927705e+00 -8.56050968e-01 -6.16489589e-01 -1.72598556e-01 -2.03937247e-01 3.58640015e-01 -6.03354611e-02 4.22608107e-01 -6.09859705e-01 1.31940648e-01 -1.73930481e-01 -2.85135657e-01 5.74515820e-01 -2.08906949e-01 1.44751012e+00 -2.82085449e-01 -3.56064677e-01 6.26612186e-01 1.72311413e+00 -6.82957619e-02 7.60758519e-01 3.73223603e-01 5.87889671e-01 5.29645741e-01 6.49137497e-01 4.54855919e-01 6.88773334e-01 6.44309878e-01 5.97864866e-01 3.57814997e-01 -2.67576743e-02 -6.69164896e-01 3.45475376e-01 1.25119162e+00 3.44659358e-01 -9.41197425e-02 -1.20629120e+00 8.15795302e-01 -1.59647584e+00 -6.72964215e-01 -6.31980002e-01 2.22933173e+00 1.28035676e+00 2.46052712e-01 -2.51398236e-01 -3.81442487e-01 4.78879273e-01 -4.24510241e-02 -2.49455944e-01 -8.75456870e-01 2.04083711e-01 5.95023990e-01 7.50615478e-01 6.94455981e-01 -9.34725702e-01 1.15838230e+00 5.61907673e+00 4.53756541e-01 -7.75545895e-01 3.31433833e-01 1.16117977e-01 1.51009575e-01 -8.09939206e-01 5.71945548e-01 -9.31595087e-01 4.00786370e-01 1.15667689e+00 -3.28537077e-01 4.59616631e-01 1.09984970e+00 -9.44660902e-02 -6.52761683e-02 -1.36305535e+00 6.09034896e-01 -1.73240319e-01 -1.35803521e+00 -1.56383663e-01 -4.09282953e-01 6.27608657e-01 4.18812394e-01 -5.35531998e-01 5.13597250e-01 5.92313290e-01 -8.99996221e-01 8.95668089e-01 4.67153817e-01 6.04969501e-01 -4.00100797e-01 6.91626728e-01 1.11039318e-01 -1.01144218e+00 2.98991472e-01 -4.55346555e-01 2.58177638e-01 -7.25135580e-02 6.76679611e-01 -6.73800230e-01 6.41102552e-01 8.92838776e-01 8.22682321e-01 -9.87920582e-01 8.21865261e-01 -2.52549976e-01 6.83736265e-01 -2.47815862e-01 -2.65773982e-02 -1.05355807e-01 1.87190138e-02 2.60921210e-01 1.63954151e+00 4.31700945e-01 -1.79071650e-01 -4.42592762e-02 1.04463303e+00 -4.51698005e-01 1.85125947e-01 -5.29962420e-01 -4.47581708e-01 7.47805893e-01 9.17364240e-01 -2.16059223e-01 -3.03756505e-01 -7.75052547e-01 1.05663776e+00 5.91230512e-01 7.13869333e-02 -1.07017720e+00 -8.77117872e-01 8.38709056e-01 2.14389861e-01 3.95327717e-01 -3.32634836e-01 -5.48218191e-01 -1.37212348e+00 7.65160143e-01 -1.07791090e+00 3.46723735e-01 -6.81521475e-01 -1.12179434e+00 2.65164226e-01 -2.37192482e-01 -6.59191847e-01 -2.43565813e-01 -6.20103717e-01 -6.83910191e-01 8.22185695e-01 -1.40001392e+00 -8.84382427e-01 -2.34988660e-01 1.62356168e-01 5.29171824e-01 1.47220880e-01 1.00615585e+00 6.14537656e-01 -3.23546827e-01 9.15593863e-01 2.10426986e-01 5.37584782e-01 8.97275567e-01 -1.42322433e+00 1.04998159e+00 1.24369681e+00 1.52095288e-01 1.20145130e+00 7.61970162e-01 -6.82221115e-01 -1.58183968e+00 -1.27833426e+00 1.56354630e+00 -9.38878119e-01 9.04320657e-01 -5.63621700e-01 -1.29122078e+00 1.12645245e+00 2.92538941e-01 1.00708798e-01 4.77276772e-01 1.62423000e-01 -9.16916072e-01 -2.51730718e-02 -9.55674052e-01 3.96914840e-01 1.11883140e+00 -1.00328946e+00 -7.19772816e-01 5.03891170e-01 9.36498165e-01 -7.11858571e-01 -1.13109791e+00 1.89363465e-01 2.63995349e-01 -9.86896515e-01 7.66731799e-01 -9.83008921e-01 8.48210275e-01 -2.15679318e-01 -5.83962202e-01 -9.54656422e-01 1.98720708e-01 -4.72980767e-01 9.29541737e-02 1.53316605e+00 6.79648995e-01 -5.38749039e-01 7.32420504e-01 7.58478045e-01 -6.38134539e-01 -5.11995614e-01 -4.76764560e-01 -7.50790596e-01 3.52469057e-01 -8.42166603e-01 4.96291935e-01 1.16529381e+00 2.24182919e-01 2.22610701e-02 1.72215432e-01 1.76108375e-01 3.65590930e-01 1.45497352e-01 8.12795103e-01 -8.55720699e-01 -6.53428257e-01 -3.56168658e-01 -1.80575311e-01 -9.72817242e-01 1.66942656e-01 -1.28267276e+00 9.32488441e-02 -1.57467198e+00 2.59602576e-01 -6.23611271e-01 1.31631941e-01 7.09776163e-01 -2.02077746e-01 -1.89100564e-01 7.26051778e-02 -4.79429550e-02 -5.22112250e-01 2.11942405e-01 3.15202355e-01 -1.95243955e-02 1.36775345e-01 -2.68146634e-01 -8.44359398e-01 6.48273766e-01 1.01436353e+00 -9.96238351e-01 -1.08910903e-01 -1.14944446e+00 5.11606753e-01 2.86077578e-02 2.85458088e-01 -8.75203311e-01 1.76135436e-01 -1.54835671e-01 -3.01337689e-01 -2.60666758e-01 -2.22305745e-01 -6.49561644e-01 2.41031736e-01 4.04025555e-01 -6.34999871e-01 4.87734944e-01 3.55183452e-01 3.12746257e-01 -3.28077644e-01 -9.49510574e-01 6.19978249e-01 -4.16668147e-01 -6.79932714e-01 -1.46628758e-02 -1.48278967e-01 7.83055365e-01 6.50848866e-01 1.32966623e-01 -6.69198751e-01 -1.76599640e-02 -3.17065001e-01 7.19180107e-02 9.45197940e-01 5.33582270e-01 2.05645308e-01 -8.53927851e-01 -5.58309913e-01 -2.40839981e-02 5.00080764e-01 -8.21811259e-02 -1.80163682e-01 7.28758395e-01 -1.07378936e+00 5.00845551e-01 1.43883198e-01 -2.34990016e-01 -1.32146370e+00 4.47224021e-01 3.18888515e-01 -2.62835503e-01 -2.63471037e-01 8.46431255e-01 -2.04184547e-01 -1.04581785e+00 2.63201492e-03 -7.27080524e-01 6.04620874e-01 -5.92823625e-01 2.11778447e-01 1.55150488e-01 5.34201145e-01 -1.52613744e-01 -6.74588978e-01 5.11984229e-01 -2.19107836e-01 2.95794457e-01 1.42044628e+00 1.51279390e-01 -5.34302354e-01 3.19651902e-01 1.43089449e+00 6.16457939e-01 -6.28092647e-01 -2.81300336e-01 6.92564428e-01 -5.51189542e-01 -4.41298008e-01 -8.93495440e-01 -9.16544974e-01 1.10485899e+00 3.74960937e-02 -5.90028428e-02 8.24669003e-01 1.36950493e-01 8.21474791e-01 6.45466626e-01 4.73938763e-01 -1.09206474e+00 -1.31828666e-01 6.59697711e-01 6.07978702e-01 -1.26856863e+00 -1.78934112e-01 -6.58911288e-01 -3.62287998e-01 1.19649398e+00 7.37757027e-01 7.71269901e-03 3.14686596e-01 5.67806840e-01 -2.97597069e-02 -1.63070798e-01 -6.94423795e-01 6.32797778e-02 -8.05917755e-02 4.68645930e-01 1.06885207e+00 -1.09814063e-01 -4.17852461e-01 6.24310434e-01 -1.67268530e-01 1.58348437e-02 7.45888829e-01 1.30267453e+00 -3.84152800e-01 -1.70051670e+00 3.90500389e-02 5.07898331e-01 -9.05710757e-01 -6.36509061e-01 -2.40379408e-01 7.60484338e-01 -2.00202137e-01 7.99407959e-01 3.45290415e-02 -3.04232717e-01 6.61579728e-01 3.18709344e-01 3.90182823e-01 -1.04826343e+00 -9.17591214e-01 -4.89190280e-01 4.60824758e-01 -9.55327988e-01 -1.09365679e-01 -7.74304152e-01 -1.64787149e+00 -5.50352156e-01 -1.65723324e-01 2.39188433e-01 8.13981891e-01 5.68812788e-01 7.14777291e-01 3.53936136e-01 8.28694850e-02 9.23106372e-02 -6.91008925e-01 -6.45807862e-01 -5.88365570e-02 5.32482862e-01 -1.09460592e-01 -3.37013346e-03 -8.98121521e-02 3.43494654e-01]
[7.947274684906006, 7.867146015167236]
e446c51a-7df3-4ab0-9f68-83ec0c36e515
from-private-to-public-benchmarking-gans-in
2303.15916
null
https://arxiv.org/abs/2303.15916v2
https://arxiv.org/pdf/2303.15916v2.pdf
From Private to Public: Benchmarking GANs in the Context of Private Time Series Classification
Deep learning has proven to be successful in various domains and for different tasks. However, when it comes to private data several restrictions are making it difficult to use deep learning approaches in these application fields. Recent approaches try to generate data privately instead of applying a privacy-preserving mechanism directly, on top of the classifier. The solution is to create public data from private data in a manner that preserves the privacy of the data. In this work, two very prominent GAN-based architectures were evaluated in the context of private time series classification. In contrast to previous work, mostly limited to the image domain, the scope of this benchmark was the time series domain. The experiments show that especially GSWGAN performs well across a variety of public datasets outperforming the competitor DPWGAN. An analysis of the generated datasets further validates the superiority of GSWGAN in the context of time series generation.
['Sheraz Ahmed', 'Andreas Dengel', 'Dominique Mercier']
2023-03-28
null
null
null
null
['time-series-classification']
['time-series']
[ 2.84239948e-01 1.88645557e-01 1.48685396e-01 -2.69914210e-01 -8.33257258e-01 -6.98505580e-01 9.64637816e-01 -1.84491612e-02 -3.33516210e-01 1.00096655e+00 1.32328764e-01 -1.82563543e-01 3.15314010e-02 -1.14724839e+00 -6.09403551e-01 -1.00915611e+00 7.49391541e-02 1.28271833e-01 -3.81612688e-01 -2.47434899e-01 2.51740916e-03 4.10712808e-01 -1.53516340e+00 2.41702631e-01 3.35717857e-01 1.11958778e+00 -5.42645276e-01 1.44761011e-01 1.37883216e-01 4.71068233e-01 -9.64456677e-01 -7.39954770e-01 9.20858860e-01 -6.19629025e-01 -6.01498723e-01 -2.09083542e-01 6.76214322e-02 -2.36730397e-01 -2.58020937e-01 8.54492009e-01 6.31227434e-01 -4.35827747e-02 1.83919579e-01 -1.51382720e+00 -5.15671551e-01 4.28449661e-01 -2.69949883e-01 -1.63343832e-01 1.29765779e-01 -4.26146239e-02 9.53438759e-01 1.23272398e-02 7.48165727e-01 8.16537499e-01 6.28542423e-01 6.57020092e-01 -1.49817383e+00 -8.55317295e-01 -2.24710926e-01 -1.71789378e-02 -1.13244307e+00 -1.43702522e-01 9.01415706e-01 -2.74306208e-01 7.54081845e-01 4.19071227e-01 4.82796937e-01 1.84905541e+00 1.62074104e-01 6.90930068e-01 1.66183114e+00 -2.83795029e-01 6.21963859e-01 1.37324989e-01 -2.74992079e-01 -1.57915503e-01 2.42764264e-01 2.52519399e-01 -5.04842937e-01 -4.77368742e-01 2.45894194e-01 8.00986663e-02 -2.04566360e-01 -5.52722335e-01 -1.14338720e+00 1.04769576e+00 2.70595491e-01 5.46116054e-01 -3.88205111e-01 2.71450371e-01 7.88361728e-01 6.65672481e-01 8.09835255e-01 4.73898530e-01 -3.45290512e-01 -1.72185987e-01 -7.39388049e-01 4.15254503e-01 8.42831910e-01 6.26334786e-01 4.20093119e-01 6.22471757e-02 -4.42185938e-01 2.46025488e-01 -2.96751112e-01 1.80403173e-01 8.43965173e-01 -5.63829482e-01 6.24704301e-01 3.22154850e-01 -1.21106654e-01 -9.67713058e-01 1.50045455e-01 -4.59929138e-01 -1.08790660e+00 3.93689126e-01 5.37553787e-01 -3.03421497e-01 -7.29963243e-01 2.02356315e+00 2.20280185e-01 7.38073587e-02 3.71390074e-01 5.77790380e-01 4.39194322e-01 6.16542816e-01 -7.39035234e-02 6.02299981e-02 1.02094436e+00 -5.20114005e-01 -5.59848189e-01 1.14218794e-01 5.31070590e-01 -4.49443340e-01 8.64627004e-01 5.06351471e-01 -4.73455578e-01 -2.43162096e-01 -9.09783721e-01 3.01829055e-02 -7.19651043e-01 -7.37407655e-02 6.38073146e-01 8.92953694e-01 -1.03564095e+00 6.18925214e-01 -6.57151639e-01 -3.33113551e-01 6.73771918e-01 2.87218541e-01 -8.24629307e-01 1.14811569e-01 -1.35380006e+00 5.02597690e-01 4.95530635e-01 -2.19951347e-01 -8.14580321e-01 -5.93378484e-01 -6.10780835e-01 1.79245085e-01 8.10586661e-03 -6.52279735e-01 1.06149542e+00 -1.05134356e+00 -1.53446472e+00 1.00393856e+00 3.82105142e-01 -1.30507708e+00 1.25185800e+00 -4.49261256e-02 -3.36060166e-01 -1.70903787e-01 -3.65785845e-02 5.04057586e-01 1.02124107e+00 -8.48757088e-01 -5.89128971e-01 -4.32110488e-01 5.60812186e-03 -2.69276530e-01 -4.68308419e-01 -3.10785830e-01 7.53708631e-02 -1.03567696e+00 -3.67413014e-01 -9.69868422e-01 -2.06939012e-01 -3.31923366e-01 -4.20925617e-01 -4.26920690e-02 1.41340852e+00 -5.94046235e-01 7.48451173e-01 -2.26309395e+00 -2.06598081e-03 3.17325890e-01 4.11693528e-02 1.65975809e-01 4.76476140e-02 7.52326012e-01 -3.85561377e-01 1.27029374e-01 -3.92172098e-01 -6.53069794e-01 3.48032206e-01 2.80001283e-01 -7.63738692e-01 7.05430746e-01 -4.94668186e-02 8.85529101e-01 -5.39524317e-01 7.91792497e-02 1.58157349e-01 5.08988798e-01 -2.84935296e-01 1.45902172e-01 -2.20815539e-01 7.09662914e-01 -3.49300712e-01 -2.24436596e-02 5.81260443e-01 1.32773006e-02 1.19878650e-01 2.91280687e-01 2.22335905e-01 8.38873908e-02 -7.28269517e-01 1.69405484e+00 -4.71118361e-01 6.63752794e-01 -1.60204500e-01 -1.08191431e+00 1.00269377e+00 7.01344132e-01 6.54630423e-01 -7.61763513e-01 1.70205027e-01 1.29011810e-01 -2.80720830e-01 -1.35944292e-01 2.03256071e-01 -2.18259141e-01 -3.21880758e-01 6.91060305e-01 -1.54991090e-01 -4.14861962e-02 -2.14647472e-01 -1.38481289e-01 1.19482100e+00 1.43085435e-01 2.08677232e-01 6.59454316e-02 4.31075275e-01 -2.38441721e-01 4.23174292e-01 4.92336333e-01 8.65870435e-03 7.66596437e-01 6.33412480e-01 -6.21855319e-01 -1.08960140e+00 -7.34732330e-01 1.18901752e-01 5.66230536e-01 -4.15729195e-01 -4.12824929e-01 -8.82766068e-01 -9.66845453e-01 -1.43377976e-02 7.80546904e-01 -8.59363854e-01 -2.18121797e-01 -4.31446642e-01 -7.62208879e-01 6.55560136e-01 2.20143959e-01 8.84736061e-01 -1.13273561e+00 -1.02085853e+00 1.54549941e-01 -1.13161355e-01 -1.17348766e+00 -1.39020626e-02 2.33086776e-02 -8.66329789e-01 -7.28500903e-01 -8.14160049e-01 -2.47425869e-01 3.24012816e-01 -1.85594276e-01 8.28817546e-01 -3.89270633e-01 -3.09034377e-01 4.39964920e-01 -4.85367835e-01 -6.90614641e-01 -5.43762565e-01 4.43226308e-01 -5.61973989e-01 4.74172354e-01 3.95909011e-01 -7.97679126e-01 -6.14565909e-01 1.69191684e-03 -1.16543794e+00 -1.05381109e-01 3.28835994e-01 8.80804658e-01 4.28107053e-01 2.18485534e-01 6.74834013e-01 -1.00613177e+00 7.29167640e-01 -6.50215387e-01 -6.76100731e-01 -5.14373071e-02 -7.64037251e-01 -1.30387932e-01 8.50207984e-01 -1.11676969e-01 -8.28943849e-01 9.44029316e-02 -1.44821674e-01 -3.50893587e-01 -4.65314120e-01 1.35291278e-01 -2.18940571e-01 1.02571346e-01 4.97912914e-01 3.52576286e-01 3.37276608e-01 -5.24746776e-01 5.68938330e-02 5.90054750e-01 2.48906314e-01 -3.69513243e-01 7.89303124e-01 7.88101137e-01 2.32770637e-01 -5.23384392e-01 -3.84683102e-01 6.71258345e-02 -1.33988187e-01 3.33073974e-01 8.61642182e-01 -8.00823271e-01 -4.11651522e-01 6.68437362e-01 -7.24930763e-01 -2.57075012e-01 -8.69408131e-01 3.56437355e-01 -8.66275609e-01 2.09978461e-01 -3.00943613e-01 -6.57775104e-01 -5.65340281e-01 -9.32587504e-01 1.03997242e+00 -1.79559723e-01 -2.76180357e-01 -9.46625113e-01 7.58064985e-02 2.37058759e-01 6.50783658e-01 1.08193862e+00 8.12138557e-01 -8.45651627e-01 -5.47810793e-01 -7.56323218e-01 1.97791621e-01 6.22577012e-01 1.76012769e-01 -4.46182758e-01 -1.48815036e+00 -4.52396452e-01 5.39286733e-01 -2.17634439e-01 5.88836312e-01 4.79491465e-02 1.49336529e+00 -5.46537936e-01 -2.43939925e-02 7.45738626e-01 1.50762331e+00 1.66663378e-01 1.00830030e+00 5.82413375e-01 3.27464521e-01 7.95427024e-01 2.82061696e-01 5.84095776e-01 1.82588175e-02 7.47672081e-01 6.72024727e-01 -1.73692167e-01 2.78883070e-01 -3.37846756e-01 3.14665675e-01 7.38944784e-02 2.45610029e-01 -5.02666771e-01 -6.11501694e-01 6.36172771e-01 -1.86289835e+00 -1.17321277e+00 1.23212323e-01 2.27313304e+00 5.42114675e-01 -1.76253200e-01 1.15161814e-01 4.07119930e-01 4.24558938e-01 3.82398516e-01 -3.08837175e-01 -5.08740366e-01 -3.50371689e-01 4.83816296e-01 6.62291825e-01 -2.61428952e-01 -1.20303893e+00 5.44419169e-01 5.82127285e+00 7.71248817e-01 -1.30516636e+00 2.87879258e-01 8.90367150e-01 -2.00367182e-01 -3.79968047e-01 -3.42294611e-02 -2.95954108e-01 6.71076238e-01 9.16432500e-01 -3.05625349e-01 3.88563693e-01 8.12595725e-01 -4.50984053e-02 3.01757008e-01 -1.16317666e+00 1.10792446e+00 -2.64736056e-01 -1.26036119e+00 -1.30068779e-01 4.08516139e-01 7.06745923e-01 7.03709275e-02 3.75339746e-01 7.37000108e-02 2.60218799e-01 -1.28066850e+00 6.63893878e-01 2.80997127e-01 5.90968609e-01 -7.95907140e-01 8.13314617e-01 3.41731608e-01 -5.20005405e-01 -6.55204430e-02 4.52791229e-02 -8.25407654e-02 -1.64545000e-01 5.92512608e-01 -8.98591936e-01 7.62880564e-01 9.92969453e-01 5.80642939e-01 -4.20280576e-01 6.86616242e-01 -2.31871754e-01 7.26788938e-01 -3.58480692e-01 2.62507468e-01 3.63506496e-01 -1.60944343e-01 5.46540678e-01 8.84864688e-01 6.10491335e-01 -3.26837540e-01 -3.45044613e-01 9.54180658e-01 -1.25551954e-01 1.39166579e-01 -1.02427971e+00 -1.96730465e-01 6.31196573e-02 1.11766577e+00 -3.35795760e-01 3.52233909e-02 -2.63501883e-01 1.08530879e+00 -3.51946428e-02 3.03193957e-01 -7.21378505e-01 -3.19809854e-01 7.50628650e-01 1.23749375e-01 4.29616481e-01 -5.89658283e-02 -1.70112193e-01 -1.06784308e+00 3.86393636e-01 -1.05327511e+00 7.44340599e-01 -4.97317612e-01 -1.51769018e+00 6.68753743e-01 1.08195789e-01 -1.37121141e+00 -6.02902830e-01 -2.65343636e-01 -5.00812352e-01 9.90937114e-01 -1.24033701e+00 -1.26699960e+00 -2.13951543e-01 9.00949597e-01 7.51393847e-03 -4.37470794e-01 1.17166376e+00 1.36800632e-01 -2.68553764e-01 7.25849092e-01 4.81820643e-01 1.34768829e-01 5.36716580e-01 -1.01814473e+00 6.53556466e-01 8.75704288e-01 2.90998369e-01 3.04338604e-01 6.42892241e-01 -3.89449656e-01 -1.26319587e+00 -1.16080415e+00 7.77957678e-01 -1.62530869e-01 4.04001594e-01 -7.12442040e-01 -7.62197852e-01 7.08564997e-01 6.66922271e-01 1.30512029e-01 8.66277337e-01 -3.79182309e-01 -3.81004125e-01 -3.61476332e-01 -1.67448175e+00 3.89829665e-01 7.59552658e-01 -5.74610949e-01 -1.74636990e-01 2.61217415e-01 4.58984643e-01 -2.66925246e-01 -8.72928321e-01 1.96654677e-01 2.88284779e-01 -1.09899557e+00 6.84330940e-01 -5.97712576e-01 4.04558569e-01 -2.04965815e-01 -1.89378455e-01 -1.34297109e+00 1.30996734e-01 -9.94694710e-01 -1.26182869e-01 1.42468393e+00 -2.19154339e-02 -1.22105753e+00 1.01363266e+00 6.46252155e-01 4.36256081e-01 -3.27563226e-01 -1.31070483e+00 -8.75102639e-01 1.36503592e-01 -2.41758406e-01 1.28997684e+00 1.06843698e+00 -5.91226518e-01 -5.56736104e-02 -5.80248296e-01 -1.65785670e-01 5.48639238e-01 3.78862530e-01 1.10854650e+00 -1.06534851e+00 -4.11682129e-01 -2.57595986e-01 -8.83628786e-01 -3.86088379e-02 3.53102714e-01 -9.26844954e-01 -3.95816863e-01 -1.09854090e+00 -4.02280122e-01 -4.32502866e-01 -3.53930116e-01 6.35388374e-01 4.37585443e-01 2.88568199e-01 2.35216051e-01 5.33935167e-02 2.95514166e-01 5.99498451e-01 8.79626989e-01 -1.67328760e-01 1.08349189e-01 2.81211674e-01 -7.68687308e-01 1.78000152e-01 1.23430037e+00 -5.53521216e-01 -6.18670166e-01 -3.39281470e-01 -2.76164897e-02 -2.00070232e-01 7.66047716e-01 -1.16706622e+00 -1.49022654e-01 1.34385571e-01 2.08308384e-01 -3.16748112e-01 2.48386428e-01 -1.24698734e+00 8.37583244e-01 3.47662717e-01 -3.36954564e-01 1.04173288e-01 1.39862448e-01 6.61563873e-01 -4.61666584e-01 1.20010518e-01 5.99428296e-01 -1.50433518e-02 -6.10876203e-01 3.49069566e-01 -4.63986360e-02 2.19999924e-02 1.52654898e+00 -2.18045309e-01 -1.91968828e-01 -7.08761692e-01 -5.80256522e-01 -2.70847142e-01 6.26956463e-01 6.03815198e-01 1.72953427e-01 -1.26107192e+00 -6.32186472e-01 4.86743093e-01 2.45619833e-01 -2.00924620e-01 1.65890574e-01 4.94473338e-01 -3.05860132e-01 4.54866260e-01 -4.36967611e-01 -3.91547054e-01 -1.04291654e+00 7.62406170e-01 3.30889285e-01 -3.89795393e-01 -1.05890751e+00 3.49026084e-01 1.38986871e-01 -1.71196535e-01 2.83951700e-01 -1.38286576e-01 1.71110660e-01 3.20174843e-02 2.09988132e-01 2.25437269e-01 4.79166597e-01 -5.08415163e-01 3.25355977e-02 1.03963241e-01 2.02872038e-01 -3.81012082e-01 1.66819715e+00 1.35990933e-01 -1.61675736e-01 1.60471469e-01 1.55397058e+00 -8.47152844e-02 -1.20601046e+00 -2.51240376e-02 -1.22724995e-02 -6.36860073e-01 -3.04499865e-01 -6.67570829e-01 -1.36806321e+00 8.73669326e-01 7.78613865e-01 4.95254785e-01 1.48858297e+00 -3.73528421e-01 8.46745074e-01 -6.27379958e-03 8.09615135e-01 -9.18919921e-01 -4.60987031e-01 1.49721667e-01 8.27420712e-01 -9.85863864e-01 -2.00468391e-01 1.03169819e-02 -5.92788994e-01 8.46687794e-01 2.66341507e-01 -1.34046495e-01 4.71322626e-01 1.45452857e-01 4.59000580e-02 1.14473313e-01 -5.86940467e-01 2.88818568e-01 -1.74838547e-02 6.88348293e-01 1.71606153e-01 2.70653944e-02 -4.96543497e-01 3.43050689e-01 -4.90428060e-01 2.70467073e-01 3.55285585e-01 1.14834869e+00 5.88081956e-01 -1.69950330e+00 -2.89819539e-01 3.80439639e-01 -9.44166839e-01 3.89569938e-01 -5.47781408e-01 9.84061480e-01 1.79444805e-01 6.85498118e-01 -7.51657709e-02 -2.78098196e-01 1.20866597e-01 1.80108145e-01 4.97217700e-02 -3.41058910e-01 -1.17483866e+00 -3.03502053e-01 -1.26180972e-03 -6.41659558e-01 -4.50776458e-01 -7.46878266e-01 -6.78083956e-01 -3.70102704e-01 1.79414675e-01 2.28964180e-01 7.73751736e-01 4.95641828e-01 6.93756282e-01 1.12622648e-01 6.22256935e-01 -3.05497438e-01 -7.88292825e-01 -5.43837965e-01 -6.27885640e-01 9.04217005e-01 3.29403222e-01 -1.73571855e-01 -2.21896082e-01 -1.49470896e-01]
[6.156399250030518, 6.85683012008667]
d6d5f9fa-3325-49e8-8877-f1dde7cad282
keyphrase-generation-a-text-summarization
1904.00110
null
http://arxiv.org/abs/1904.00110v2
http://arxiv.org/pdf/1904.00110v2.pdf
Keyphrase Generation: A Text Summarization Struggle
Authors' keyphrases assigned to scientific articles are essential for recognizing content and topic aspects. Most of the proposed supervised and unsupervised methods for keyphrase generation are unable to produce terms that are valuable but do not appear in the text. In this paper, we explore the possibility of considering the keyphrase string as an abstractive summary of the title and the abstract. First, we collect, process and release a large dataset of scientific paper metadata that contains 2.2 million records. Then we experiment with popular text summarization neural architectures. Despite using advanced deep learning models, large quantities of data and many days of computation, our systematic evaluation on four test datasets reveals that the explored text summarization methods could not produce better keyphrases than the simpler unsupervised methods, or the existing supervised ones.
['Ondřej Bojar', 'Erion Çano']
2019-03-29
null
null
null
null
['keyphrase-generation']
['natural-language-processing']
[ 2.52821833e-01 3.72204036e-01 -2.43698359e-01 1.72535822e-01 -8.92603815e-01 -7.44069219e-01 1.06100500e+00 9.76241231e-01 -4.30160016e-01 1.18686390e+00 8.96992147e-01 -2.98821598e-01 -2.72700995e-01 -7.58429825e-01 -8.27791870e-01 -4.79728103e-01 -4.55349013e-02 3.50210279e-01 -1.54886425e-01 1.50828257e-01 1.07330835e+00 4.12623733e-01 -1.64063764e+00 3.60709280e-01 1.11660826e+00 7.45199740e-01 1.44342288e-01 8.67774725e-01 -8.53762448e-01 7.96094656e-01 -1.24599636e+00 -2.23801449e-01 -1.88532874e-01 -4.94418710e-01 -9.34125960e-01 -1.80246174e-01 1.03548825e+00 -3.56986225e-01 -3.96281481e-01 1.04163706e+00 5.46486020e-01 7.18740150e-02 8.74231219e-01 -9.29125488e-01 -9.25099015e-01 1.39770198e+00 -5.71269453e-01 4.37574714e-01 4.49715018e-01 -3.57198000e-01 1.22250283e+00 -5.68985164e-01 9.71521199e-01 9.30357695e-01 5.48171878e-01 2.45069951e-01 -8.53473485e-01 -3.42309266e-01 -1.26395836e-01 -8.00008476e-02 -8.70212078e-01 -4.68085915e-01 9.15519655e-01 -2.91280568e-01 1.00788689e+00 5.45569062e-01 5.62776864e-01 1.21384311e+00 5.55928826e-01 8.82011354e-01 7.72656620e-01 -3.79427612e-01 3.71929407e-01 -7.75076076e-02 7.04792619e-01 3.92435282e-01 1.01639366e+00 -6.05045319e-01 -7.16654897e-01 -3.92066032e-01 7.62232617e-02 -4.08957005e-02 -1.37753323e-01 3.57267290e-01 -1.59969699e+00 6.34641230e-01 -1.15329385e-01 3.56142849e-01 -7.97502339e-01 7.17737377e-02 6.33333266e-01 2.61014968e-01 5.52314341e-01 1.07298839e+00 -3.34288776e-01 -6.47015646e-02 -1.53453016e+00 6.85343623e-01 1.22306466e+00 1.08268309e+00 5.01165986e-01 9.32514071e-02 -7.07529128e-01 4.22909826e-01 -1.55148223e-01 3.28516990e-01 6.91461086e-01 -8.59684885e-01 5.47834218e-01 8.42374980e-01 1.48300186e-01 -1.16652501e+00 -3.94165099e-01 -5.49461722e-01 -1.05927682e+00 -4.47811782e-01 -2.08447967e-02 -3.85335207e-01 -9.73759890e-01 9.80720937e-01 -2.94676483e-01 -4.52826113e-01 3.26021820e-01 1.53561026e-01 1.71507943e+00 1.08851171e+00 -4.43677604e-02 -6.85058892e-01 1.34131336e+00 -8.20400357e-01 -1.15953946e+00 1.94409415e-01 2.99808800e-01 -8.63703847e-01 5.08991897e-01 6.92034066e-01 -1.33153188e+00 -4.94867623e-01 -1.07311714e+00 -3.33645940e-01 -7.64757931e-01 4.59741116e-01 6.56262279e-01 1.80931523e-01 -1.17328024e+00 8.77822995e-01 -3.10114384e-01 -5.54936469e-01 5.11630893e-01 -8.51275846e-02 -1.93229705e-01 4.79027212e-01 -9.58751500e-01 5.42089224e-01 8.42858672e-01 -2.22705662e-01 -5.04691482e-01 -8.60826731e-01 -4.66425568e-01 5.10731995e-01 2.91359514e-01 -7.41507471e-01 1.27827287e+00 -3.93952906e-01 -1.33215463e+00 7.53886044e-01 -1.44063979e-02 -7.25653291e-01 3.17007959e-01 -6.37796700e-01 -1.58672303e-01 5.03357410e-01 -2.02206988e-03 6.18057430e-01 6.71931624e-01 -1.06926620e+00 -8.42952847e-01 -1.42005295e-01 -2.97386110e-01 -8.20773914e-02 -6.82625771e-01 -2.29764879e-02 -1.84532493e-01 -1.07242358e+00 -1.90543607e-02 -3.58541429e-01 -4.24216464e-02 -7.11935639e-01 -9.34945941e-01 -6.78377211e-01 5.23959339e-01 -9.63147700e-01 1.58151340e+00 -1.47651041e+00 2.78772146e-01 -1.41947448e-01 4.06095833e-01 3.97756994e-02 1.28890246e-01 1.01794934e+00 1.81846797e-01 5.26645660e-01 -1.30026147e-01 -1.35855138e-01 6.35946319e-02 -1.56627342e-01 -1.00848627e+00 6.62864819e-02 -1.56607106e-01 9.05252278e-01 -8.22142720e-01 -7.04370499e-01 -4.18351322e-01 -1.35685459e-01 -2.63713956e-01 2.81829655e-01 -5.65459669e-01 -5.89369573e-02 -6.81960523e-01 6.69939518e-01 5.32096326e-01 -2.00982571e-01 -1.41174272e-01 -3.22115302e-01 -3.87083173e-01 6.69131517e-01 -8.40991378e-01 1.70741987e+00 1.19742759e-01 1.11614108e+00 -4.31108236e-01 -1.05390537e+00 8.95590425e-01 2.97369808e-01 7.31430709e-01 -3.97958100e-01 1.93050995e-01 2.53054976e-01 -3.97772640e-01 -5.23098528e-01 1.24789321e+00 5.37969351e-01 -2.42227226e-01 7.23656595e-01 2.74010003e-01 -1.89867586e-01 7.38264978e-01 7.25259602e-01 1.03967392e+00 -5.64251393e-02 3.59127969e-01 -5.11436760e-01 4.17134941e-01 1.67511538e-01 1.19940490e-01 1.35205400e+00 4.43122804e-01 7.12523103e-01 8.65072191e-01 -3.94388229e-01 -1.17777956e+00 -7.30720401e-01 7.21339881e-03 8.11566889e-01 -2.84233958e-01 -7.65881479e-01 -8.77727985e-01 -5.39316595e-01 7.97321647e-02 9.54553127e-01 -4.61946696e-01 -1.00664847e-01 -4.94492173e-01 -7.94671893e-01 8.88066530e-01 2.05474675e-01 3.14147532e-01 -1.23129249e+00 -7.44930625e-01 2.95673072e-01 -1.84843585e-01 -7.46604621e-01 -2.58138359e-01 2.19259113e-01 -9.64662135e-01 -6.33086562e-01 -1.11257005e+00 -6.75705492e-01 6.37986839e-01 5.72818369e-02 1.22719073e+00 -2.02205017e-01 -1.10668018e-01 3.71529281e-01 -5.49300492e-01 -7.25207686e-01 -6.32499576e-01 6.64951384e-01 5.75693464e-03 -4.18477565e-01 2.95539588e-01 -7.09077418e-01 -3.35483730e-01 -9.71021593e-01 -1.16079974e+00 1.01792835e-01 9.30271506e-01 6.72010422e-01 2.56645381e-01 -5.45453951e-02 8.67078960e-01 -1.01519620e+00 1.30388498e+00 -4.53605086e-01 -3.62652868e-01 3.07135373e-01 -9.16116238e-01 5.06167054e-01 6.97507739e-01 -3.99203718e-01 -9.81907487e-01 -5.61373711e-01 1.58720911e-01 7.76161626e-02 -2.36128047e-01 9.25088704e-01 1.89953998e-01 7.07728446e-01 6.13150120e-01 7.54719138e-01 -3.23693722e-01 -1.01011765e+00 5.21396041e-01 8.75024796e-01 6.45937741e-01 -6.24139905e-01 4.81106669e-01 1.82624385e-01 -1.80276185e-01 -1.01126373e+00 -8.86223078e-01 -4.20885116e-01 -5.65023720e-01 7.34519213e-02 6.07041359e-01 -7.12921441e-01 -5.38331389e-01 4.33737516e-01 -1.72887874e+00 5.62164664e-01 -4.60970610e-01 3.18135381e-01 -2.17366695e-01 7.27995694e-01 -5.78379035e-01 -3.58913034e-01 -1.21912265e+00 -5.91165066e-01 1.00773323e+00 6.25621140e-01 -6.10764444e-01 -4.67438102e-01 1.48704499e-01 1.30462527e-01 3.46074790e-01 2.27905825e-01 1.01397908e+00 -1.30089593e+00 -2.56549001e-01 -3.87025326e-01 -1.35512501e-01 1.39943019e-01 3.33293706e-01 5.28063297e-01 -7.49853075e-01 -3.17166112e-02 -1.75260231e-01 -2.28590295e-01 1.46963310e+00 4.82858747e-01 1.53961492e+00 -1.05907893e+00 -2.43283778e-01 4.06778872e-01 1.05490339e+00 2.28118718e-01 5.38346887e-01 4.96920675e-01 8.26385140e-01 7.63651967e-01 1.26158148e-01 5.03189206e-01 3.13645571e-01 -1.04521506e-01 -1.05759457e-01 2.26746947e-01 4.77278307e-02 -2.68256187e-01 2.43170232e-01 1.31997478e+00 -1.35193631e-01 -6.45923436e-01 -8.89824033e-01 7.49389768e-01 -1.72933221e+00 -1.28699863e+00 -2.79700130e-01 1.71565390e+00 1.32616508e+00 3.16273779e-01 -1.47397906e-01 -1.28860185e-02 3.46392900e-01 5.42466283e-01 -2.54653096e-01 -6.74182057e-01 -3.37088883e-01 1.65997684e-01 5.88840902e-01 1.50544390e-01 -1.11347485e+00 8.85390937e-01 6.74366140e+00 6.97285056e-01 -9.31315660e-01 -3.90344203e-01 5.49839616e-01 -3.13252024e-02 -5.61482310e-01 -9.92473289e-02 -9.42457795e-01 4.38602060e-01 1.07556999e+00 -7.58691192e-01 -1.17228746e-01 7.98690319e-01 1.66822612e-01 -3.44824493e-01 -1.16641426e+00 7.57449627e-01 3.62550050e-01 -1.99904490e+00 8.45707536e-01 -7.51481056e-02 1.09382296e+00 -2.47501805e-01 -2.56741673e-01 4.86883782e-02 2.96765894e-01 -8.99388194e-01 6.48192167e-01 1.05732465e+00 2.73824841e-01 -6.41295433e-01 8.21021438e-01 2.34074131e-01 -2.94410050e-01 1.70259655e-01 -5.71423173e-01 -7.87074491e-02 -3.37165833e-01 8.59476149e-01 -5.30005455e-01 7.63537347e-01 6.94748223e-01 8.84441972e-01 -8.53265405e-01 9.69233871e-01 -1.74632184e-02 7.95843720e-01 -5.44406362e-02 -5.82728028e-01 2.97593117e-01 -1.05424926e-01 7.51881778e-01 1.65365028e+00 3.59201729e-01 -2.73958724e-02 -7.95344040e-02 9.21830773e-01 -6.07729197e-01 2.77938783e-01 -6.55667841e-01 -8.22608173e-01 3.84478658e-01 1.33442402e+00 -1.03678513e+00 -9.81188893e-01 -3.93277556e-02 7.69531071e-01 -2.87708221e-03 4.81846750e-01 -2.76090711e-01 -8.75432909e-01 4.71501239e-02 -1.79595411e-01 1.71203151e-01 -2.00630277e-01 -6.14459574e-01 -1.23956943e+00 1.11174256e-01 -9.99711752e-01 3.20748538e-01 -8.61456096e-01 -1.12587643e+00 5.18970847e-01 1.81941360e-01 -9.76025105e-01 -3.58723491e-01 -2.70380795e-01 -8.73220921e-01 5.45985043e-01 -1.03149772e+00 -9.04659271e-01 -1.37270927e-01 -2.22174466e-01 8.16193938e-01 -6.02124393e-01 7.12737918e-01 -1.83845058e-01 -4.51187491e-01 2.33729303e-01 6.48642421e-01 7.66546726e-02 6.73568606e-01 -1.49126327e+00 4.53030288e-01 8.17987144e-01 1.57915130e-01 1.04362118e+00 1.18165064e+00 -9.60307181e-01 -1.72393942e+00 -6.62478447e-01 1.16992784e+00 -3.62965256e-01 7.27496982e-01 -1.73564136e-01 -1.01639974e+00 3.36860061e-01 1.22139513e+00 -8.37910771e-01 5.16835153e-01 -1.13621973e-01 -4.68103886e-02 -1.43987715e-01 -6.21607304e-01 7.31101573e-01 6.77302778e-01 -1.38254017e-01 -1.47246969e+00 3.85178417e-01 1.12559569e+00 -1.03705600e-01 -7.90641189e-01 1.68616831e-01 4.38627452e-01 -3.29437017e-01 7.98569739e-01 -8.63026202e-01 1.11887884e+00 9.67726260e-02 3.06544840e-01 -1.29585826e+00 -1.46708190e-01 -8.03230226e-01 -7.76801467e-01 1.69148195e+00 3.22245598e-01 -3.18241328e-01 5.15902936e-01 6.53033331e-02 -1.47372559e-01 -5.26196301e-01 -5.61525226e-01 -2.40183413e-01 4.05607551e-01 1.42747924e-01 4.40755934e-01 9.58777070e-01 1.75847873e-01 5.50164223e-01 -2.58302063e-01 -4.66071308e-01 7.39602089e-01 4.38559145e-01 7.31663287e-01 -1.58195782e+00 3.40780228e-01 -9.81112599e-01 6.10980541e-02 -6.76496267e-01 6.51497468e-02 -8.42694283e-01 -1.04653053e-01 -2.24541187e+00 4.27566290e-01 3.07094872e-01 -3.07723343e-01 3.35022211e-01 -3.64032954e-01 -5.06229699e-01 -3.45040619e-01 4.42733496e-01 -6.36206210e-01 5.77204764e-01 9.30935085e-01 -7.09602416e-01 -2.48676881e-01 -2.51332521e-01 -1.21294951e+00 6.43715501e-01 8.28558803e-01 -5.72115064e-01 -1.32087708e-01 -1.89013138e-01 5.04510522e-01 -1.93533584e-01 1.10252731e-01 -8.40105712e-01 6.59260213e-01 -3.76015872e-01 5.32621682e-01 -1.42896557e+00 -4.96579289e-01 -1.59273282e-01 -2.52583593e-01 2.89922506e-01 -1.00313294e+00 2.35336676e-01 3.19583625e-01 3.65397364e-01 -2.23907635e-01 -4.38293189e-01 1.21168286e-01 -3.76536161e-01 -3.81244510e-01 -1.24295913e-01 -8.65360856e-01 9.59219262e-02 2.88445979e-01 9.77047980e-02 -5.89700162e-01 -1.72464609e-01 -7.74779096e-02 1.33418635e-01 1.70187742e-01 5.70310354e-01 6.38606191e-01 -9.46387827e-01 -1.16984177e+00 -2.71605521e-01 -5.05599640e-02 1.33238941e-01 -1.08952321e-01 4.57563967e-01 -8.86211395e-01 9.08105850e-01 -3.13550472e-01 -4.68723988e-03 -8.73083889e-01 7.07044959e-01 -3.72405171e-01 -4.67391640e-01 -9.57085609e-01 3.30252290e-01 -2.10150838e-01 -9.93042886e-02 5.96171141e-01 -7.73176730e-01 -7.66778171e-01 8.24112117e-01 6.92950547e-01 5.19622624e-01 1.10566668e-01 -1.97955042e-01 3.87648903e-02 2.13376433e-01 -5.01746535e-01 -2.50291526e-02 1.66904414e+00 2.57840063e-02 -6.02217495e-01 6.39062583e-01 1.13222504e+00 3.57687563e-01 -3.95331234e-01 -1.74531445e-01 5.78272581e-01 2.99456984e-01 3.12701970e-01 -6.72120154e-01 -6.51712537e-01 7.05356896e-01 1.40944019e-01 6.40314579e-01 9.28148329e-01 -7.23864436e-02 8.39285016e-01 1.08665812e+00 -3.03077877e-01 -1.40728331e+00 -4.07397151e-02 5.80165744e-01 1.04874158e+00 -1.00509608e+00 6.78713918e-01 3.29760760e-01 -2.54889965e-01 1.41198969e+00 2.95262754e-01 -3.45769897e-02 1.57027259e-01 8.01507086e-02 -1.53794572e-01 -5.61747611e-01 -8.34699392e-01 1.85697347e-01 6.14042640e-01 2.77759749e-02 6.70365572e-01 -1.93285227e-01 -9.68720019e-01 8.26466143e-01 -5.62818527e-01 -3.08174305e-02 1.08478713e+00 9.39078033e-01 -7.15490282e-01 -4.51793969e-01 -4.38030601e-01 1.02008235e+00 -1.05155051e+00 -2.27484271e-01 -9.15372193e-01 3.52268517e-01 -3.84625345e-01 7.01198816e-01 1.16772741e-01 -5.54431267e-02 6.74990118e-02 1.55469850e-01 7.30673298e-02 -5.53531229e-01 -6.86909914e-01 8.92125890e-02 3.95398922e-02 1.87025890e-01 -4.52970892e-01 -6.03093088e-01 -1.12855887e+00 -3.78223300e-01 4.08630446e-02 5.77387094e-01 6.96897924e-01 8.13589454e-01 4.75506693e-01 9.16660488e-01 4.52827096e-01 -8.37275088e-01 -6.05722666e-01 -1.45275056e+00 -2.28372082e-01 2.39619613e-01 4.78141785e-01 -4.33446057e-02 -2.49925926e-01 5.20125210e-01]
[12.490280151367188, 9.343859672546387]
2826c839-23f0-440e-8d28-1ca20524f0e4
rank-1-matrix-completion-with-gradient
2212.09396
null
https://arxiv.org/abs/2212.09396v2
https://arxiv.org/pdf/2212.09396v2.pdf
Rank-1 Matrix Completion with Gradient Descent and Small Random Initialization
The nonconvex formulation of matrix completion problem has received significant attention in recent years due to its affordable complexity compared to the convex formulation. Gradient descent (GD) is the simplest yet efficient baseline algorithm for solving nonconvex optimization problems. The success of GD has been witnessed in many different problems in both theory and practice when it is combined with random initialization. However, previous works on matrix completion require either careful initialization or regularizers to prove the convergence of GD. In this work, we study the rank-1 symmetric matrix completion and prove that GD converges to the ground truth when small random initialization is used. We show that in logarithmic amount of iterations, the trajectory enters the region where local convergence occurs. We provide an upper bound on the initialization size that is sufficient to guarantee the convergence and show that a larger initialization can be used as more samples are available. We observe that implicit regularization effect of GD plays a critical role in the analysis, and for the entire trajectory, it prevents each entry from becoming much larger than the others.
['Hye Won Chung', 'Daesung Kim']
2022-12-19
null
null
null
null
['matrix-completion']
['methodology']
[-8.18518102e-02 -1.12748884e-01 -1.92684934e-01 6.68558897e-03 -8.21225047e-01 -6.57869339e-01 1.24360994e-01 9.22029391e-02 -3.97594333e-01 7.55960405e-01 3.42328936e-01 -3.35933059e-01 -2.64549434e-01 -4.40705687e-01 -8.78929555e-01 -9.78827655e-01 -2.49630809e-01 3.58367711e-01 -1.61955342e-01 -3.72515500e-01 4.36809920e-02 2.29744315e-01 -7.90853500e-01 -2.93703347e-01 1.00789797e+00 6.90974712e-01 2.79176503e-01 5.11471212e-01 2.58065075e-01 3.98932397e-01 -1.62694946e-01 -1.70876890e-01 8.41702819e-01 -2.94441581e-01 -7.02322662e-01 2.98955172e-01 2.65215367e-01 -2.24787906e-01 -2.69057065e-01 1.27687657e+00 4.65874434e-01 3.75726014e-01 3.73071879e-01 -1.15514112e+00 -2.55550265e-01 6.08116090e-01 -9.28051174e-01 -2.12062579e-02 1.50345802e-01 -1.81478396e-01 1.16471386e+00 -1.26702952e+00 6.00184679e-01 9.08143938e-01 7.51332521e-01 9.44742411e-02 -1.20067978e+00 -4.62858170e-01 2.97620207e-01 -1.67565987e-01 -1.54418349e+00 -2.61997938e-01 6.19507849e-01 -3.75394672e-01 2.63235807e-01 3.40048283e-01 5.09137988e-01 5.18835843e-01 -2.42320389e-01 8.11093688e-01 8.90012741e-01 -4.62116033e-01 1.55847400e-01 -9.32535306e-02 2.70691812e-01 6.18892193e-01 7.30741262e-01 -1.88650966e-01 -3.31238687e-01 -2.98621982e-01 7.43203282e-01 -3.16752791e-02 -5.84362090e-01 -5.45359850e-01 -1.18827605e+00 1.00685668e+00 5.11767268e-01 1.46019086e-01 -3.48381460e-01 3.22068214e-01 1.96137697e-01 3.65935117e-01 5.45396388e-01 3.98383528e-01 -1.77809328e-01 -1.64804190e-01 -1.00413764e+00 2.69114673e-01 7.25572109e-01 1.01895034e+00 7.32625008e-01 -2.07987186e-02 -3.10874060e-02 6.60634875e-01 1.79510042e-02 7.19853699e-01 -1.62579995e-02 -8.72984231e-01 9.34747338e-01 3.79294932e-01 4.07805026e-01 -1.36075413e+00 -3.52486372e-01 -7.42346346e-01 -1.04299486e+00 -7.47067332e-02 7.34174907e-01 -4.09352511e-01 -4.43294287e-01 1.81408882e+00 4.77084905e-01 3.19669187e-01 -1.72982082e-01 1.17687905e+00 2.88882345e-01 7.18071461e-01 -5.91348648e-01 -3.22944909e-01 8.04060936e-01 -8.00953031e-01 -7.06816018e-01 -1.20663583e-01 7.73403883e-01 -7.88077176e-01 1.10853434e+00 3.86851460e-01 -1.12350023e+00 -3.50031741e-02 -1.06924450e+00 -6.15922920e-02 2.45020106e-01 3.09381723e-01 5.97535312e-01 6.55606151e-01 -9.11800623e-01 4.63189572e-01 -9.79716003e-01 -1.56686068e-01 1.32213995e-01 6.26970828e-01 -3.77362669e-01 -3.27756941e-01 -6.67123914e-01 4.72158879e-01 6.47840053e-02 5.30016899e-01 -6.68470740e-01 -7.00803220e-01 -6.01514637e-01 -1.59717854e-02 5.61490119e-01 -4.28559124e-01 7.45314300e-01 -7.61157453e-01 -1.19190276e+00 6.78195536e-01 -3.94883543e-01 -3.07686865e-01 8.02695811e-01 -4.94386524e-01 2.97352701e-01 2.17504967e-02 1.08803503e-01 -6.42591417e-02 9.33253169e-01 -1.24501956e+00 -4.35121775e-01 -3.74306679e-01 2.98114657e-01 2.94422209e-01 -5.58223546e-01 -1.72782689e-01 -8.00370514e-01 -6.39207780e-01 4.10078853e-01 -1.40536559e+00 -6.52916133e-01 -2.14963645e-01 -5.44797957e-01 8.95770863e-02 3.50601375e-01 -5.04670620e-01 1.44023025e+00 -2.29528594e+00 5.55674314e-01 6.66666150e-01 5.96416473e-01 7.92272314e-02 -1.96674913e-02 5.96452892e-01 6.49461225e-02 1.44651785e-01 -3.47120553e-01 -5.65993428e-01 -1.46579584e-02 2.45079413e-01 -4.21201974e-01 1.08500314e+00 -3.06415588e-01 6.02014422e-01 -9.04875159e-01 -2.16250326e-02 3.65440696e-02 1.19589239e-01 -9.34450686e-01 -1.06737465e-01 2.19507277e-01 5.23449242e-01 -4.49919313e-01 1.82289600e-01 1.02965128e+00 -4.52684939e-01 3.35322708e-01 -2.44725972e-01 -1.26675591e-01 -7.41141513e-02 -1.70451915e+00 1.59773529e+00 -4.45191532e-01 5.90358615e-01 8.11612785e-01 -1.07406986e+00 4.23661679e-01 2.47912541e-01 7.04937279e-01 -1.96792081e-01 1.63434356e-01 4.90373492e-01 -9.34723392e-02 -1.73982829e-01 5.48515141e-01 -4.55196649e-02 1.36361882e-01 5.76007545e-01 -6.22837961e-01 1.78058416e-01 5.28410196e-01 3.86383057e-01 1.00811911e+00 -2.84596026e-01 5.56377657e-02 -6.34057164e-01 5.80594540e-01 6.43490553e-02 8.02821040e-01 7.34700859e-01 1.82230204e-01 6.67221427e-01 4.86255556e-01 -2.40162969e-01 -1.09656680e+00 -8.32707644e-01 -1.31606594e-01 9.31239069e-01 4.44877207e-01 -5.96578002e-01 -6.26691222e-01 -3.44223619e-01 9.94002149e-02 1.01893336e-01 -6.13548756e-01 6.43037781e-02 -9.21518147e-01 -1.10138559e+00 1.97039455e-01 3.52204084e-01 3.66814733e-01 -1.51833653e-01 -6.16916232e-02 1.57611385e-01 -2.54854858e-01 -1.19514024e+00 -8.31901073e-01 -7.03506321e-02 -1.16040266e+00 -1.27882886e+00 -8.37315381e-01 -6.19206309e-01 1.33850813e+00 7.18586624e-01 8.07811439e-01 3.98066014e-01 7.76526704e-02 4.35166597e-01 -2.39325136e-01 -3.66801880e-02 -3.47952805e-02 1.53404832e-01 1.69219792e-01 2.16091156e-01 -2.94724613e-01 -5.89017630e-01 -6.64414763e-01 4.00241077e-01 -9.87881422e-01 2.49668267e-02 3.01406890e-01 8.56176734e-01 5.73798478e-01 8.15326050e-02 2.50601202e-01 -1.09434021e+00 8.95168245e-01 -4.90563393e-01 -8.49262238e-01 6.84259161e-02 -4.47928876e-01 2.21826360e-01 7.10401714e-01 -2.57702440e-01 -5.88785172e-01 1.51529402e-01 9.70874056e-02 -4.41764683e-01 6.93079591e-01 8.58254611e-01 6.73203841e-02 -3.20356727e-01 6.50124013e-01 7.61241000e-03 4.11717966e-03 -5.16168535e-01 4.81794268e-01 2.02179715e-01 3.23425740e-01 -9.63472009e-01 1.19933712e+00 8.90774310e-01 2.80166984e-01 -8.56504261e-01 -9.23849583e-01 -7.53186166e-01 -5.12831450e-01 -5.23503944e-02 3.19929123e-01 -9.77060318e-01 -7.84925222e-01 1.48922242e-02 -7.96752512e-01 -3.60212296e-01 -1.12716720e-01 6.72959507e-01 -2.71555752e-01 4.79011744e-01 -5.16525209e-01 -7.04603255e-01 -1.29943937e-01 -1.13642895e+00 7.18601882e-01 -1.67390630e-01 1.38089299e-01 -1.29703104e+00 1.74858823e-01 8.33873823e-02 3.17573339e-01 4.07788754e-01 4.89099205e-01 -1.62793010e-01 -6.13255739e-01 -3.92464012e-01 -1.77325323e-01 2.45058045e-01 9.39899012e-02 -2.30849117e-01 -3.88123184e-01 -7.76141822e-01 2.97357470e-01 1.35228053e-01 7.97972322e-01 5.08889198e-01 8.01219642e-01 -5.43597579e-01 -2.96216726e-01 8.57390165e-01 1.70837104e+00 -4.37854141e-01 4.59616750e-01 2.81822681e-01 9.39090848e-01 1.77298442e-01 5.76024950e-01 6.37853384e-01 2.23732769e-01 6.00619435e-01 3.10066849e-01 -3.84036630e-01 2.00540379e-01 -2.01233122e-02 3.27282906e-01 1.01467061e+00 -2.86900789e-01 3.65301920e-03 -8.40507865e-01 5.98532021e-01 -2.13871980e+00 -5.70744336e-01 -5.14395237e-01 2.75488687e+00 8.09982955e-01 -1.19906880e-01 2.61119455e-01 2.30299264e-01 7.73504317e-01 -9.85305011e-02 -3.97790790e-01 -3.12014855e-02 -2.54065603e-01 9.50225741e-02 1.02148175e+00 8.58075798e-01 -9.70388114e-01 7.53991902e-01 6.99367857e+00 6.58828497e-01 -1.13017917e+00 6.41261935e-02 3.94984692e-01 -3.25949311e-01 -3.08451533e-01 2.20092565e-01 -6.97913408e-01 3.31498682e-01 3.86804074e-01 -4.14173007e-01 6.08654618e-01 8.01916361e-01 6.18578970e-01 -1.13824315e-01 -9.95550096e-01 1.15018523e+00 -1.20886758e-01 -1.21013069e+00 -3.68581474e-01 3.94058228e-01 1.19124877e+00 5.44658676e-02 1.05710045e-01 -1.49475545e-01 4.70183730e-01 -9.90797043e-01 5.55044115e-01 7.92711973e-02 5.49428463e-01 -7.49637365e-01 6.14378035e-01 4.36447889e-01 -1.29920530e+00 -1.24496043e-01 -6.17023110e-01 -2.76785702e-01 3.52224499e-01 1.07898915e+00 -6.45899057e-01 7.25012660e-01 1.35336637e-01 7.39550531e-01 -2.49881044e-01 1.17519724e+00 -2.97857344e-01 8.76796365e-01 -8.02762270e-01 3.12263966e-01 3.70336324e-01 -7.94037163e-01 8.37680995e-01 1.00321174e+00 2.76077062e-01 9.60029960e-02 6.60626650e-01 5.30977964e-01 -2.35443771e-01 4.76835340e-01 -4.23873335e-01 1.65871397e-01 2.29202166e-01 1.17911899e+00 -5.97545981e-01 -8.55749547e-02 -5.23658216e-01 8.63820314e-01 4.79938865e-01 5.83475113e-01 -8.10890734e-01 -1.81008711e-01 5.15409052e-01 2.64799714e-01 3.69720966e-01 -8.36416006e-01 -5.12068748e-01 -1.29222786e+00 3.90379965e-01 -7.38949001e-01 3.18607897e-01 -2.79727951e-02 -1.11259377e+00 2.54609495e-01 -1.22737110e-01 -1.22350371e+00 -5.44903725e-02 -4.28992778e-01 -4.78816032e-01 6.68076873e-01 -1.26598680e+00 -7.48820543e-01 -3.07341754e-01 6.03820622e-01 8.05669054e-02 2.57842243e-01 3.96759331e-01 7.16027856e-01 -9.28124905e-01 6.53935552e-01 6.08757257e-01 1.76867902e-01 5.62706232e-01 -1.24179411e+00 -1.19141825e-01 1.27350318e+00 1.19340591e-01 9.91640091e-01 9.08453822e-01 -5.89978039e-01 -1.88521254e+00 -8.41439009e-01 5.48051476e-01 -2.69933760e-01 7.83495188e-01 -3.29824120e-01 -6.02040708e-01 7.84093022e-01 -8.94654095e-02 1.60754502e-01 5.50913870e-01 3.26535255e-01 -8.61245841e-02 -2.25164428e-01 -8.06214392e-01 6.67585552e-01 8.57652247e-01 -3.58177334e-01 -3.30469273e-02 6.27981186e-01 3.57028484e-01 -7.02006578e-01 -7.78626204e-01 2.45836020e-01 2.19646454e-01 -6.75930321e-01 9.03150618e-01 -5.10106444e-01 1.57025784e-01 -6.15866005e-01 -2.27679253e-01 -1.05242693e+00 -1.60691574e-01 -1.18078804e+00 -1.83339417e-01 8.59377742e-01 4.54687953e-01 -5.31355977e-01 8.86508048e-01 7.56280839e-01 -1.24647103e-01 -8.95701349e-01 -9.19637084e-01 -1.05263758e+00 1.65054873e-01 -3.50169331e-01 1.59902766e-01 8.56093228e-01 -5.01082372e-03 2.74357051e-01 -7.50647783e-01 3.43242824e-01 7.12328732e-01 2.61974305e-01 1.05813372e+00 -9.48126078e-01 -2.84064353e-01 -2.08630666e-01 -9.48875025e-02 -1.62697279e+00 -1.00478195e-01 -1.10105681e+00 1.80769209e-02 -1.45349908e+00 1.92160949e-01 -9.09142673e-01 1.41444209e-03 1.46209106e-01 -3.99947762e-01 2.41369009e-01 3.71976018e-01 5.07984340e-01 -5.65651953e-01 5.05419075e-01 1.42341459e+00 5.61826192e-02 -4.89197880e-01 9.53967422e-02 -7.45238960e-01 5.96458673e-01 6.27733946e-01 -4.68714207e-01 -4.35523093e-01 -5.52619219e-01 7.42945552e-01 -7.86993802e-02 -3.02350968e-02 -7.68819809e-01 2.55213261e-01 -5.10544553e-02 -1.61601558e-01 -3.28064531e-01 1.89360499e-01 -7.54374564e-01 1.36951059e-01 3.55103731e-01 -2.21304879e-01 1.05592661e-01 -5.61738536e-02 7.22130299e-01 -1.21519044e-01 -4.35228527e-01 7.56351829e-01 8.75150189e-02 -8.03818777e-02 6.05438530e-01 -5.97467013e-02 3.63755077e-01 8.09678555e-01 -1.16519700e-03 2.05312103e-01 -6.08552754e-01 -7.74317980e-01 3.85411888e-01 5.30980527e-01 -1.33527070e-01 3.03321958e-01 -1.43187940e+00 -8.68818343e-01 -2.34133765e-01 -3.25380027e-01 1.25514805e-01 -9.05217454e-02 1.44532859e+00 -6.94302440e-01 3.06328595e-01 4.41167653e-01 -5.91778696e-01 -1.12749290e+00 5.03431082e-01 2.74702534e-02 -4.71759140e-01 -8.46223474e-01 7.32008278e-01 2.46175751e-01 -5.24238497e-02 3.49220097e-01 -5.01687765e-01 2.73307860e-01 -2.04462975e-01 5.47578633e-01 6.34615123e-01 4.18649539e-02 -4.53035355e-01 -1.81006774e-01 6.66601479e-01 -5.22603728e-02 -1.49317563e-01 1.48301792e+00 -4.16368067e-01 -2.87048131e-01 1.88460931e-01 1.28177524e+00 6.74106121e-01 -1.13866639e+00 -2.34096661e-01 -9.51960459e-02 -6.86368763e-01 1.05563641e-01 2.86524743e-02 -1.28736389e+00 6.25809848e-01 2.38240123e-01 2.48285279e-01 1.00215459e+00 -3.63054782e-01 9.24732804e-01 5.19570470e-01 5.72960913e-01 -1.17423868e+00 -4.43174541e-02 6.44917130e-01 9.34283435e-01 -1.18464649e+00 4.93874401e-01 -9.08582270e-01 -3.74280721e-01 9.58145082e-01 5.16720340e-02 -5.81213653e-01 6.44211650e-01 1.01930313e-01 -3.76145154e-01 -9.09960270e-02 -2.69754767e-01 -9.28261578e-02 3.21140200e-01 1.90700769e-01 3.43012184e-01 3.60589661e-02 -6.17869854e-01 3.02797616e-01 -2.23662630e-01 -2.74993598e-01 6.16252303e-01 7.92521656e-01 -3.44118148e-01 -1.38319600e+00 -5.36003828e-01 3.35578144e-01 -5.81456721e-01 -2.60807425e-01 -9.69262868e-02 7.32829154e-01 -3.81257445e-01 1.04158568e+00 -3.36436957e-01 -5.01136389e-03 1.63645416e-01 -5.25430501e-01 6.43516719e-01 -5.55229306e-01 -4.35205519e-01 2.65444100e-01 1.71815343e-02 -5.86624622e-01 -4.45630044e-01 -6.87641859e-01 -1.41469443e+00 -6.14853382e-01 -5.36916316e-01 4.27872509e-01 4.56574440e-01 9.05387342e-01 3.12298000e-01 6.67054206e-02 9.32598710e-01 -6.34645760e-01 -5.44859588e-01 -6.09893441e-01 -6.38839781e-01 4.15662527e-01 3.98447096e-01 -6.46336079e-01 -6.49987698e-01 -1.20419189e-01]
[7.006585121154785, 4.5205183029174805]
da953a83-25eb-4636-b0f4-ba946c173b07
multi-modal-multi-correlation-learning-for
2207.01197
null
https://arxiv.org/abs/2207.01197v1
https://arxiv.org/pdf/2207.01197v1.pdf
Multi-Modal Multi-Correlation Learning for Audio-Visual Speech Separation
In this paper we propose a multi-modal multi-correlation learning framework targeting at the task of audio-visual speech separation. Although previous efforts have been extensively put on combining audio and visual modalities, most of them solely adopt a straightforward concatenation of audio and visual features. To exploit the real useful information behind these two modalities, we define two key correlations which are: (1) identity correlation (between timbre and facial attributes); (2) phonetic correlation (between phoneme and lip motion). These two correlations together comprise the complete information, which shows a certain superiority in separating target speaker's voice especially in some hard cases, such as the same gender or similar content. For implementation, contrastive learning or adversarial training approach is applied to maximize these two correlations. Both of them work well, while adversarial training shows its advantage by avoiding some limitations of contrastive learning. Compared with previous research, our solution demonstrates clear improvement on experimental metrics without additional complexity. Further analysis reveals the validity of the proposed architecture and its good potential for future extension.
['Yan Lu', 'Xiulian Peng', 'Xiangyu Kong', 'Xiaoyu Wang']
2022-07-04
null
null
null
null
['speech-separation']
['speech']
[ 1.11001365e-01 -1.57818899e-01 -1.53961703e-01 -2.39207342e-01 -1.07392180e+00 -5.43320775e-01 8.09558511e-01 -1.42946452e-01 -2.87686795e-01 5.69642365e-01 2.82587260e-01 1.25324847e-02 -1.30480319e-01 -1.71533450e-01 -2.11491257e-01 -1.00855482e+00 -9.79471877e-02 -8.95705223e-02 3.09492815e-02 -1.51137978e-01 1.16866924e-01 4.74746197e-01 -2.04777884e+00 3.05378735e-01 4.66988891e-01 1.16597116e+00 -3.13475020e-02 7.28837430e-01 -1.06948040e-01 6.34205163e-01 -4.38303769e-01 -3.46035749e-01 2.35109046e-01 -4.70626056e-01 -5.39432585e-01 -2.99043339e-02 5.03015637e-01 2.77750101e-02 -1.42348737e-01 9.87064719e-01 9.50911999e-01 1.14536718e-01 7.03160405e-01 -1.45782483e+00 -2.82188743e-01 2.89708555e-01 -6.92064404e-01 2.01786548e-01 6.00778818e-01 1.96739569e-01 1.09926462e+00 -9.46114480e-01 1.54040113e-01 1.30155826e+00 6.79269850e-01 5.21967232e-01 -1.06060147e+00 -8.23646069e-01 -9.05938596e-02 3.35670233e-01 -1.47230661e+00 -8.97581220e-01 9.88059819e-01 -4.90138531e-01 5.57377756e-01 4.44145083e-01 3.74042928e-01 1.10830307e+00 -8.46269876e-02 7.96831071e-01 1.60250437e+00 -5.01791954e-01 -1.42545685e-01 6.13380373e-01 -3.06562334e-01 2.60530949e-01 -5.23534715e-01 2.90389866e-01 -5.72230041e-01 -9.50499773e-02 2.16733694e-01 -2.87472606e-01 -3.48823875e-01 -3.44194233e-01 -8.52772772e-01 5.19568622e-01 1.55010149e-01 5.20977497e-01 -9.70645323e-02 -1.09880082e-01 4.65766430e-01 3.30601722e-01 2.49085620e-01 1.04160771e-01 -1.79868028e-01 -1.39439777e-01 -1.02914488e+00 -5.32410666e-02 4.23572302e-01 6.77048504e-01 6.34974062e-01 2.37681985e-01 -1.21304542e-01 1.09756124e+00 4.25550580e-01 5.69235563e-01 6.31143212e-01 -7.15056598e-01 3.46771747e-01 1.29823208e-01 -1.57528564e-01 -9.48575318e-01 -4.01371986e-01 -3.14489007e-01 -7.69882321e-01 4.57451463e-01 3.35690290e-01 -1.77068830e-01 -6.17523491e-01 1.88838899e+00 1.82563603e-01 4.93080378e-01 1.33391082e-01 8.92597258e-01 9.87452984e-01 4.19325709e-01 1.31878451e-01 -3.17191482e-01 1.31773186e+00 -7.45216191e-01 -9.11659896e-01 -6.03816807e-02 -3.07386648e-02 -1.23268795e+00 8.89368474e-01 2.51946956e-01 -1.07531321e+00 -7.50408471e-01 -1.03448284e+00 2.27414697e-01 -5.00166833e-01 1.84969455e-01 5.72044075e-01 1.05522907e+00 -9.97231185e-01 4.49379116e-01 -3.72022301e-01 -2.36365944e-01 1.08301029e-01 5.70097327e-01 -4.73642230e-01 2.59726793e-01 -1.07513022e+00 6.00639582e-01 -9.56745446e-03 4.31258604e-02 -6.89696670e-01 -3.92569304e-01 -7.55980134e-01 -2.63486058e-02 1.62262425e-01 -4.51363832e-01 9.13783252e-01 -1.21315694e+00 -1.67596889e+00 8.42934906e-01 -1.47875682e-01 -1.78043708e-01 5.79430461e-01 -2.18659848e-01 -6.86548650e-01 1.38577059e-01 -2.55882949e-01 7.33085752e-01 1.30310845e+00 -1.51750922e+00 -6.34870648e-01 -2.74431735e-01 1.26815019e-02 3.09168071e-01 -7.07342386e-01 2.53549278e-01 -5.05764186e-01 -7.54470229e-01 -2.50900742e-02 -9.91304040e-01 2.82251865e-01 -1.43365011e-01 -3.45667452e-01 -5.06286919e-02 9.95477617e-01 -6.39362752e-01 1.06388307e+00 -2.39124060e+00 1.46967009e-01 8.04964602e-02 -1.80669930e-02 4.96537209e-01 -1.90305784e-01 4.54887152e-01 -2.57842511e-01 2.51572654e-02 2.16999818e-02 -6.17536664e-01 -1.26050726e-01 -5.78837506e-02 -1.84200078e-01 5.23457587e-01 2.37112820e-01 5.43145716e-01 -6.60901725e-01 -5.97381115e-01 3.99474949e-01 8.78153741e-01 -3.63048971e-01 2.46649340e-01 3.17197353e-01 6.43900692e-01 -1.20267518e-01 6.13852859e-01 8.52620006e-01 3.91109258e-01 1.83089405e-01 -3.85196537e-01 1.99386273e-02 1.19147561e-01 -1.32149255e+00 1.34777033e+00 -6.44203305e-01 7.39109516e-01 4.23913509e-01 -9.67533767e-01 9.87571776e-01 6.88655734e-01 6.45615995e-01 -5.10993838e-01 9.64872837e-02 1.00041278e-01 4.76462655e-02 -5.89756012e-01 4.13206726e-01 -3.26342225e-01 1.41647086e-01 2.19427064e-01 3.30347210e-01 -1.10449933e-01 -1.93861738e-01 -2.07522914e-01 5.03826618e-01 1.57912061e-01 3.21497738e-01 1.40546868e-02 1.06211317e+00 -7.94793665e-01 4.30749029e-01 3.29813242e-01 -5.81166923e-01 7.67492294e-01 3.81774157e-01 2.91477472e-01 -8.47589612e-01 -1.04385173e+00 -3.12123895e-02 1.07057643e+00 6.29337132e-02 -4.03844059e-01 -4.32187885e-01 -4.84999955e-01 -8.62638801e-02 2.23089918e-01 -5.18359244e-01 -1.61728691e-02 -4.28568453e-01 -4.95257586e-01 7.57022738e-01 4.48670715e-01 4.71110970e-01 -9.26211059e-01 -2.57807970e-01 -2.47313708e-01 -2.06418037e-01 -1.36115932e+00 -3.36365402e-01 9.27067250e-02 -5.28896570e-01 -7.69943595e-01 -8.65569472e-01 -7.60296762e-01 1.45374209e-01 4.09733415e-01 6.44475162e-01 -1.58124879e-01 -2.26661757e-01 5.87777674e-01 -3.14160228e-01 -2.96617419e-01 -5.08080125e-01 -1.85413122e-01 1.54818103e-01 5.02561688e-01 2.01287538e-01 -8.72587442e-01 -4.87911314e-01 2.82704562e-01 -5.48309267e-01 -4.19379234e-01 5.63472688e-01 8.35768223e-01 2.57445961e-01 1.42011747e-01 6.73867404e-01 -3.83193105e-01 6.96263194e-01 -3.09587806e-01 -1.41119167e-01 -3.14157791e-02 -4.50256914e-01 -1.78684875e-01 6.57330990e-01 -4.93428469e-01 -1.09650230e+00 1.68898135e-01 -3.53278190e-01 -6.35986269e-01 -5.17903209e-01 2.08131745e-01 -2.82745034e-01 -8.09801370e-02 1.89580947e-01 2.00103194e-01 2.06924409e-01 -5.16635358e-01 3.63851696e-01 9.48710978e-01 7.03197479e-01 -3.87339592e-01 7.58207202e-01 5.71750283e-01 9.84983295e-02 -9.97450471e-01 -3.21323246e-01 -8.07460725e-01 -7.62393415e-01 -3.32981735e-01 8.14032495e-01 -9.53317225e-01 -1.01774526e+00 4.84817088e-01 -7.76703715e-01 2.58436650e-01 -3.58481593e-02 7.07276523e-01 -5.84930718e-01 6.64697945e-01 -3.53030831e-01 -1.26657176e+00 -1.25847831e-01 -1.17026031e+00 1.09300208e+00 2.20669925e-01 -9.99262109e-02 -9.85646009e-01 9.46067870e-02 4.84406173e-01 3.57321143e-01 9.90638807e-02 6.72284603e-01 -5.64090729e-01 -1.93742856e-01 -1.89474583e-01 -1.55128688e-01 7.07641125e-01 3.60277861e-01 1.59459293e-01 -1.58536780e+00 -3.83777708e-01 6.24882104e-03 -4.04221952e-01 8.60917807e-01 2.06069216e-01 7.46751428e-01 6.70431927e-02 -1.44753486e-01 4.97992784e-01 1.22339499e+00 3.03641438e-01 5.06926358e-01 6.94817156e-02 6.12093508e-01 8.18085432e-01 5.74531674e-01 4.51075524e-01 1.11954667e-01 1.11621332e+00 5.50413549e-01 -3.44855368e-01 -4.85534340e-01 -7.46644288e-02 6.45305634e-01 9.04167950e-01 -2.17244118e-01 -5.91939362e-03 -7.05195725e-01 5.17562091e-01 -1.62125623e+00 -1.28519666e+00 1.72776714e-01 2.38013530e+00 5.85308850e-01 3.08282636e-02 4.20729190e-01 5.66213131e-01 7.38336205e-01 4.35906380e-01 -1.74969718e-01 -5.35708904e-01 -3.58011097e-01 2.44263336e-01 7.68483803e-02 4.63523090e-01 -1.40141475e+00 6.26576245e-01 6.41668844e+00 1.04538059e+00 -1.46101058e+00 1.35932282e-01 2.62418926e-01 -5.50496243e-02 -2.48670056e-01 -1.98330909e-01 -5.53299189e-01 3.78840178e-01 6.65031731e-01 2.14596704e-01 3.72225076e-01 5.45819581e-01 7.48863518e-02 -6.34940015e-03 -8.27324808e-01 1.24971461e+00 2.93806165e-01 -7.21728623e-01 -2.29163200e-01 1.23303765e-02 4.60274130e-01 -3.49447787e-01 5.45810521e-01 2.21609846e-01 -2.50441641e-01 -1.11036348e+00 7.06825554e-01 3.14925373e-01 7.45944023e-01 -8.86704564e-01 7.27068543e-01 1.05229370e-01 -1.48498333e+00 -1.85813159e-01 4.51879539e-02 1.70095563e-01 7.69481063e-02 1.83667064e-01 -7.70393431e-01 8.23821723e-01 6.76736832e-01 6.04562700e-01 -5.41060925e-01 9.25346792e-01 -2.07454473e-01 5.03637075e-01 -1.39057726e-01 2.71315128e-01 1.16076574e-01 -7.81344324e-02 7.97026277e-01 1.48938715e+00 2.19957277e-01 -4.37065154e-01 7.16159940e-02 3.88098598e-01 2.65739590e-01 4.07800019e-01 -6.55437350e-01 6.16338179e-02 4.26052600e-01 1.50302482e+00 -5.10247707e-01 -1.20606743e-01 -6.24200761e-01 8.47022116e-01 5.54833887e-03 1.44692063e-01 -8.49115014e-01 -1.66755289e-01 8.95774722e-01 -1.79862946e-01 5.36330700e-01 -1.12890162e-01 -2.19100863e-01 -7.91179240e-01 5.81484055e-03 -1.01234889e+00 2.36769110e-01 -7.09364355e-01 -1.06703639e+00 7.77988315e-01 -1.08268604e-01 -1.55847573e+00 -4.29633617e-01 -5.13363957e-01 -5.81693053e-01 9.26397145e-01 -1.59857225e+00 -1.43266046e+00 -1.49640083e-01 8.50790322e-01 4.80932474e-01 -5.04325330e-01 8.60907614e-01 6.54503047e-01 -4.98663604e-01 1.00618374e+00 5.00232168e-02 -4.72535789e-02 1.06609190e+00 -1.12380016e+00 -2.19871506e-01 6.47885323e-01 3.61317247e-01 4.88343209e-01 7.42419243e-01 -1.50202215e-01 -1.15538847e+00 -6.03010893e-01 7.77153432e-01 -1.50565565e-01 5.28137147e-01 -2.71438360e-01 -7.15329945e-01 2.49363750e-01 4.67285812e-01 -1.02842905e-01 9.37895060e-01 1.84020713e-01 -6.56742990e-01 -3.42923373e-01 -9.82592940e-01 4.91738319e-01 5.06988227e-01 -9.50305045e-01 -4.73030269e-01 -6.50165975e-02 2.02141985e-01 -9.57678705e-02 -8.60706747e-01 5.78975677e-01 8.80994797e-01 -1.43901873e+00 1.12331724e+00 -3.47062379e-01 2.34913245e-01 -3.57777685e-01 -3.72560233e-01 -1.18399978e+00 -5.77470623e-02 -8.34316194e-01 5.95846027e-02 1.91640365e+00 1.92818418e-01 -4.86636102e-01 4.77817953e-01 -1.36882961e-02 4.19031195e-02 -5.74217260e-01 -1.00465333e+00 -9.02686894e-01 -4.48315851e-02 -6.23020530e-01 4.95399654e-01 1.06778288e+00 5.69710806e-02 4.27951604e-01 -8.24487031e-01 2.36479893e-01 3.54056716e-01 8.25226679e-02 8.59503210e-01 -1.05449200e+00 -3.76994789e-01 -6.29413009e-01 -5.35764396e-01 -6.87828541e-01 3.24001342e-01 -6.70267642e-01 -1.08131491e-01 -1.00301957e+00 2.27629796e-01 -3.48523706e-01 -5.71511269e-01 3.05363864e-01 -2.06910178e-01 5.15542984e-01 4.17766243e-01 1.45322099e-01 -2.26715297e-01 4.47594076e-01 1.09033167e+00 -9.86013189e-02 -2.59210736e-01 3.07644755e-01 -4.97796386e-01 7.62758791e-01 6.96198642e-01 -2.33488336e-01 -4.65268910e-01 6.41928539e-02 -2.13469535e-01 1.97334185e-01 4.11812305e-01 -1.12765729e+00 1.79266855e-01 7.24280551e-02 6.25256225e-02 -3.39521706e-01 9.17163372e-01 -8.85184288e-01 3.80922519e-02 1.40014037e-01 -2.54180640e-01 -1.40083060e-01 3.49079221e-01 4.50695634e-01 -7.25035310e-01 -1.32170424e-01 9.94212747e-01 1.94076031e-01 -6.33620262e-01 -2.93364115e-02 -3.20692241e-01 -2.86500901e-01 1.00756800e+00 -3.01639169e-01 -1.26097545e-01 -7.81636894e-01 -9.28528607e-01 -2.38135248e-01 1.33456752e-01 6.94851279e-01 4.08637404e-01 -1.42207575e+00 -5.87301970e-01 2.78106868e-01 5.94130531e-02 -8.73448431e-01 2.89409161e-01 1.17476559e+00 1.86471388e-01 3.97434205e-01 -4.31501567e-01 -6.68499827e-01 -1.85019338e+00 6.88261390e-01 2.62146562e-01 -7.55177736e-02 -1.82047889e-01 6.61458492e-01 1.49929553e-01 -1.32523835e-01 4.47426319e-01 1.70541182e-01 -6.02524161e-01 5.27839959e-01 3.33785355e-01 4.23897117e-01 2.54379883e-02 -1.20790470e+00 -5.44456065e-01 9.27377760e-01 1.75600871e-01 -2.96344787e-01 1.10199416e+00 -3.96302581e-01 1.18643604e-01 6.46077693e-01 1.35956609e+00 6.35868430e-01 -1.05157781e+00 -1.38790607e-01 -1.98388919e-01 -4.56728220e-01 -2.13381946e-02 -6.17867768e-01 -1.07990348e+00 1.28160369e+00 9.37095344e-01 4.07637328e-01 1.50468099e+00 -8.73003155e-02 4.34519589e-01 -2.47467518e-01 -4.93808389e-02 -8.92267525e-01 2.63972044e-01 2.28107020e-01 7.91094720e-01 -1.26299310e+00 3.34915146e-02 -5.36740363e-01 -8.23735535e-01 1.11533535e+00 4.66592282e-01 1.65187821e-01 6.31785333e-01 2.87415653e-01 4.03580129e-01 1.54624939e-01 -5.77475429e-01 -7.56674051e-01 6.26972914e-01 9.50347900e-01 6.19532883e-01 -1.23547979e-01 -1.73063710e-01 2.92946517e-01 -1.86511070e-01 -4.64650959e-01 1.33257240e-01 4.80035335e-01 -2.33184963e-01 -1.23964536e+00 -6.34432316e-01 -7.71568045e-02 -7.67533004e-01 -1.36910109e-02 -3.65668058e-01 9.00946796e-01 2.59956539e-01 1.22673190e+00 -8.78385529e-02 -6.67057395e-01 2.84997076e-01 2.32167453e-01 4.52703625e-01 -2.41129220e-01 -7.65565097e-01 5.91602981e-01 5.05972654e-02 -3.92476469e-01 -8.60535085e-01 -7.99428046e-01 -7.80210674e-01 -3.64988074e-02 -3.26072663e-01 -1.05556339e-01 7.56670535e-01 9.03505445e-01 8.66082236e-02 4.91333067e-01 9.65335250e-01 -9.05583084e-01 -4.96930391e-01 -1.01570392e+00 -6.13313317e-01 5.88240504e-01 5.81868410e-01 -7.85914302e-01 -5.34844458e-01 1.07728519e-01]
[14.411968231201172, 5.107040882110596]
786bfd49-a47c-4ed3-be35-f5315db1fc07
inference-with-hybrid-bio-hardware-neural
1905.11594
null
https://arxiv.org/abs/1905.11594v2
https://arxiv.org/pdf/1905.11594v2.pdf
Inference with Hybrid Bio-hardware Neural Networks
To understand the learning process in brains, biologically plausible algorithms have been explored by modeling the detailed neuron properties and dynamics. On the other hand, simplified multi-layer models of neural networks have shown great success on computational tasks such as image classification and speech recognition. However, the computational models that can achieve good accuracy for these learning applications are very different from the bio-plausible models. This paper studies whether a bio-plausible model of a in vitro living neural network can be used to perform machine learning tasks and achieve good inference accuracy. A novel two-layer bio-hardware hybrid neural network is proposed. The biological layer faithfully models variations of synapses, neurons, and network sparsity in in vitro living neural networks. The hardware layer is a computational fully-connected layer that tunes parameters to optimize for accuracy. Several techniques are proposed to improve the inference accuracy of the proposed hybrid neural network. For instance, an adaptive pre-processing technique helps the proposed neural network to achieve good learning accuracy for different living neural network sparsity. The proposed hybrid neural network with realistic neuron parameters and variations achieves a 98.3% testing accuracy for the handwritten digit recognition task on the full MNIST dataset.
['Zubayer Ibne Ferdous', 'Zhiyuan Yan', 'Yuan Zeng', 'Weixiang Zhang', 'Xiaochen Guo', 'Drew Patel', 'Yevgeny Berdichevsky', 'Mufan Xu', 'Anlan Yu']
2019-05-28
null
null
null
null
['handwritten-digit-recognition']
['computer-vision']
[ 3.21419448e-01 1.46616399e-01 9.13819596e-02 -2.14603081e-01 4.58858818e-01 1.85710490e-01 4.96858478e-01 -9.53527689e-02 -5.73153079e-01 8.14217329e-01 -4.95974511e-01 -7.46768340e-02 -2.38242179e-01 -6.58938408e-01 -1.05252707e+00 -1.18854761e+00 1.05262913e-01 4.46453631e-01 4.50187773e-01 -6.09227791e-02 4.23101187e-01 7.11412013e-01 -1.90167856e+00 3.78895313e-01 7.47875988e-01 1.00330663e+00 5.43710470e-01 6.69542253e-01 -1.73905581e-01 8.64289343e-01 -7.02036262e-01 8.22818130e-02 8.90285671e-02 -2.66141027e-01 -1.21250987e-01 -2.92981893e-01 1.30789220e-01 7.14517161e-02 -1.50029153e-01 1.02381730e+00 6.65951490e-01 -2.10693702e-01 9.45344031e-01 -9.45254683e-01 -5.27533889e-01 9.35962617e-01 9.77009684e-02 4.06262547e-01 -5.05262554e-01 1.93808705e-01 -7.89589360e-02 -5.35757244e-01 1.73544228e-01 1.24663842e+00 6.65884435e-01 9.37478483e-01 -1.13884377e+00 -1.07577491e+00 -2.84451693e-01 1.85532689e-01 -1.43878818e+00 -2.96902001e-01 4.85139430e-01 -4.07922834e-01 1.12632883e+00 -1.97605252e-01 8.47146332e-01 1.03253126e+00 1.08741486e+00 1.75485224e-01 1.20186150e+00 -3.30610693e-01 7.06582606e-01 1.68525040e-01 5.01964986e-01 8.95543873e-01 8.12955201e-01 3.75094488e-02 -6.38525009e-01 1.30631328e-01 1.22411299e+00 6.89969957e-02 -1.41577050e-01 2.75763720e-01 -1.03962839e+00 5.14410019e-01 3.58398259e-01 6.83757901e-01 -5.57020366e-01 4.69987184e-01 1.06292658e-01 -1.74017772e-01 -1.42258406e-01 5.73974609e-01 -4.98292565e-01 3.91000777e-01 -8.80283773e-01 -1.62294991e-02 7.78704822e-01 5.35311639e-01 5.74521601e-01 6.71885669e-01 2.77411878e-01 9.18397903e-01 3.81579518e-01 3.47733706e-01 1.10044563e+00 -9.11911309e-01 -9.82560068e-02 6.97478890e-01 -5.00422716e-01 -7.96200931e-01 -6.05360389e-01 -5.94684601e-01 -1.43170440e+00 4.87071365e-01 3.89499724e-01 -2.00173989e-01 -9.44776475e-01 1.42589915e+00 1.04188621e-01 5.24074137e-01 4.27027375e-01 6.59745991e-01 8.91074002e-01 1.01292408e+00 -5.47580309e-02 -3.61653447e-01 1.18743300e+00 -7.71397948e-01 -7.13638008e-01 -1.73101887e-01 4.81303662e-01 -1.13207787e-01 6.94798410e-01 5.71064353e-01 -1.13321865e+00 -5.70164025e-01 -1.59526300e+00 1.35273516e-01 -6.61290288e-01 7.05854520e-02 4.22311723e-01 5.79954743e-01 -8.66601825e-01 8.01760256e-01 -1.14576447e+00 -4.51044410e-01 3.99238974e-01 8.72068226e-01 6.74777925e-02 3.73827368e-02 -8.95974576e-01 9.46753263e-01 9.33586359e-01 2.31576607e-01 -1.02502477e+00 -6.17585003e-01 -2.84005135e-01 2.26439148e-01 -1.98864952e-01 -9.00571465e-01 6.36614084e-01 -8.21146548e-01 -1.82843113e+00 4.83841896e-01 -1.76960006e-01 -9.62861359e-01 -7.10041374e-02 6.00818992e-01 -1.65536329e-01 9.94297788e-02 -7.85915434e-01 9.18660223e-01 5.66077769e-01 -1.15067029e+00 -1.54933125e-01 -3.48122895e-01 -3.29224139e-01 -2.37948611e-01 -5.95451295e-01 -3.83997232e-01 1.36812478e-01 -6.41537845e-01 3.11915904e-01 -8.39062929e-01 5.22804111e-02 4.41944629e-01 5.83518073e-02 -4.18059602e-02 7.88144410e-01 -2.73703188e-01 6.53686762e-01 -1.81186903e+00 9.50173475e-04 3.76810841e-02 2.00237706e-01 5.25887251e-01 -9.20498669e-02 -1.68238878e-01 2.61233836e-01 1.95089485e-02 -9.66661200e-02 1.82753783e-02 -5.67284107e-01 5.76643288e-01 2.23868176e-01 3.69122684e-01 1.29533529e-01 7.08740115e-01 -2.17584908e-01 -5.38021326e-01 8.99428651e-02 1.05707395e+00 -4.38639164e-01 1.08190320e-01 -2.42070287e-01 2.96369433e-01 -2.40332007e-01 5.42087734e-01 3.31631571e-01 -2.62629539e-01 1.39682502e-01 -2.99513429e-01 2.27178098e-03 -7.60377944e-02 -9.14500356e-01 1.28431976e+00 -4.96107578e-01 6.96852922e-01 1.20741285e-01 -1.37451506e+00 1.40298617e+00 2.74368256e-01 1.17286421e-01 -8.43342066e-01 4.55009878e-01 3.39578331e-01 5.79289556e-01 -3.38810682e-01 -4.06733960e-01 -3.28779787e-01 7.07368255e-01 1.45756096e-01 3.13530803e-01 -9.39229652e-02 4.41250876e-02 -4.10151482e-01 8.17320168e-01 -2.21509561e-01 1.64304599e-01 -8.92342985e-01 7.56728590e-01 -1.18674956e-01 6.77772462e-01 5.70497096e-01 -5.92792779e-02 3.12977999e-01 7.27611482e-02 -6.65362418e-01 -1.13775098e+00 -5.41040480e-01 -3.63790363e-01 3.90424103e-01 1.67340353e-01 5.14114141e-01 -1.09078526e+00 4.09523696e-01 -2.36780420e-01 5.51324248e-01 -7.24455833e-01 -5.06852865e-01 -5.70041835e-01 -1.13427615e+00 8.55934739e-01 3.29831809e-01 9.54580665e-01 -1.19117796e+00 -8.96861017e-01 4.75038588e-01 3.86162192e-01 -1.02366447e+00 4.35692579e-01 6.22909844e-01 -1.30123734e+00 -8.69726002e-01 -4.82499927e-01 -1.12176895e+00 6.68729126e-01 -1.57700643e-01 4.42940384e-01 4.31754202e-01 -4.93205279e-01 -1.85583413e-01 4.05833535e-02 -8.53647888e-01 -6.09756529e-01 1.81908421e-02 4.87879127e-01 -1.88505620e-01 9.51348618e-02 -7.68814385e-01 -3.69719774e-01 2.32671455e-01 -9.02174652e-01 4.04298306e-01 7.95889080e-01 1.14923215e+00 9.50447321e-01 4.48965371e-01 7.69332767e-01 -4.85827386e-01 3.19229215e-01 -2.65256763e-01 -4.68969584e-01 4.00471687e-01 -8.00427139e-01 5.36621928e-01 9.65126872e-01 -1.03537190e+00 -8.90526891e-01 2.77801026e-02 -2.65872538e-01 -1.73282966e-01 -2.20385775e-01 5.60861051e-01 -9.45684686e-02 -4.04209793e-01 7.44405329e-01 5.80232799e-01 3.16544235e-01 -2.24905804e-01 -6.49558783e-01 5.80042601e-01 3.95586878e-01 -5.53419173e-01 1.04823187e-01 3.39295536e-01 3.51399302e-01 -1.21297228e+00 -3.69893819e-01 3.24084878e-01 -4.88624543e-01 -1.81961060e-01 6.81793332e-01 -5.91389775e-01 -9.53573883e-01 9.80607629e-01 -1.08770132e+00 -7.92551458e-01 1.79693282e-01 7.99380541e-01 -4.37624007e-01 -2.02926487e-01 -9.40771282e-01 -9.47878420e-01 -6.16701901e-01 -1.17119384e+00 2.71527320e-01 7.23874688e-01 1.19598664e-01 -1.11043501e+00 -2.26100191e-01 1.54167950e-01 6.12314999e-01 2.43934751e-01 1.32069850e+00 -6.34767413e-01 -3.13003480e-01 7.82573372e-02 3.94285321e-02 2.18584314e-01 -8.46140087e-02 2.89342910e-01 -1.06184697e+00 -1.87134683e-01 3.00144821e-01 -3.29730213e-01 8.32679093e-01 7.65329480e-01 1.20368743e+00 -5.81778109e-01 -4.55738813e-01 4.64678764e-01 1.62090468e+00 6.28501534e-01 7.43340969e-01 3.61138761e-01 3.07068199e-01 5.60371220e-01 -1.50506079e-01 2.07979321e-01 -2.55365651e-02 2.28438944e-01 6.05435133e-01 1.61061019e-01 -1.35288417e-01 1.62911624e-01 2.73819417e-01 1.25648153e+00 -1.34829998e-01 -3.69450688e-01 -1.08957410e+00 2.01216727e-01 -1.65337634e+00 -9.13711488e-01 -9.40014888e-03 1.86298609e+00 7.85615325e-01 4.72227335e-01 -2.26707876e-01 4.96319026e-01 7.92681158e-01 -5.42418480e-01 -1.02178478e+00 -3.15741956e-01 -5.90750158e-01 2.61752754e-01 4.71275777e-01 2.52305269e-01 -3.10186863e-01 5.17013073e-01 6.08582878e+00 6.18073821e-01 -1.69686067e+00 -4.65488918e-02 7.03021586e-01 -2.85048097e-01 2.88206339e-01 -4.24335688e-01 -1.21722353e+00 6.24082327e-01 1.47417295e+00 -6.37339503e-02 5.97428441e-01 4.32362944e-01 3.58768880e-01 -1.03929304e-01 -1.06626606e+00 1.16111219e+00 -5.81353754e-02 -1.89810836e+00 5.91707110e-01 7.46105686e-02 5.76013505e-01 -3.38644907e-02 -7.97213167e-02 1.48821607e-01 -2.19797701e-01 -1.23275506e+00 7.74762690e-01 1.03713691e+00 2.43636027e-01 -7.13619828e-01 1.02283418e+00 9.71291244e-01 -8.53468895e-01 -3.47539008e-01 -8.82383466e-01 -3.01183641e-01 -3.09484422e-01 5.69458544e-01 -6.56200826e-01 -2.02884302e-01 7.79597700e-01 5.48278093e-01 -6.72587335e-01 1.04522109e+00 5.55054069e-01 7.80544996e-01 -5.91608346e-01 -6.14710152e-01 -3.26453075e-02 7.80206248e-02 -3.22234109e-02 1.03999579e+00 5.30960202e-01 3.60848367e-01 -4.50174034e-01 1.18434131e+00 7.89913461e-02 -1.28988355e-01 -3.53083998e-01 -4.99607146e-01 6.93933070e-01 1.05119312e+00 -1.07932842e+00 -4.35298830e-01 5.86211197e-02 1.87788159e-01 4.35401738e-01 1.54396534e-01 -6.89888477e-01 -1.94798440e-01 4.67604160e-01 2.85410315e-01 1.61631867e-01 -4.59745884e-01 -7.29088664e-01 -7.75117338e-01 -4.85350609e-01 -6.08958960e-01 -2.02859417e-01 -7.66832709e-01 -8.01062346e-01 5.19776583e-01 -3.54987800e-01 -6.23903334e-01 3.10701262e-02 -1.22527742e+00 -6.39543712e-01 5.30817151e-01 -1.07503867e+00 -7.03533649e-01 -3.36152881e-01 4.07781899e-01 5.64299524e-01 -5.50206006e-01 7.63129771e-01 3.03696585e-03 -1.12775350e+00 5.19649923e-01 2.39352897e-01 -1.45891279e-01 1.44838616e-01 -5.37846863e-01 -1.18217640e-01 6.60929859e-01 -8.96906629e-02 8.90735865e-01 7.48785257e-01 -4.95754361e-01 -1.54511666e+00 -1.15688062e+00 3.74899954e-01 3.83759081e-01 1.64937884e-01 -3.01119208e-01 -1.31278813e+00 2.34127596e-01 4.14723344e-02 -2.41679493e-02 4.48566675e-01 -7.32244849e-01 2.97845844e-02 -3.57621163e-01 -1.36337137e+00 6.16123855e-01 6.31877124e-01 -1.66538376e-02 -4.74344701e-01 1.34808317e-01 3.36838275e-01 9.38816965e-02 -1.12020183e+00 4.98268217e-01 8.74108195e-01 -7.13640273e-01 9.21914816e-01 -1.35695606e-01 3.03521901e-01 -3.83993804e-01 -9.47940499e-02 -9.70501304e-01 -2.09218860e-01 -1.08576842e-01 -4.76502806e-01 9.17076945e-01 3.97868961e-01 -1.03755713e+00 8.87730479e-01 6.42233849e-01 -3.06951385e-02 -1.25017047e+00 -8.53963852e-01 -9.73501384e-01 3.35350007e-01 1.73238069e-01 3.65879953e-01 6.04469180e-01 -1.25821561e-01 2.31545717e-01 1.82937711e-01 1.33767635e-01 8.70919108e-01 -5.57715952e-01 2.65371472e-01 -1.58601189e+00 -1.68710932e-01 -6.00424528e-01 -6.80347562e-01 -6.85399592e-01 2.77523369e-01 -6.09944522e-01 3.12877931e-02 -1.48811519e+00 1.80069104e-01 -5.26570022e-01 -2.07278281e-01 3.36340159e-01 3.06559831e-01 2.65791744e-01 -1.68054014e-01 4.26783890e-01 7.04768971e-02 3.21118563e-01 1.09424365e+00 -3.00290585e-01 -6.86811581e-02 -4.64293122e-01 -2.06723049e-01 7.07324266e-01 1.27971947e+00 -7.18390584e-01 -4.22566772e-01 -5.47396898e-01 -1.58044606e-01 7.53737688e-02 4.20329124e-01 -1.76139855e+00 8.51385295e-01 -1.50483742e-01 7.56957948e-01 -3.11927646e-01 5.11381269e-01 -7.56647348e-01 4.89182472e-01 1.19401360e+00 -5.24641931e-01 -1.59351364e-01 2.29132771e-01 4.39031839e-01 -2.35867947e-02 -5.12122512e-01 1.35294926e+00 -2.77423710e-01 -2.68721879e-01 -4.95545147e-03 -9.12781000e-01 -5.25466859e-01 9.83362138e-01 -8.04955721e-01 -5.72136581e-01 2.58596629e-01 -6.72554255e-01 -1.38995051e-01 3.24542940e-01 -4.89385016e-02 7.86096215e-01 -9.78783906e-01 -3.56149256e-01 2.46018708e-01 -3.45901579e-01 -3.26226130e-02 4.28420193e-02 6.61776960e-01 -9.41038370e-01 6.84253037e-01 -1.06657672e+00 -6.76013887e-01 -9.61616457e-01 5.88065982e-01 1.01855850e+00 8.10700431e-02 -2.15292305e-01 6.80258393e-01 -4.56004180e-02 -2.09556714e-01 3.98747057e-01 -6.27005517e-01 -3.82310480e-01 -5.39332628e-01 6.36353076e-01 4.44184214e-01 1.20565901e-02 -4.12377566e-01 -2.37327442e-01 6.62872195e-01 6.25319183e-02 1.20354876e-01 1.55837333e+00 7.56119490e-02 -5.64690493e-02 6.92960441e-01 8.61340642e-01 -9.23409998e-01 -1.23359191e+00 -2.81476714e-02 -1.26705289e-01 5.50296724e-01 2.46578708e-01 -9.31356549e-01 -1.17207301e+00 1.21662366e+00 8.21826100e-01 -1.72337145e-01 9.00388956e-01 -5.20047843e-01 6.42685890e-01 1.01835954e+00 6.52162194e-01 -1.03372848e+00 1.54105276e-01 4.77780610e-01 7.76187479e-01 -8.99437308e-01 -7.44078904e-02 -1.83192715e-01 -1.17612794e-01 1.63656271e+00 1.05908680e+00 -5.89207232e-01 8.81858885e-01 8.36118817e-01 -1.75700665e-01 -2.35034395e-02 -1.07473564e+00 4.57021475e-01 -2.18184423e-02 5.87535024e-01 3.82291734e-01 -2.80885130e-01 -3.20075959e-01 8.67984951e-01 -8.22407380e-02 3.16987365e-01 8.54415834e-01 8.30657184e-01 -8.35547328e-01 -3.94689620e-01 -3.25456232e-01 6.12265408e-01 -2.75244445e-01 1.22145422e-01 -3.30234706e-01 5.87259829e-01 1.96761459e-01 7.15817988e-01 1.88047692e-01 -5.31207740e-01 1.46767264e-02 2.82837480e-01 6.68628693e-01 -5.00415325e-01 -7.39380538e-01 -2.91729063e-01 -4.31790560e-01 -9.36315805e-02 -5.22181749e-01 -1.93151295e-01 -1.94463396e+00 -4.11785722e-01 -4.18332607e-01 -3.29481103e-02 1.09563410e+00 1.17367947e+00 3.36181074e-01 6.91482246e-01 -6.73130155e-02 -1.17061770e+00 -1.65490553e-01 -9.56445634e-01 -5.83490968e-01 -2.37348244e-01 6.60978183e-02 -9.07089353e-01 -3.64985138e-01 4.70407605e-01]
[8.157161712646484, 2.7445082664489746]
5bb058a2-2001-46ed-94f8-9fc46cb9a3bc
renderdiffusion-text-generation-as-image
2304.12519
null
https://arxiv.org/abs/2304.12519v2
https://arxiv.org/pdf/2304.12519v2.pdf
GlyphDiffusion: Text Generation as Image Generation
Diffusion models have become a new generative paradigm for text generation. Considering the discrete categorical nature of text, in this paper, we propose GlyphDiffusion, a novel diffusion approach for text generation via text-guided image generation. Our key idea is to render the target text as a glyph image containing visual language content. In this way, conditional text generation can be cast as a glyph image generation task, and it is then natural to apply continuous diffusion models to discrete texts. Specially, we utilize a cascaded architecture (ie a base and a super-resolution diffusion model) to generate high-fidelity glyph images, conditioned on the input text. Furthermore, we design a text grounding module to transform and refine the visual language content from generated glyph images into the final texts. In experiments over four conditional text generation tasks and two classes of metrics (ie quality and diversity), GlyphDiffusion can achieve comparable or even better results than several baselines, including pretrained language models. Our model also makes significant improvements compared to the recent diffusion model.
['Ji-Rong Wen', 'Jian-Yun Nie', 'Wayne Xin Zhao', 'Junyi Li']
2023-04-25
null
null
null
null
['conditional-text-generation']
['natural-language-processing']
[ 5.36697268e-01 2.30190694e-01 2.32091650e-01 -1.59777731e-01 -7.45849609e-01 -4.75472003e-01 1.28658724e+00 -1.67521805e-01 -5.54398634e-02 7.44663060e-01 7.10460663e-01 -1.85737759e-01 4.39300895e-01 -1.21589255e+00 -6.29179895e-01 -6.34453833e-01 6.26702964e-01 5.56452274e-01 9.22613367e-02 -3.16274881e-01 3.61180037e-01 1.90143421e-01 -1.28259015e+00 6.11246049e-01 1.30115509e+00 5.12941837e-01 6.67874813e-01 8.97022545e-01 -5.92760742e-01 9.21405256e-01 -8.35756540e-01 -5.55487335e-01 -4.68622241e-03 -1.07953393e+00 -7.23881781e-01 4.82255459e-01 4.28293586e-01 -4.59373444e-01 -3.29869807e-01 9.16177690e-01 5.61743081e-01 1.28262937e-01 1.05458629e+00 -1.08781528e+00 -1.54586780e+00 8.80063891e-01 -6.70789123e-01 -2.72255003e-01 4.62987393e-01 1.42420277e-01 1.00951552e+00 -1.36573339e+00 1.00675356e+00 1.56577444e+00 2.11641788e-01 8.28066766e-01 -1.41189122e+00 -3.34234476e-01 4.20048237e-01 -3.59271437e-01 -1.13475692e+00 -2.47018561e-01 8.43449831e-01 -5.16207218e-01 5.30616999e-01 1.54931903e-01 6.58412576e-01 1.52180040e+00 3.41595829e-01 1.13318074e+00 9.28029895e-01 -5.26198924e-01 2.28455916e-01 1.94909111e-01 -6.89012766e-01 7.61832774e-01 1.35866096e-02 -2.68866777e-01 -6.64209664e-01 -1.41843081e-01 1.01155293e+00 -5.01082912e-02 -2.83873051e-01 -2.14216188e-01 -1.47680652e+00 1.08256626e+00 4.22112226e-01 2.04580560e-01 -4.34269428e-01 1.71571866e-01 -2.77698096e-02 7.21339434e-02 9.68823552e-01 3.52340460e-01 4.70997185e-01 2.24731207e-01 -1.24992716e+00 6.17199242e-01 5.85710466e-01 1.09449065e+00 6.57224417e-01 2.19361797e-01 -7.80601084e-01 7.98447251e-01 5.09946167e-01 5.12785733e-01 5.29227972e-01 -4.98336077e-01 7.51703203e-01 5.51801205e-01 9.34219733e-02 -9.45395172e-01 1.56636178e-01 -1.65833198e-02 -1.19516349e+00 2.39310861e-01 -3.86226736e-02 -4.39481497e-01 -1.32753050e+00 1.43597603e+00 1.41606078e-01 -1.85032666e-01 1.02128290e-01 6.67836964e-01 8.41241419e-01 1.34639537e+00 -1.32802995e-02 -1.45935491e-01 9.47650313e-01 -1.28541183e+00 -8.36500525e-01 -1.55277699e-01 2.22453773e-01 -9.53095257e-01 1.26883101e+00 1.82025373e-01 -1.40800226e+00 -7.00390458e-01 -8.41299832e-01 -3.49457085e-01 -2.67789155e-01 1.49371058e-01 2.59359181e-01 3.94186884e-01 -1.58075428e+00 2.72147834e-01 -4.91035849e-01 -3.27659935e-01 2.30258301e-01 -2.90932924e-01 1.74357593e-01 -4.72922623e-02 -1.13793647e+00 3.69109392e-01 3.22694123e-01 -6.50753975e-02 -1.15544605e+00 -5.06147742e-01 -8.13944519e-01 -1.93314284e-01 -6.17195442e-02 -1.08463645e+00 1.02100396e+00 -8.01222265e-01 -1.56212080e+00 7.04417109e-01 -2.35977352e-01 -4.01169747e-01 9.77108717e-01 1.16134249e-02 -8.05207342e-02 8.72501880e-02 3.34921181e-01 1.19848967e+00 1.34372663e+00 -1.61387956e+00 -6.67441130e-01 7.48063549e-02 -1.41351640e-01 5.62150180e-01 -3.61601830e-01 -2.80585676e-01 -5.86968720e-01 -1.24086022e+00 -1.38759106e-01 -7.44217575e-01 -4.67089802e-01 1.37862727e-01 -6.52346015e-01 -2.36311659e-01 9.34615672e-01 -7.00006902e-01 1.06870019e+00 -1.87313199e+00 5.15601337e-01 1.80058833e-02 4.55673873e-01 -1.42250627e-01 -3.30936283e-01 7.04323769e-01 1.31268442e-01 4.27967280e-01 -3.95880550e-01 -7.98218191e-01 2.01706484e-01 -6.38538897e-02 -8.44444752e-01 -2.74515420e-01 2.97554046e-01 1.26819849e+00 -7.80904770e-01 -6.44248068e-01 1.49613604e-01 7.28196859e-01 -6.16612136e-01 3.29675168e-01 -8.10262382e-01 5.71987689e-01 -4.73210245e-01 1.97104111e-01 6.71816766e-01 -4.12824959e-01 -3.26945409e-02 7.83168599e-02 -5.25142401e-02 -4.30741534e-03 -8.03363681e-01 2.00381994e+00 -5.43399990e-01 7.85001218e-01 -3.29076886e-01 -4.09819126e-01 1.21906722e+00 2.92154670e-01 1.70506969e-01 -5.40313244e-01 -1.26839653e-01 -1.31453350e-01 -5.99570453e-01 -8.55109766e-02 1.16217506e+00 -4.32224795e-02 -8.15623850e-02 8.81174862e-01 -1.52616084e-01 -6.49309456e-01 4.22763079e-01 6.94919944e-01 6.30868495e-01 2.35904783e-01 -3.01175267e-01 -8.18953291e-02 3.22266340e-01 -3.65234613e-02 -1.23546019e-01 8.06117594e-01 5.12852788e-01 1.16334701e+00 5.19013286e-01 -2.74358481e-01 -1.26046038e+00 -1.27605009e+00 2.81090260e-01 9.23089385e-01 6.50739819e-02 -4.70638037e-01 -1.00596964e+00 -7.22610116e-01 -3.32821310e-01 7.87417114e-01 -7.67731428e-01 -8.49267915e-02 -5.37092030e-01 -9.65352952e-01 3.80555481e-01 4.70133692e-01 7.60822296e-01 -1.18755698e+00 -7.84485489e-02 2.46494129e-01 -5.28319895e-01 -9.71748948e-01 -9.02943373e-01 -3.95018399e-01 -7.84035563e-01 -2.60676384e-01 -1.59565425e+00 -1.09330690e+00 8.64175975e-01 1.65335998e-01 1.18524599e+00 -4.91394177e-02 -1.65047824e-01 1.13794550e-01 -5.08777678e-01 -1.42680913e-01 -8.27328682e-01 8.10258016e-02 -5.09904385e-01 3.51956725e-01 -3.90998423e-01 -2.96504200e-01 -7.28600323e-01 -1.10951863e-01 -1.27108502e+00 7.11048543e-01 5.46262145e-01 8.39300871e-01 7.56016970e-01 4.05829251e-02 4.05819595e-01 -9.50872481e-01 1.40336335e+00 -2.18795136e-01 -4.52089220e-01 3.61165106e-01 -5.93876779e-01 2.71575302e-01 7.03014672e-01 -4.22584414e-01 -1.58524668e+00 -2.04817995e-01 -1.12097956e-01 -2.80012071e-01 8.00234824e-02 4.33847010e-01 1.08633358e-02 3.53167355e-01 7.60614216e-01 6.66516006e-01 -3.35516065e-01 -4.41359162e-01 8.21693659e-01 5.37339211e-01 2.60955960e-01 -5.61537325e-01 9.10587966e-01 6.43496215e-01 -3.67942125e-01 -7.13474274e-01 -6.01428330e-01 3.50752652e-01 -4.64791477e-01 -1.25315711e-01 1.26964414e+00 -9.14675236e-01 2.87641436e-02 5.91730475e-01 -1.38079143e+00 -6.83710515e-01 -4.42222476e-01 -1.00764886e-01 -5.73946416e-01 4.45827276e-01 -6.67074978e-01 -6.77064836e-01 -5.41181505e-01 -1.04770267e+00 1.42286921e+00 1.44934565e-01 3.74425314e-02 -1.28128994e+00 1.74001530e-01 6.25023246e-02 3.95055532e-01 4.36579168e-01 8.75091970e-01 2.04230435e-02 -7.72433937e-01 1.08937830e-01 -4.23670173e-01 1.37244239e-01 2.13690966e-01 1.96308047e-01 -6.67806447e-01 -2.44852543e-01 -1.66953653e-01 -2.70050347e-01 1.16944575e+00 2.90376991e-01 1.03329945e+00 -4.09487098e-01 -3.50039929e-01 6.34275615e-01 1.21774566e+00 1.49862766e-01 7.46499121e-01 5.84919564e-03 1.00985813e+00 5.40697157e-01 3.24775875e-01 5.36711335e-01 5.11762261e-01 4.95584399e-01 5.32092229e-02 -5.05359471e-01 -5.99697828e-01 -9.82617855e-01 4.25729960e-01 7.48285294e-01 5.60462885e-02 -1.00267041e+00 -8.11194241e-01 5.60913384e-01 -1.78816915e+00 -9.68181968e-01 -1.94024995e-01 1.73495817e+00 8.82951379e-01 6.66608289e-02 5.14379330e-02 -1.79064229e-01 8.73511493e-01 5.93024611e-01 -3.63258392e-01 -2.67626792e-01 -2.35228270e-01 3.73325944e-02 -8.15072209e-02 6.20533407e-01 -7.47836590e-01 1.29213572e+00 6.70024633e+00 8.60489547e-01 -1.07092643e+00 -3.34951989e-02 8.97228360e-01 8.17745924e-02 -8.45539927e-01 -2.67390579e-01 -8.68971825e-01 4.87936497e-01 4.90115106e-01 -5.31912446e-01 2.69893706e-01 6.02063179e-01 2.94208467e-01 5.07438555e-02 -8.59948218e-01 8.46800029e-01 3.98891598e-01 -1.54037690e+00 1.00746417e+00 4.14920039e-02 1.44232631e+00 -4.47672307e-01 4.94323581e-01 2.94913314e-02 8.23095322e-01 -1.02062011e+00 8.77687573e-01 6.33535743e-01 1.05739367e+00 -7.41856337e-01 7.63694793e-02 3.25776249e-01 -1.31198883e+00 3.68979812e-01 -4.50940937e-01 3.22410583e-01 5.21961868e-01 6.76187992e-01 -7.97649264e-01 5.14101505e-01 4.13638264e-01 7.83877075e-01 -5.81777990e-01 6.26690567e-01 -5.00561118e-01 2.48609215e-01 2.45609283e-01 -1.48739740e-01 3.98724973e-01 -1.45598754e-01 3.71995687e-01 1.31114924e+00 5.83101034e-01 -2.30780259e-01 1.54927269e-01 1.43438947e+00 -2.66349196e-01 1.02175415e-01 -7.08237350e-01 -2.66406715e-01 5.84809482e-02 1.24606371e+00 -8.59539628e-01 -6.90026224e-01 -1.62123770e-01 1.69091511e+00 1.28279120e-01 7.43158758e-01 -8.93073916e-01 -5.69634497e-01 2.74748266e-01 -6.43606335e-02 2.87225187e-01 -1.81724697e-01 -2.82315046e-01 -1.37643051e+00 -2.29328256e-02 -8.27858746e-01 -3.11820256e-03 -1.23693967e+00 -1.17594504e+00 1.01964784e+00 -2.08742186e-01 -1.11620641e+00 -4.76060271e-01 -2.61237442e-01 -6.68346524e-01 1.21327317e+00 -1.33468640e+00 -1.30626988e+00 -5.28660476e-01 7.50176609e-01 9.77795601e-01 -3.49304490e-02 5.44446647e-01 -1.49622306e-01 -1.25122085e-01 3.66857678e-01 9.70593318e-02 8.91653076e-02 6.32426202e-01 -1.43668306e+00 1.17325306e+00 1.04516327e+00 3.93573523e-01 3.07375133e-01 5.19939601e-01 -1.05613518e+00 -9.35548604e-01 -1.48565364e+00 9.08861876e-01 -4.10458058e-01 4.67946798e-01 -7.53952324e-01 -7.73114264e-01 6.27332389e-01 7.41444111e-01 -5.95491052e-01 1.28228948e-01 -6.09986007e-01 -7.04588294e-02 3.01655561e-01 -8.14390242e-01 1.16493285e+00 1.24990749e+00 -4.71084386e-01 -3.97691756e-01 5.57495952e-01 9.75809395e-01 -5.49370110e-01 -4.17914033e-01 -2.63223294e-02 7.58850351e-02 -9.40345943e-01 7.72767782e-01 -9.40450951e-02 9.11720395e-01 -2.29668021e-01 2.85956323e-01 -1.53892541e+00 -2.85723835e-01 -9.39118087e-01 8.75225365e-02 1.34707344e+00 6.94984913e-01 -2.60577768e-01 8.26265275e-01 2.68512040e-01 -2.96716802e-02 -3.69478792e-01 -2.80259848e-01 -4.51224774e-01 2.74712116e-01 -7.68421888e-02 6.51108146e-01 7.81093121e-01 -2.44396657e-01 5.34703910e-01 -6.79852784e-01 -2.64823586e-01 5.87240458e-01 1.63908467e-01 8.34194422e-01 -8.39900315e-01 -1.92134306e-01 -5.42589605e-01 1.94646955e-01 -1.72772837e+00 -7.71679580e-02 -1.02803278e+00 3.02482575e-01 -2.25882721e+00 1.41513124e-01 -2.04837963e-01 2.90227830e-01 1.21652186e-01 -4.05383915e-01 4.01438385e-01 2.73083210e-01 3.04867238e-01 -2.87182987e-01 9.70020413e-01 1.97773433e+00 -3.37381899e-01 -3.18714827e-01 -2.18820825e-01 -8.60696733e-01 4.28661495e-01 6.11205220e-01 -1.46627411e-01 -7.61736095e-01 -9.25767422e-01 3.44571620e-01 2.07334280e-01 2.88006395e-01 -8.50426614e-01 2.97877669e-01 -2.12765291e-01 6.34080470e-01 -7.71111012e-01 3.44606340e-01 -2.52356023e-01 3.01710274e-02 2.59523720e-01 -7.39960849e-01 3.50956887e-01 -9.31967646e-02 6.07425034e-01 -2.14030743e-01 2.77485196e-02 5.58438599e-01 -2.89176971e-01 -3.99842203e-01 4.43526894e-01 -5.92363536e-01 1.50674388e-01 7.72978008e-01 -1.07808948e-01 -3.19198549e-01 -7.04022408e-01 -7.50243425e-01 7.66050741e-02 5.35294294e-01 7.29944527e-01 1.08325434e+00 -1.52199972e+00 -9.97413158e-01 2.79865205e-01 -3.98872979e-02 1.66146219e-01 1.26938060e-01 3.33244652e-01 -5.28647065e-01 2.97700465e-01 7.74428844e-02 -4.75983739e-01 -9.11492705e-01 5.09483457e-01 1.39676198e-01 -3.98146659e-01 -7.63120830e-01 9.99697387e-01 8.84839416e-01 -1.84956938e-01 5.24885319e-02 -3.68994117e-01 -7.30986595e-02 7.71751925e-02 7.73398399e-01 -3.36240493e-02 -2.82027811e-01 -4.10272747e-01 3.05820167e-01 5.97783208e-01 -3.02945912e-01 -7.47068048e-01 1.05272913e+00 -3.36979121e-01 3.16980295e-04 3.50946009e-01 7.57754743e-01 6.36940151e-02 -1.45403826e+00 -1.07870162e-01 -2.91256815e-01 -3.94302428e-01 -1.88895240e-02 -7.48735189e-01 -1.11264384e+00 1.11473918e+00 2.91220546e-01 4.96466190e-01 1.05356753e+00 3.40869986e-02 7.66255140e-01 5.70803182e-03 5.30081391e-02 -9.53888655e-01 6.98445499e-01 4.48993564e-01 1.35916126e+00 -1.02233696e+00 -2.56313801e-01 -2.26328671e-01 -9.27935660e-01 9.62695718e-01 5.71492612e-01 -1.08542278e-01 2.85511255e-01 2.49329507e-01 9.42370445e-02 -6.83193281e-02 -9.58972335e-01 -1.82194665e-01 4.76899773e-01 7.38264561e-01 6.58680797e-01 -1.66015238e-01 -3.00349623e-01 -1.29347788e-02 -3.79572839e-01 -7.98706785e-02 5.08193851e-01 6.35891855e-01 -4.24218774e-01 -1.17746758e+00 -2.94421911e-01 1.95923641e-01 -1.92106277e-01 -4.48156863e-01 -8.66871357e-01 5.16491175e-01 -1.21625185e-01 8.85612071e-01 9.74734724e-02 -3.04364920e-01 1.46979913e-01 -8.39630440e-02 2.59172499e-01 -7.97935545e-01 -3.37529212e-01 2.94053704e-01 -2.89024681e-01 -1.60346806e-01 -2.48303726e-01 -3.97375971e-01 -1.11535192e+00 -3.54635239e-01 -1.30618542e-01 1.84864342e-01 5.39550304e-01 5.91670573e-01 3.28896463e-01 7.51203597e-01 6.30655646e-01 -8.85218203e-01 3.28838937e-02 -9.21170354e-01 -4.23535407e-01 3.69622380e-01 2.24181160e-01 -5.42950034e-02 -2.86885232e-01 6.00088239e-01]
[11.31615924835205, -0.07377715408802032]
1c45e4d2-9c66-4408-ad91-0a77b6b551eb
from-behavioral-theories-to-econometrics
2112.15151
null
https://arxiv.org/abs/2112.15151v1
https://arxiv.org/pdf/2112.15151v1.pdf
From Behavioral Theories to Econometrics: Inferring Preferences of Human Agents from Data on Repeated Interactions
We consider the problem of estimating preferences of human agents from data of strategic systems where the agents repeatedly interact. Recently, it was demonstrated that a new estimation method called "quantal regret" produces more accurate estimates for human agents than the classic approach that assumes that agents are rational and reach a Nash equilibrium; however, this method has not been compared to methods that take into account behavioral aspects of human play. In this paper we leverage equilibrium concepts from behavioral economics for this purpose and ask how well they perform compared to the quantal regret and Nash equilibrium methods. We develop four estimation methods based on established behavioral equilibrium models to infer the utilities of human agents from observed data of normal-form games. The equilibrium models we study are quantal-response equilibrium, action-sampling equilibrium, payoff-sampling equilibrium, and impulse-balance equilibrium. We show that in some of these concepts the inference is achieved analytically via closed formulas, while in the others the inference is achieved only algorithmically. We use experimental data of 2x2 games to evaluate the estimation success of these behavioral equilibrium methods. The results show that the estimates they produce are more accurate than the estimates of the Nash equilibrium. The comparison with the quantal-regret method shows that the behavioral methods have better hit rates, but the quantal-regret method performs better in terms of the overall mean squared error, and we discuss the differences between the methods.
['Gali Noti']
2021-12-30
null
null
null
null
['econometrics']
['miscellaneous']
[-5.44061661e-01 1.06275424e-01 -3.01017076e-01 -1.83443606e-01 -4.89194185e-01 -7.26818442e-01 3.47553819e-01 -2.07740054e-01 -8.94801676e-01 1.10904360e+00 4.96780388e-02 -5.51990449e-01 -6.50820076e-01 -5.49582183e-01 -2.85514891e-01 -4.41631496e-01 -2.49422804e-01 6.45563126e-01 -3.77107388e-03 -2.78031528e-01 3.59447867e-01 -1.03756659e-01 -1.20374644e+00 -1.21150672e-01 8.70377243e-01 1.02265942e+00 4.27889973e-02 7.42499828e-01 1.86849028e-01 1.08459830e+00 -4.20211166e-01 -7.32786179e-01 8.30814540e-01 -5.21053791e-01 -7.24070787e-01 -3.42094064e-01 -6.62498355e-01 -8.09083164e-01 -1.33511975e-01 1.27456713e+00 4.01002228e-01 1.91013724e-01 7.05652654e-01 -1.44073188e+00 -3.86837929e-01 7.34288990e-01 -5.02548695e-01 2.52312720e-01 5.35593271e-01 2.43091807e-01 1.34157288e+00 -7.36948848e-02 4.46926296e-01 1.32442188e+00 7.75521398e-01 5.53128958e-01 -1.34154189e+00 -5.68944395e-01 5.29450178e-02 3.81472223e-02 -1.08550680e+00 -2.55712777e-01 2.54886776e-01 -4.11270589e-01 6.98430300e-01 3.58284771e-01 7.93611884e-01 7.26634204e-01 3.07288468e-01 8.81920338e-01 1.45171249e+00 -1.77261055e-01 6.04937434e-01 1.56142727e-01 1.34673804e-01 3.26626599e-01 4.15349275e-01 7.74558783e-01 -9.42887291e-02 -5.05434871e-01 8.82333875e-01 -2.06844248e-02 -6.39426038e-02 -3.00700188e-01 -5.23061037e-01 1.06670594e+00 1.25408694e-01 1.01620043e-02 -9.57793772e-01 4.26721931e-01 2.42472008e-01 6.94949210e-01 6.59873545e-01 7.27061987e-01 -4.18360740e-01 -5.00765562e-01 -6.94397748e-01 7.44014442e-01 1.19179630e+00 5.53179681e-01 5.03411531e-01 -8.52258801e-02 -1.78740397e-01 5.30644178e-01 2.13545278e-01 3.63433629e-01 4.00760889e-01 -1.61133313e+00 2.52560884e-01 3.40240091e-01 1.01092100e+00 -8.03248942e-01 -5.62038243e-01 -3.45274836e-01 -4.02490854e-01 4.67380732e-01 8.55910003e-01 -6.61068976e-01 -2.48385698e-01 2.06920481e+00 4.09533866e-02 -2.17111960e-01 1.72467917e-01 8.69305015e-01 -2.65363418e-02 5.57220459e-01 -1.38739556e-01 -7.46635556e-01 1.02690387e+00 -4.86903936e-01 -7.72380650e-01 -6.18649833e-02 3.95527720e-01 -5.04185915e-01 7.45438874e-01 3.86810869e-01 -1.46676433e+00 1.46095380e-01 -5.23776352e-01 6.68816209e-01 4.39923443e-02 -4.58081216e-01 9.30195451e-01 1.02331722e+00 -1.16383636e+00 6.65473163e-01 -6.25483632e-01 -1.25959590e-01 -2.53714174e-02 5.59558272e-01 2.79375821e-01 6.09584749e-01 -1.12151647e+00 9.80487883e-01 1.00710616e-01 -1.75266489e-01 -7.99865484e-01 -6.24494970e-01 -4.11743611e-01 4.21545237e-01 9.38093066e-01 -6.23725295e-01 1.99823892e+00 -1.33720839e+00 -1.96779847e+00 4.05601025e-01 1.87064722e-01 -6.57824337e-01 9.68494594e-01 1.92921266e-01 3.28510970e-01 -2.43691117e-01 1.30357847e-01 2.15364665e-01 6.18909933e-02 -9.30421352e-01 -6.68934643e-01 -1.86584681e-01 7.06084371e-01 4.26440537e-01 2.62116909e-01 1.91483438e-01 3.74715418e-01 -1.07097678e-01 -3.30428571e-01 -1.01719022e+00 -6.07904375e-01 -4.77045745e-01 -2.59590179e-01 -2.71279424e-01 -3.18394244e-01 -2.78452545e-01 1.13271701e+00 -1.71363676e+00 -1.00497395e-01 3.56636763e-01 3.20427179e-01 -3.07986438e-01 -2.67946115e-03 5.86887360e-01 1.10658161e-01 2.67725974e-01 2.98003882e-01 -1.58414200e-01 6.12861991e-01 -1.05571337e-01 -1.50810465e-01 5.17664909e-01 -6.95285320e-01 8.85981858e-01 -7.55577207e-01 -3.54042426e-02 2.78324857e-02 -5.64911246e-01 -1.07107985e+00 2.33692810e-01 -7.96043351e-02 -1.90701768e-01 -5.57056963e-01 7.62081221e-02 4.23485070e-01 -1.72598377e-01 7.31670201e-01 4.05659914e-01 -3.75969231e-01 3.58794004e-01 -1.24846935e+00 6.96942151e-01 -2.55250156e-01 2.67762750e-01 1.77695841e-01 -9.46320117e-01 1.69203058e-01 3.96048546e-01 6.48155212e-01 -6.00135505e-01 5.59146464e-01 3.46868485e-01 3.53439569e-01 -2.62562156e-01 4.67586637e-01 -5.44056654e-01 -4.34813857e-01 9.74504232e-01 -1.63511902e-01 -1.95807163e-02 1.64139614e-01 1.69964463e-01 9.71359193e-01 -1.81681231e-01 8.56923282e-01 -6.58095479e-01 2.12897621e-02 6.42202348e-02 5.50860107e-01 1.33310902e+00 -5.11333227e-01 -2.29834720e-01 1.14508486e+00 -2.57401913e-01 -9.91502643e-01 -1.14739358e+00 4.28764075e-01 1.07992399e+00 3.61700356e-01 -2.11190730e-01 -1.07584202e+00 -1.54997960e-01 2.24128440e-01 8.36054385e-01 -7.38743305e-01 -9.64492708e-02 1.09141491e-01 -1.00640523e+00 1.25223294e-01 3.64587575e-01 4.79132175e-01 -6.91766143e-01 -8.63824844e-01 1.67166427e-01 -3.87155503e-01 -6.57289445e-01 -5.99307239e-01 -2.83498287e-01 -4.51710671e-01 -1.10703719e+00 -5.37771523e-01 2.23906279e-01 3.19856703e-01 2.20169902e-01 9.53524470e-01 2.09228713e-02 2.23615840e-01 6.99342251e-01 -1.29645273e-01 -5.64017117e-01 -4.06194508e-01 -2.17197239e-01 5.80767930e-01 -8.22526887e-02 5.76525688e-01 -3.40702415e-01 -7.62018144e-01 3.83779287e-01 -4.00777102e-01 -4.25054617e-02 3.47315818e-01 7.43117809e-01 -1.61911532e-01 -1.64649114e-02 7.71893144e-01 -8.48121166e-01 1.25392735e+00 -6.99243546e-01 -1.15324914e+00 1.35962963e-01 -8.20800006e-01 1.11986786e-01 6.79689586e-01 -4.36177254e-01 -1.10691392e+00 -1.98910415e-01 1.91139340e-01 1.86303854e-01 4.07707453e-01 6.07595861e-01 2.86042631e-01 2.33869508e-01 5.92832565e-01 -1.08908057e-01 4.70223308e-01 -2.77849287e-01 2.15937085e-02 8.01349163e-01 3.07256859e-02 -8.11482906e-01 4.20120388e-01 1.57532424e-01 -1.54841140e-01 -3.28637242e-01 -7.14717090e-01 -1.57126755e-01 2.56481111e-01 -5.17730057e-01 6.26215696e-01 -6.04526103e-01 -1.84650922e+00 5.00208795e-01 -9.07338560e-01 -6.59188509e-01 -4.15751219e-01 8.11414659e-01 -1.33760929e+00 1.37514830e-01 -7.60156572e-01 -1.87738216e+00 2.57523894e-01 -1.16273189e+00 3.21420521e-01 4.64500785e-01 -3.06289345e-01 -1.01735902e+00 1.72563046e-01 1.22207440e-01 4.66837287e-01 -2.02810511e-01 5.35578609e-01 -6.76775038e-01 -6.40216887e-01 -1.32433727e-01 3.88002954e-02 -1.75688192e-01 -1.17190771e-01 -1.85047820e-01 -4.77139950e-01 -3.72975260e-01 1.57474115e-01 -2.94415861e-01 2.87391514e-01 1.08327198e+00 5.38202286e-01 -6.90246284e-01 -9.73631665e-02 -5.03435060e-02 1.51850951e+00 6.26251101e-01 5.74634969e-01 2.58116424e-01 -3.88516635e-01 7.34131157e-01 3.96215320e-01 7.81061649e-01 5.49104750e-01 7.18239009e-01 4.11638081e-01 4.28971708e-01 9.03034985e-01 -2.18060449e-01 6.47698343e-01 2.09778562e-01 -4.94321764e-01 -3.27488422e-01 -4.36697513e-01 3.71726424e-01 -2.37629271e+00 -1.48585546e+00 1.41811758e-01 2.54358912e+00 8.23575497e-01 2.60681272e-01 1.02245665e+00 -2.48929083e-01 6.27341032e-01 -3.29273283e-01 -5.67983806e-01 -6.77359581e-01 3.33412439e-01 -1.95748493e-01 8.69569778e-01 9.97755826e-01 -4.99101967e-01 6.88206494e-01 8.03014755e+00 7.07870960e-01 -3.37024719e-01 2.78071612e-01 8.39496970e-01 -4.73871469e-01 -1.63241357e-01 5.17293811e-02 -4.41969186e-01 5.33802509e-01 1.00960183e+00 -8.72927427e-01 9.87085402e-01 7.91581869e-01 6.58535719e-01 -5.69363952e-01 -1.06777179e+00 7.36025751e-01 -6.37308776e-01 -1.15701151e+00 -7.20408559e-01 5.37402034e-01 7.14405417e-01 -2.45370075e-01 3.89449485e-03 5.84328294e-01 1.32875597e+00 -9.36864913e-01 9.39046741e-01 8.38990211e-01 2.27472782e-01 -7.95636356e-01 9.64172244e-01 8.34069669e-01 -8.00700665e-01 -5.96833348e-01 -5.18857598e-01 -1.01436722e+00 1.79393157e-01 3.28402489e-01 -5.04130900e-01 3.35857242e-01 3.33040774e-01 2.57278886e-02 1.18568830e-01 1.03502011e+00 2.45538414e-01 4.64078486e-01 -4.61505830e-01 -5.79183221e-01 4.77398843e-01 -6.49836659e-01 3.45976323e-01 3.91618669e-01 3.44845504e-01 4.77428615e-01 1.63721889e-01 1.20012772e+00 1.51716903e-01 9.92292240e-02 -4.68850732e-01 -7.08651692e-02 3.83198798e-01 8.31171215e-01 -6.14669740e-01 -1.67103767e-01 -2.94413596e-01 5.55316985e-01 1.77154928e-01 2.14515820e-01 -7.89408386e-01 5.40336668e-02 1.04474604e+00 -1.34323552e-01 -1.90247387e-01 4.81618568e-02 -1.34958044e-01 -1.09772599e+00 -2.79133439e-01 -7.52849698e-01 3.85070086e-01 -6.30914569e-01 -1.42112160e+00 -5.05760536e-02 3.87226760e-01 -8.12488079e-01 -1.02068090e+00 -7.71936953e-01 -6.15133047e-01 7.16391325e-01 -7.45262980e-01 -1.16485327e-01 3.93230021e-01 1.30043000e-01 3.69417727e-01 -1.01736009e-01 4.32019502e-01 7.71480321e-04 -4.67821538e-01 3.25639129e-01 5.70942223e-01 -1.36321634e-01 -1.03886304e-02 -1.58822691e+00 -7.38385245e-02 5.33469200e-01 -3.55716288e-01 4.98681188e-01 1.05509055e+00 -4.59050626e-01 -1.07906508e+00 -2.02629104e-01 3.84741992e-01 -3.44790101e-01 7.83005595e-01 -1.56863630e-01 -5.41989654e-02 8.14400196e-01 3.01149786e-01 -6.09561384e-01 5.87964118e-01 3.46929371e-01 2.19942495e-01 8.57985020e-03 -1.23032904e+00 1.04424322e+00 9.12837863e-01 -2.33112648e-01 -4.40539718e-01 2.60938883e-01 3.41331929e-01 -6.02816045e-02 -3.62425089e-01 -9.39151049e-02 9.49201584e-01 -1.49692261e+00 7.73495734e-01 -9.13779020e-01 1.53422669e-01 1.67842120e-01 -1.54624373e-01 -1.67119849e+00 -4.43538845e-01 -1.00964022e+00 4.76144046e-01 6.05768919e-01 4.44456995e-01 -1.10026407e+00 7.78044581e-01 1.10910320e+00 4.01713192e-01 -5.01585782e-01 -1.04902804e+00 -1.09901118e+00 3.82622778e-01 -3.13292325e-01 5.77717543e-01 5.71075141e-01 5.79248667e-01 7.39384443e-02 -6.09402895e-01 -3.28175545e-01 8.02388728e-01 1.21322513e-01 6.87920630e-01 -1.17281306e+00 -6.89109027e-01 -8.25632453e-01 -2.46170700e-01 -1.19857657e+00 3.57581317e-01 -4.43007648e-01 6.31682128e-02 -1.17179942e+00 7.21933782e-01 -2.12270439e-01 -8.01937133e-02 1.13781579e-01 -3.06349788e-02 -1.15072518e-03 3.58850837e-01 -5.47238067e-03 -7.85771370e-01 2.85464048e-01 7.92750239e-01 1.72955960e-01 -3.01989198e-01 4.54376996e-01 -1.02313173e+00 8.99045587e-01 7.81744123e-01 -4.27373976e-01 -5.92358768e-01 2.01562241e-01 7.19672203e-01 7.72802413e-01 4.16579396e-01 -4.39114779e-01 2.69188493e-01 -8.51821542e-01 -8.19367841e-02 -1.65582970e-01 1.87207088e-01 -6.78265929e-01 4.72758025e-01 8.95710289e-01 -4.54677850e-01 1.59252197e-01 -1.95934370e-01 5.89290917e-01 3.67518812e-01 -6.90701962e-01 7.10867047e-01 -4.86138076e-01 -1.95169985e-01 4.90774959e-03 -1.09320390e+00 1.85091123e-01 1.10770941e+00 -1.03121273e-01 -2.08708838e-01 -1.44699967e+00 -6.62802994e-01 3.39214027e-01 4.00286794e-01 -2.49401778e-01 1.85794532e-01 -1.17007828e+00 -6.28228962e-01 -3.85444373e-01 -4.99281973e-01 -1.02395093e+00 2.36829147e-01 1.01463675e+00 -2.78234720e-01 3.73757035e-01 -1.84402436e-01 -5.12118675e-02 -8.49424899e-01 4.79872614e-01 9.49988127e-01 -5.81908345e-01 2.29385778e-01 4.69707996e-01 4.30875510e-01 -4.66454715e-01 -1.24072418e-01 -2.35690102e-01 1.28046080e-01 -3.63689139e-02 4.76985663e-01 7.03461766e-01 -7.00609505e-01 -2.39943892e-01 -1.96871877e-01 -1.51586905e-02 1.94824874e-01 -6.63100183e-01 1.20782280e+00 -4.12920654e-01 -1.78867672e-02 4.58968639e-01 5.99139988e-01 -2.77911983e-02 -1.27703309e+00 -4.79230583e-02 1.95026457e-01 -6.38335407e-01 -2.09737018e-01 -7.58729637e-01 -7.74967492e-01 2.20462918e-01 2.85161614e-01 1.09493911e+00 8.38736773e-01 -3.82146984e-01 1.78736195e-01 3.93191248e-01 8.57169390e-01 -1.35302317e+00 -1.50662139e-01 3.44781339e-01 4.93370414e-01 -9.67791021e-01 -5.96622527e-02 -1.85439885e-01 -6.42020643e-01 7.05008745e-01 4.06221122e-01 -3.46368611e-01 7.42274821e-01 2.08320558e-01 -2.64352709e-01 3.72273810e-02 -1.14139903e+00 -4.87904608e-01 -2.20587283e-01 3.45102251e-01 3.32847424e-02 4.99254227e-01 -8.02500546e-01 1.17412984e+00 -4.80914563e-01 2.29821876e-01 1.14135396e+00 4.37181711e-01 -4.63493377e-01 -9.15757060e-01 -2.15800658e-01 7.44692743e-01 -7.60978580e-01 5.67731224e-02 -5.40225208e-01 9.05951142e-01 -5.88083804e-01 1.30945361e+00 1.78965911e-01 -2.93713659e-01 3.01012456e-01 -6.95151985e-02 5.52123249e-01 -3.44965577e-01 -6.33337498e-01 1.19361266e-01 2.58527756e-01 -6.74015343e-01 -4.65697765e-01 -7.39165962e-01 -7.52332628e-01 -1.06018507e+00 -5.15586972e-01 6.23614252e-01 2.12207347e-01 1.05266929e+00 4.52826768e-02 1.36923060e-01 9.21707094e-01 -7.75475979e-01 -1.48889577e+00 -7.48199403e-01 -1.01830482e+00 1.59998924e-01 5.40135205e-02 -8.42826009e-01 -7.04000652e-01 -6.68980718e-01]
[4.256002426147461, 2.945192337036133]
1f1593e9-3850-4b36-9b3d-bccaec02a67c
geometric-processing-for-image-based-3d
2106.14307
null
https://arxiv.org/abs/2106.14307v1
https://arxiv.org/pdf/2106.14307v1.pdf
Geometric Processing for Image-based 3D Object Modeling
Image-based 3D object modeling refers to the process of converting raw optical images to 3D digital representations of the objects. Very often, such models are desired to be dimensionally true, semantically labeled with photorealistic appearance (reality-based modeling). Laser scanning was deemed as the standard (and direct) way to obtaining highly accurate 3D measurements of objects, while one would have to abide the high acquisition cost and its unavailability on some of the platforms. Nowadays the image-based methods backboned by the recently developed advanced dense image matching algorithms and geo-referencing paradigms, are becoming the dominant approaches, due to its high flexibility, availability and low cost. The largely automated geometric processing of images in a 3D object reconstruction workflow, from ordered/unordered raw imagery to textured meshes, is becoming a critical part of the reality-based 3D modeling. This article summarizes the overall geometric processing workflow, with focuses on introducing the state-of-the-art methods of three major components of geometric processing: 1) geo-referencing; 2) Image dense matching 3) texture mapping. Finally, we will draw conclusions and share our outlooks of the topics discussed in this article.
['Xu Huang', 'Rongjun Qin']
2021-06-27
null
null
null
null
['3d-object-reconstruction', 'object-reconstruction']
['computer-vision', 'computer-vision']
[ 4.76404577e-01 -3.92634347e-02 4.47625428e-01 -3.26406121e-01 -4.33372587e-01 -3.31214130e-01 8.21743131e-01 1.88224241e-01 -1.22233495e-01 2.68489033e-01 -3.63506198e-01 -5.53265698e-02 -4.82796848e-01 -1.09765005e+00 -5.83902657e-01 -3.13416690e-01 -2.02778005e-03 1.21203864e+00 3.38320166e-01 -1.40959084e-01 4.42432106e-01 1.40687108e+00 -2.21093702e+00 -1.53884431e-02 6.14205062e-01 1.34641325e+00 3.58821303e-01 5.04661262e-01 -5.85348666e-01 3.12608033e-01 1.98203977e-02 -3.10476393e-01 4.67879295e-01 1.06246269e-03 -6.41007245e-01 4.54273164e-01 6.26661897e-01 -2.68356681e-01 5.98846860e-02 9.22639310e-01 8.22510421e-02 -5.44597693e-02 6.36961102e-01 -8.14841866e-01 -3.53461027e-01 -2.35193625e-01 -5.14890373e-01 -4.59952801e-01 5.78171849e-01 -1.86045527e-01 3.45693976e-01 -1.21509409e+00 9.95676994e-01 1.22219789e+00 6.69946611e-01 6.66123703e-02 -1.23153257e+00 -2.06412062e-01 -2.43682504e-01 2.30983406e-01 -1.69674432e+00 -4.02687848e-01 8.94622922e-01 -8.36206973e-01 6.97235227e-01 5.75330496e-01 8.62186253e-01 3.35111141e-01 1.03685461e-01 -8.07996243e-02 1.39647472e+00 -7.86548793e-01 1.55433878e-01 2.17881814e-01 -1.54538289e-01 3.45086843e-01 1.68297410e-01 3.15206468e-01 -3.83111596e-01 -1.45735607e-01 1.07036626e+00 1.16523415e-01 -2.44464912e-03 -8.13412666e-01 -9.99771416e-01 3.85862082e-01 2.82177508e-01 2.54355133e-01 -6.11319542e-01 2.47874502e-02 3.48592773e-02 1.73803017e-01 8.39395881e-01 1.74944010e-02 -2.46162396e-02 -3.86873595e-02 -8.96947026e-01 3.57491255e-01 4.12269026e-01 1.14702845e+00 1.09811020e+00 5.31766117e-02 4.53057975e-01 5.07778823e-01 5.48642814e-01 6.08066916e-01 -1.64607212e-01 -9.69968438e-01 -9.34513882e-02 9.72606182e-01 2.31911987e-01 -1.20010388e+00 -3.00814658e-01 -1.32719859e-01 -6.69785619e-01 6.18765891e-01 1.43017411e-01 7.09670722e-01 -7.81482577e-01 8.21305990e-01 7.95252740e-01 -7.23392144e-02 -4.25765872e-01 7.86834419e-01 6.36060536e-01 3.67871106e-01 -1.63631156e-01 2.72791591e-02 1.25685418e+00 -1.51185930e-01 -5.36410153e-01 3.39036398e-02 3.08583289e-01 -1.28240347e+00 8.21812570e-01 2.55820364e-01 -1.17258298e+00 -5.05782664e-01 -7.32051611e-01 -1.22606844e-01 -4.09224868e-01 -1.75387129e-01 3.59550476e-01 6.20434761e-01 -9.97964382e-01 7.63015568e-01 -6.30369425e-01 -6.50870085e-01 2.76701301e-01 2.53748953e-01 -5.85182250e-01 -2.26483777e-01 -6.19690001e-01 1.14720714e+00 2.79717743e-01 2.19867155e-01 -3.32225502e-01 -8.04757953e-01 -6.19063079e-01 -3.51745307e-01 9.41108689e-02 -6.99081957e-01 9.29531753e-01 -7.55894721e-01 -1.35833693e+00 1.85048831e+00 -8.80495738e-03 -2.21882254e-01 8.43904197e-01 -1.27092702e-02 -3.48683476e-01 1.95641086e-01 -2.20087156e-01 3.28801870e-01 6.79909945e-01 -1.67699039e+00 -2.90877134e-01 -7.54272699e-01 -8.73191804e-02 1.57240897e-01 2.45068267e-01 8.90789554e-02 -2.26136610e-01 -2.48479590e-01 7.46588111e-01 -7.14797676e-01 -5.99398883e-03 6.13091588e-01 -1.69818342e-01 2.48851240e-01 1.11239195e+00 -6.86128259e-01 7.40751922e-01 -2.03642416e+00 2.57802363e-02 4.35643882e-01 1.99615359e-01 1.37566909e-01 3.17614108e-01 6.85245514e-01 -5.41445389e-02 -1.99865982e-01 -1.59778714e-01 -5.33880651e-01 -3.91731933e-02 -1.81410240e-03 -1.30323961e-01 7.77785003e-01 -1.61116228e-01 6.38513505e-01 -5.63097179e-01 -5.21652520e-01 9.90226984e-01 6.10986352e-01 3.51189524e-02 3.30467016e-01 -2.15469971e-01 7.45041072e-01 -3.80431145e-01 7.43889153e-01 1.13576019e+00 4.33517955e-02 5.98131530e-02 -4.89971370e-01 -7.94350505e-01 -8.91171843e-02 -1.39365089e+00 1.77159619e+00 -4.44954515e-01 1.49935886e-01 3.47252607e-01 -7.23915875e-01 1.42763078e+00 2.97095895e-01 6.76110327e-01 -8.38141918e-01 8.47276077e-02 6.46523416e-01 -6.27625525e-01 -3.69861543e-01 5.81235170e-01 -3.15693170e-01 2.60346830e-01 3.57128918e-01 -4.71749663e-01 -9.20428395e-01 -3.37696701e-01 -2.23951578e-01 3.82375360e-01 6.02223039e-01 3.31021786e-01 -5.18025637e-01 5.77163279e-01 5.61512709e-01 -5.81034608e-02 2.13456467e-01 3.34590852e-01 9.18457091e-01 -1.56441420e-01 -7.71184564e-01 -1.59973133e+00 -1.07072222e+00 -6.03295803e-01 -1.82388863e-03 2.54485518e-01 -7.77234286e-02 -6.54131770e-01 2.07770780e-01 1.69695809e-01 5.09212017e-01 -5.84367514e-01 2.49218449e-01 -4.37523544e-01 -2.72725463e-01 -6.84508011e-02 -1.50405206e-02 4.65123802e-01 -6.07715786e-01 -8.53604972e-01 9.71224830e-02 3.00325453e-01 -9.96808231e-01 1.89894661e-01 -4.16237712e-01 -1.14655411e+00 -1.16259837e+00 -7.01546848e-01 -4.02138293e-01 7.55442262e-01 4.89831656e-01 1.25111675e+00 3.22305441e-01 -5.04352033e-01 6.46669209e-01 -4.14325565e-01 -4.36586678e-01 -4.58624959e-01 -3.18145633e-01 7.13285208e-02 1.58084020e-01 2.72233903e-01 -8.31134260e-01 -6.41584635e-01 5.45558035e-01 -1.13591421e+00 5.45721948e-01 3.64078104e-01 2.03880653e-01 1.12997973e+00 -2.03814924e-01 -2.31579781e-01 -8.39393497e-01 -7.31462836e-02 -1.23972252e-01 -9.62789297e-01 2.85876215e-01 -4.21638012e-01 -4.56457496e-01 1.83516741e-01 3.45838368e-02 -9.84614551e-01 2.56518185e-01 -1.78667337e-01 -6.91131353e-01 -4.84551817e-01 3.07731509e-01 -8.82453769e-02 -5.57237446e-01 5.81935346e-01 1.81651741e-01 3.70256156e-01 -9.20259297e-01 2.56632209e-01 7.32936859e-01 4.58354890e-01 -5.00652373e-01 1.15995014e+00 1.08348739e+00 4.06712383e-01 -1.08220661e+00 -4.27325755e-01 -5.06202340e-01 -1.21268594e+00 -7.23854482e-01 8.61628115e-01 -5.70874035e-01 -3.95760715e-01 6.39154375e-01 -1.27880228e+00 -2.02509955e-01 -5.00724435e-01 4.52654690e-01 -7.25307226e-01 3.73312175e-01 -3.42100598e-02 -9.78799343e-01 -1.46965057e-01 -1.14143026e+00 1.29934311e+00 -1.50059257e-03 -2.27635369e-01 -8.51581872e-01 2.20181998e-02 4.39874381e-01 6.05821431e-01 7.83259630e-01 1.07433963e+00 2.22267836e-01 -8.47963512e-01 -4.06900525e-01 -4.28739488e-01 -1.76461451e-02 1.55956954e-01 3.34940821e-01 -1.11849117e+00 1.76344723e-01 -2.95716599e-02 -8.74447357e-03 -1.26204953e-01 2.97949165e-01 8.62405360e-01 4.34421182e-01 -1.72576621e-01 5.24306953e-01 1.82488871e+00 1.57776058e-01 9.94178355e-01 4.43271935e-01 5.50957918e-01 1.02450502e+00 8.29007566e-01 4.30386275e-01 1.88431621e-01 1.26622868e+00 7.77027607e-01 -1.11699544e-01 -2.75037825e-01 -2.82388985e-01 -4.06856954e-01 8.07853043e-01 -5.88576078e-01 2.96378404e-01 -1.05858314e+00 4.20648456e-02 -1.55616605e+00 -7.21569955e-01 -9.69787359e-01 2.76205492e+00 1.42527178e-01 -7.98657760e-02 -1.87508777e-01 3.09963852e-01 8.00808191e-01 -2.38974094e-01 -2.90414244e-01 -4.12631273e-01 -1.68814346e-01 2.87749350e-01 3.83613199e-01 5.07695317e-01 -7.84105778e-01 6.95766985e-01 5.82152939e+00 6.36232197e-01 -1.05361235e+00 1.37359932e-01 4.15618777e-01 2.80589759e-01 -3.95881146e-01 1.73871368e-02 -5.86477876e-01 1.40785068e-01 7.52184749e-01 -1.54135108e-01 2.86451697e-01 8.06842685e-01 5.69546342e-01 -4.58789110e-01 -8.08348477e-01 1.24420905e+00 -7.90023282e-02 -1.50213552e+00 1.97583601e-01 1.97293013e-01 5.49319804e-01 -9.28812698e-02 -3.35705608e-01 -4.46989000e-01 -2.64578044e-01 -9.58723187e-01 1.34475315e+00 9.76209104e-01 1.28355646e+00 -4.03772265e-01 3.75287026e-01 2.33726591e-01 -1.34397399e+00 6.13211930e-01 -5.09297311e-01 -1.22553237e-01 6.93586111e-01 1.04345512e+00 -2.29375213e-01 9.34964061e-01 8.34171295e-01 3.98870677e-01 -3.27590436e-01 1.18035913e+00 3.23881626e-01 -2.05863968e-01 -4.69943076e-01 4.13981587e-01 -5.74246235e-02 -7.93383777e-01 4.82512802e-01 6.94514275e-01 5.56054115e-01 1.37489542e-01 -1.60444956e-02 9.55299199e-01 3.18645149e-01 2.21209720e-01 -8.59279454e-01 1.87007338e-01 4.58346128e-01 1.15737414e+00 -9.27516699e-01 -1.67579949e-01 -3.23312879e-01 7.49723375e-01 1.71496421e-02 -2.07777575e-01 -4.63426352e-01 -2.59161461e-02 5.63744783e-01 8.75820398e-01 -2.01898441e-01 -6.40249670e-01 -4.51815397e-01 -6.67022824e-01 1.20767392e-01 -3.34514230e-01 -5.83540425e-02 -1.19788170e+00 -1.09201872e+00 5.66885769e-01 3.13585401e-01 -1.56845713e+00 4.88633700e-02 -6.59281492e-01 -1.11607231e-01 1.13899243e+00 -1.40700066e+00 -1.30032802e+00 -7.55959213e-01 4.44692254e-01 3.44543219e-01 4.02866542e-01 1.11449647e+00 5.31205416e-01 1.97730120e-02 -4.47004378e-01 2.33925700e-01 -5.20760596e-01 3.41121048e-01 -7.74155319e-01 2.65445858e-01 4.31316733e-01 -1.36904120e-01 3.16362560e-01 5.90475738e-01 -7.00829983e-01 -1.73491859e+00 -8.88512790e-01 8.80835652e-01 -2.68331885e-01 3.57616156e-01 -3.49595100e-01 -8.97928655e-01 5.49121797e-01 -3.63713324e-01 -2.08626688e-02 3.29395711e-01 -4.19737458e-01 7.14963377e-02 -9.13846865e-02 -1.42258418e+00 3.55963111e-01 1.24595046e+00 -4.70324069e-01 -5.47415078e-01 4.79450613e-01 7.54419938e-02 -7.93591201e-01 -1.13555253e+00 3.87559146e-01 7.39000916e-01 -1.38732839e+00 1.20910585e+00 -7.18914792e-02 2.17265889e-01 -4.74594742e-01 -2.74680883e-01 -7.51134813e-01 -2.17008665e-01 -2.29957953e-01 3.74230117e-01 1.07642019e+00 -1.15590058e-01 -5.57090104e-01 8.16245496e-01 9.97149646e-01 -3.47468942e-01 -4.20720518e-01 -1.13219595e+00 -7.86202550e-01 -2.77775496e-01 -6.21957600e-01 8.11603367e-01 8.81091774e-01 -5.53066790e-01 -4.54172105e-01 -1.86699256e-01 1.33940727e-01 8.41210485e-01 4.45909798e-01 9.21822965e-01 -1.86007035e+00 1.02622956e-01 -2.62385756e-01 -9.41199064e-01 -6.11640215e-01 -3.14378768e-01 -7.26953447e-01 -3.91049296e-01 -1.55633569e+00 -2.26624191e-01 -9.34138298e-01 5.38728535e-01 -3.32150728e-01 6.95365727e-01 6.84066713e-01 -3.61033455e-02 6.49099112e-01 -8.74431897e-03 3.26001644e-01 1.15554440e+00 1.95542902e-01 6.18770272e-02 -1.35907009e-01 3.60888848e-03 6.59276307e-01 4.34000254e-01 -2.54385501e-01 -1.75413489e-01 -7.50187039e-01 3.74662578e-01 -1.61502421e-01 4.95146334e-01 -1.00476778e+00 7.13200569e-02 -2.06158146e-01 2.21044794e-01 -9.94914055e-01 8.11220348e-01 -1.47481394e+00 1.37205327e+00 3.22217494e-01 2.36817136e-01 6.20562248e-02 9.04128626e-02 1.53172672e-01 -1.38382941e-01 -3.54999125e-01 1.12746513e+00 -4.88590181e-01 -8.26910973e-01 5.29457748e-01 -5.63611202e-02 -5.36201596e-01 1.28672814e+00 -8.62477243e-01 2.43341327e-01 -3.02843209e-02 -7.72415936e-01 -3.72758269e-01 1.22016418e+00 -6.39369190e-02 7.35526383e-01 -1.38687968e+00 -6.30200326e-01 2.76372254e-01 1.90048113e-01 3.13607454e-01 7.31669188e-01 8.07615638e-01 -1.28025544e+00 4.19532001e-01 -3.73858988e-01 -9.61510301e-01 -1.31969178e+00 4.49728996e-01 5.05875528e-01 1.53202564e-01 -9.36409354e-01 3.38087261e-01 -1.76022816e-02 -5.05208611e-01 -2.64949948e-01 -5.00315763e-02 -5.36374887e-03 -2.18319640e-01 4.62593853e-01 6.89251482e-01 8.10219288e-01 -1.07608140e+00 -3.34062457e-01 1.31555450e+00 4.43342537e-01 -5.89404069e-02 1.36837637e+00 -3.08786333e-01 -3.89395207e-01 4.18572485e-01 8.98719728e-01 -2.50984132e-01 -9.53453541e-01 -1.62927821e-01 -1.31163886e-02 -1.04712892e+00 3.18292886e-01 -3.43783736e-01 -8.81478250e-01 9.10888374e-01 7.02475786e-01 3.58783960e-01 9.18810010e-01 6.82582408e-02 2.72428870e-01 -1.26131296e-01 1.08842790e+00 -8.67805958e-01 -6.58734441e-01 2.69930214e-01 1.19912636e+00 -8.97398233e-01 2.97121435e-01 -1.07026994e+00 1.21686876e-01 1.39341235e+00 3.25908735e-02 -3.90571170e-02 9.60985601e-01 1.97841108e-01 5.45346923e-02 -7.42338240e-01 6.45195246e-02 -1.50255606e-01 2.72571892e-01 7.91319072e-01 3.71811986e-01 -3.39509472e-02 -2.72420555e-01 -3.29003215e-01 -1.19935825e-01 6.82468265e-02 8.02397504e-02 1.03377271e+00 -2.80927956e-01 -1.27447832e+00 -9.77462351e-01 4.48222846e-01 1.63012058e-01 2.58860230e-01 -3.39312404e-01 8.86136293e-01 -2.09152419e-02 5.94120026e-01 3.26193392e-01 -1.82033613e-01 8.47233534e-01 3.25965397e-02 7.41740108e-01 -5.18397808e-01 -2.72211492e-01 -1.37608454e-01 2.87994426e-02 -6.17229402e-01 -7.47053564e-01 -6.85259879e-01 -8.57325375e-01 -6.57921612e-01 -1.83227152e-01 -1.39499381e-01 1.21888447e+00 6.52023017e-01 4.09897029e-01 -1.59844309e-01 6.77073240e-01 -1.58450472e+00 -7.27173984e-02 -5.90589523e-01 -9.02436018e-01 5.87680817e-01 -4.09885436e-01 -1.05726409e+00 -2.02579513e-01 1.41307816e-01]
[8.43966007232666, -2.7125673294067383]
169e0e27-64a6-4fca-b0d7-8ab6690f4f30
art-authentication-with-vision-transformers
2307.03039
null
https://arxiv.org/abs/2307.03039v2
https://arxiv.org/pdf/2307.03039v2.pdf
Art Authentication with Vision Transformers
In recent years, Transformers, initially developed for language, have been successfully applied to visual tasks. Vision Transformers have been shown to push the state-of-the-art in a wide range of tasks, including image classification, object detection, and semantic segmentation. While ample research has shown promising results in art attribution and art authentication tasks using Convolutional Neural Networks, this paper examines if the superiority of Vision Transformers extends to art authentication, improving, thus, the reliability of computer-based authentication of artworks. Using a carefully compiled dataset of authentic paintings by Vincent van Gogh and two contrast datasets, we compare the art authentication performances of Swin Transformers with those of EfficientNet. Using a standard contrast set containing imitations and proxies (works by painters with styles closely related to van Gogh), we find that EfficientNet achieves the best performance overall. With a contrast set that only consists of imitations, we find the Swin Transformer to be superior to EfficientNet by achieving an authentication accuracy of over 85%. These results lead us to conclude that Vision Transformers represent a strong and promising contender in art authentication, particularly in enhancing the computer-based ability to detect artistic imitations.
['Eric Postma', 'Carina Popovici', 'Ludovica Schaerf']
2023-07-06
null
null
null
null
['object-detection']
['computer-vision']
[ 2.04571754e-01 -1.53630599e-01 7.55965859e-02 -9.08828992e-03 -3.29015225e-01 -8.68613005e-01 1.09840202e+00 -9.69351903e-02 -5.41735172e-01 2.53028393e-01 -2.88261294e-01 -2.20992371e-01 -1.04833916e-01 -7.53544629e-01 -4.88783449e-01 -3.95240605e-01 3.47094774e-01 6.77871466e-01 3.70923221e-01 -2.05406740e-01 3.89677256e-01 9.58282650e-01 -1.51283026e+00 1.66171953e-01 3.49780411e-01 1.23366976e+00 -1.36164173e-01 5.68874955e-01 -1.11234538e-01 8.57402623e-01 -9.24392760e-01 -1.34408367e+00 3.22275728e-01 8.28196183e-02 -8.31944764e-01 2.05506027e-01 8.52161288e-01 -2.63970524e-01 -4.21163261e-01 8.74382675e-01 4.78853136e-01 -3.26467276e-01 5.82420647e-01 -1.66696250e+00 -9.28693414e-01 6.83779895e-01 -3.87948781e-01 9.20513719e-02 3.29724133e-01 4.75665808e-01 9.43549156e-01 -7.25174367e-01 6.06816053e-01 1.29451525e+00 9.51011062e-01 4.03630972e-01 -1.03136945e+00 -8.67068052e-01 -3.22725177e-01 3.36162269e-01 -1.41598308e+00 -3.89830202e-01 8.54034781e-01 -4.17181939e-01 5.70750475e-01 1.39250860e-01 7.68286586e-01 1.29716051e+00 -2.82950014e-01 8.94784927e-01 1.21588767e+00 -3.52128893e-01 1.62025422e-01 1.44920468e-01 -4.21355218e-02 7.86695361e-01 2.61944711e-01 -1.67510390e-01 -5.24817228e-01 9.12128985e-02 8.86443615e-01 -2.28942141e-01 -6.67311773e-02 -6.12132736e-02 -1.29839361e+00 7.05144465e-01 5.33894062e-01 5.84916115e-01 -3.73380482e-01 3.68146181e-01 4.35007423e-01 1.79184660e-01 1.09392963e-01 8.33407700e-01 -1.39376268e-01 -2.88825542e-01 -1.04829836e+00 2.01200433e-02 6.62159920e-01 6.42236531e-01 3.34120631e-01 4.65155482e-01 -2.08685666e-01 7.31831968e-01 -1.53610315e-02 7.34679043e-01 2.48214975e-01 -1.20807815e+00 1.50547594e-01 8.84922266e-01 -1.26474410e-01 -9.19964135e-01 -9.52102989e-02 -4.83285397e-01 -7.35542953e-01 7.86931813e-01 9.39634919e-01 3.88217449e-01 -8.57353389e-01 1.29829466e+00 -1.20977953e-01 -4.51166071e-02 -1.91500232e-01 6.10349298e-01 6.84918225e-01 3.73758748e-02 1.24553300e-01 4.80976194e-01 1.75355971e+00 -6.51663661e-01 -2.88490981e-01 -7.66237900e-02 4.34850566e-02 -1.04246080e+00 1.36418831e+00 6.61275923e-01 -1.07434034e+00 -8.33894789e-01 -1.08211327e+00 8.54752436e-02 -6.02941215e-01 4.11869079e-01 6.74906850e-01 1.14998102e+00 -1.12919629e+00 7.46987820e-01 -4.65978533e-01 -5.48693717e-01 9.45631981e-01 5.22045374e-01 -5.93902647e-01 2.21521273e-01 -7.61644840e-01 9.45070267e-01 2.25565001e-01 -9.65104699e-02 -9.96967256e-01 -3.00564796e-01 -5.03711104e-01 -8.16343352e-02 3.29719245e-01 -5.71862996e-01 1.06514645e+00 -1.13443196e+00 -1.33048582e+00 1.16288292e+00 3.36052775e-01 -6.64838910e-01 8.88508439e-01 -1.26366988e-01 -4.79672134e-01 6.35936618e-01 1.75113142e-01 7.78659642e-01 1.12168443e+00 -1.14580297e+00 -3.98492008e-01 -3.11099708e-01 3.46027553e-01 -2.76591331e-01 -7.91191161e-01 1.32620111e-01 -5.09783566e-01 -6.29528999e-01 -1.61877885e-01 -9.94830370e-01 3.66390944e-01 5.52914679e-01 -4.93592560e-01 -3.72556388e-01 9.88856375e-01 -5.55846155e-01 4.87081498e-01 -2.06857133e+00 -2.69288182e-01 2.60132968e-01 3.40084910e-01 6.34084761e-01 -2.22594664e-01 9.48866680e-02 2.63652168e-02 3.25458616e-01 -2.20030360e-02 -5.14871538e-01 1.63644671e-01 2.27718607e-01 -2.47740939e-01 4.69146222e-01 2.89112151e-01 1.15897489e+00 -6.41349971e-01 -6.60774410e-01 6.37646258e-01 6.51747763e-01 2.20215037e-01 -1.35502398e-01 -8.62268507e-02 1.72382236e-01 -3.37790310e-01 9.70198810e-01 3.47276032e-01 -3.66626948e-01 -2.74718076e-01 -2.22928926e-01 1.35650977e-01 -3.14445078e-01 -9.64219809e-01 1.35978723e+00 -1.45621404e-01 1.15759075e+00 -9.33991298e-02 -3.37592006e-01 1.08643007e+00 2.19899684e-01 3.22252512e-01 -6.29326463e-01 2.51234531e-01 7.77625665e-02 -1.80324376e-01 -3.13260913e-01 7.17315376e-01 9.86508206e-02 3.14531416e-01 3.45103741e-01 -4.46275733e-02 -4.32048261e-01 1.67293265e-01 3.55766207e-01 1.06588817e+00 1.42132699e-01 -1.62117675e-01 1.36078358e-01 7.57160544e-01 -1.29677966e-01 -1.56391963e-01 8.79317760e-01 -3.29158753e-01 6.48532331e-01 3.22072715e-01 -2.53473192e-01 -1.33589387e+00 -1.44277608e+00 -9.65135731e-03 8.35384130e-01 3.52083743e-02 -2.10836425e-01 -8.19319785e-01 -5.96417248e-01 -1.10712983e-01 5.39142251e-01 -6.61829889e-01 5.31802811e-02 -3.02646726e-01 -2.34356910e-01 1.45585263e+00 6.82453275e-01 1.06761205e+00 -1.31970787e+00 -5.93235135e-01 -1.46198258e-01 -6.69641420e-02 -1.55185020e+00 -5.99149093e-02 -3.81214112e-01 -6.30930066e-01 -1.38896298e+00 -9.79227722e-01 -6.75119519e-01 2.11798653e-01 1.00687623e-01 1.23587883e+00 4.63214427e-01 -3.84468704e-01 7.88879395e-01 -2.53683895e-01 -4.47612464e-01 -5.59046328e-01 1.65549159e-01 -2.46463735e-02 1.05027683e-01 1.83566004e-01 -4.09674287e-01 -4.78363156e-01 4.83879775e-01 -9.09372568e-01 -4.10978705e-01 6.05170369e-01 4.27294046e-01 4.75633554e-02 -1.37258857e-03 1.54189572e-01 -3.51325542e-01 3.90855968e-01 2.53679335e-01 -4.69312310e-01 2.59748518e-01 -6.08143747e-01 -9.25622806e-02 4.72250998e-01 -6.29951417e-01 -6.94960773e-01 1.23170555e-01 -2.57732660e-01 -6.74110472e-01 -3.30753982e-01 -2.69017935e-01 -1.04484051e-01 -5.85812807e-01 5.33141732e-01 2.46260658e-01 7.97096714e-02 -4.24934447e-01 4.27058518e-01 6.31569624e-01 1.04871309e+00 -5.61242998e-01 1.08369064e+00 7.20375657e-01 2.51579434e-01 -1.20300138e+00 -4.28488791e-01 -2.86720514e-01 -6.25746489e-01 -7.04853415e-01 9.92399335e-01 -6.19584322e-01 -1.27700627e+00 9.02776778e-01 -1.10512781e+00 -1.45805106e-01 -5.62004924e-01 2.59756505e-01 -5.31918108e-01 7.09106922e-01 -7.35292792e-01 -1.04798210e+00 -4.64157164e-01 -1.07799244e+00 1.18727052e+00 4.57673930e-02 -2.77777195e-01 -8.36855173e-01 -4.22970444e-01 7.19541550e-01 3.09582442e-01 3.36598277e-01 6.89598143e-01 -5.76758146e-01 -6.18856966e-01 -3.25694054e-01 -4.92652893e-01 4.60675359e-01 -1.81340829e-01 3.07578057e-01 -1.33792973e+00 1.10269353e-01 -3.34498942e-01 -2.80016601e-01 7.49719560e-01 -7.07259923e-02 9.01917815e-01 1.68146744e-01 -9.13334861e-02 4.50350285e-01 1.28922439e+00 -8.13224539e-02 1.09270513e+00 8.12357545e-01 7.61698008e-01 3.81186932e-01 1.85147807e-01 2.68036395e-01 3.24702114e-02 5.32289743e-01 7.78593302e-01 -2.14255646e-01 -4.74216968e-01 -1.25083387e-01 3.34317237e-01 1.33968040e-01 -6.12123072e-01 -2.02050999e-01 -8.87933135e-01 4.53823388e-01 -1.39399552e+00 -9.83501315e-01 -3.73516858e-01 1.95204961e+00 3.02068830e-01 5.05787730e-01 4.58224744e-01 5.94063103e-01 7.67238379e-01 -4.70043067e-03 -3.01782727e-01 -4.15237397e-02 -5.71213961e-01 5.22414565e-01 6.17726386e-01 -3.43981981e-02 -1.05463135e+00 9.77837741e-01 6.87645674e+00 8.76927793e-01 -9.68611538e-01 -1.05412491e-03 5.27725995e-01 3.88975799e-01 1.79572716e-01 -2.51923352e-01 -2.79844433e-01 3.35768402e-01 5.57176352e-01 3.69556397e-01 4.33066308e-01 8.52629244e-01 -1.63787425e-01 3.24244313e-02 -9.86592174e-01 1.11378908e+00 2.72492588e-01 -1.37652540e+00 2.02206358e-01 6.21177778e-02 3.60526443e-01 -5.37010580e-02 3.51371884e-01 -1.05731554e-01 3.92233610e-01 -1.16289294e+00 1.05652392e+00 3.30649436e-01 8.28327835e-01 -5.75013041e-01 8.61745775e-01 -1.48149043e-01 -1.32778347e+00 -6.91419095e-02 -2.44742841e-01 7.19227083e-03 4.13121060e-02 1.67756319e-01 -7.51121044e-01 4.40107793e-01 8.22271466e-01 7.34468520e-01 -1.20206416e+00 1.03907657e+00 -4.37498242e-01 4.53043252e-01 -2.57555902e-01 -6.09054510e-03 2.66766816e-01 3.02452105e-03 4.78010058e-01 9.86372888e-01 1.64522916e-01 -3.90155911e-01 -1.66394874e-01 8.40603650e-01 -2.33370528e-01 -7.18629286e-02 -5.21605134e-01 -3.26249629e-01 2.52173960e-01 1.31765497e+00 -1.19925964e+00 -3.69866163e-01 7.27982167e-03 1.25549972e+00 -1.64668486e-01 6.19440377e-02 -8.06310356e-01 -2.93809563e-01 4.97863233e-01 1.55730605e-01 4.40279067e-01 -3.71603787e-01 -4.66943294e-01 -6.95223451e-01 -1.30742118e-01 -8.01445782e-01 3.07522684e-01 -1.45139766e+00 -1.37762189e+00 8.45219016e-01 -2.46889576e-01 -1.02361941e+00 1.16457082e-01 -1.12766755e+00 -5.26484072e-01 4.78632748e-01 -1.11619937e+00 -1.87411880e+00 -5.25176883e-01 9.63763595e-01 4.43718910e-01 -5.15823662e-01 6.93828881e-01 1.73167706e-01 -1.51290551e-01 6.35365248e-01 -1.25263587e-01 6.91436172e-01 4.49485362e-01 -1.14608097e+00 4.60376531e-01 7.50508249e-01 6.33187056e-01 3.19200993e-01 5.15895784e-01 -5.45075655e-01 -1.31190205e+00 -6.66835845e-01 3.36221993e-01 -9.30053353e-01 8.96022081e-01 -1.09671347e-01 -5.97501576e-01 4.53703254e-01 3.29761803e-01 -5.49796760e-01 2.19106987e-01 -1.81050360e-01 -7.53109694e-01 2.46030875e-02 -1.21585524e+00 6.65693164e-01 9.29120958e-01 -8.30136240e-01 -6.38232708e-01 2.76191086e-01 2.63480157e-01 1.20386146e-01 -9.67725635e-01 8.82058870e-03 9.14599895e-01 -9.26172316e-01 1.45836997e+00 -2.29664981e-01 3.76736075e-01 -3.56072545e-01 -3.71212848e-02 -6.43403411e-01 -1.43972039e-01 -4.80416656e-01 2.01831907e-01 1.54998696e+00 7.15879500e-02 -5.52835107e-01 1.04539394e+00 4.33644116e-01 2.19515651e-01 -8.47163275e-02 -9.45603669e-01 -1.08114552e+00 -1.25429645e-01 -7.40854681e-01 5.66413045e-01 8.26948404e-01 -6.10226035e-01 3.80764872e-01 -2.90125072e-01 -1.49806529e-01 7.68545210e-01 -2.93903410e-01 9.78220284e-01 -1.64078343e+00 -1.30216390e-01 -8.55710626e-01 -9.49733615e-01 -4.16952759e-01 9.73463282e-02 -7.53585577e-01 -4.09162164e-01 -1.49660981e+00 2.80870367e-02 -1.66790232e-01 2.22546011e-02 7.17395484e-01 1.91216603e-01 1.17133093e+00 8.37912202e-01 4.14676428e-01 -5.26516795e-01 1.84548661e-01 1.10761750e+00 -6.05842769e-01 2.88014650e-01 -5.11441343e-02 -4.86788362e-01 9.40651119e-01 7.71300077e-01 -1.33770213e-01 -6.02993369e-02 -1.69946015e-01 4.58582155e-02 -7.04005420e-01 7.90923655e-01 -1.42141926e+00 3.74583974e-02 3.09306800e-01 7.17725158e-01 -4.52867985e-01 9.00374353e-01 -1.08969343e+00 1.88390166e-01 6.97714329e-01 -2.53474623e-01 2.22074464e-01 2.09725529e-01 3.06541890e-01 1.67923883e-01 -2.90901542e-01 6.95178330e-01 -2.44465962e-01 -1.04043412e+00 1.14110619e-01 -3.77552241e-01 -9.77019295e-02 7.17214644e-01 -8.34357977e-01 -5.59038460e-01 -5.02647877e-01 -4.75511611e-01 -2.40936309e-01 8.42622876e-01 4.93516177e-01 5.66029549e-01 -1.09289348e+00 -6.88247025e-01 3.78404520e-02 1.73902050e-01 -6.91944957e-01 -1.85618386e-01 5.42094231e-01 -7.46225417e-01 -2.97809429e-02 -6.62430406e-01 -7.94105411e-01 -1.57984328e+00 5.92726409e-01 6.02977611e-02 2.26771133e-03 -6.78402483e-01 7.20698774e-01 -2.26186007e-01 6.29832223e-02 2.86361337e-01 2.63203401e-02 -1.27162412e-01 3.33612561e-02 4.98479903e-01 6.35464191e-01 -1.33734211e-01 -8.06626499e-01 -5.13423145e-01 7.23604560e-01 3.31023693e-01 -3.47210377e-01 1.12430084e+00 1.94279224e-01 -3.77261043e-02 3.53169382e-01 7.63391495e-01 1.37431845e-01 -9.60274458e-01 2.85921376e-02 2.49372236e-02 -6.20119452e-01 -1.65778548e-01 -1.01132405e+00 -1.32185829e+00 9.86568809e-01 8.73099804e-01 4.70685393e-01 1.07777214e+00 1.44843340e-01 6.40598834e-01 4.26125884e-01 4.18885350e-01 -9.32418942e-01 6.66985750e-01 3.04903209e-01 6.59856260e-01 -1.06529427e+00 -3.51096131e-02 -1.53091341e-01 -6.92149282e-01 1.29169333e+00 3.77812684e-01 -1.90903898e-02 2.22877920e-01 2.40465716e-01 1.12742819e-01 -2.58567989e-01 2.27486953e-01 -5.46739042e-01 1.98956072e-01 9.33766007e-01 -2.95085777e-02 -6.96075149e-03 2.54572600e-01 4.18092459e-02 -3.71725559e-01 1.66523065e-02 4.62914288e-01 5.76289535e-01 -3.11055332e-01 -1.16137016e+00 -8.11383247e-01 3.30503911e-01 -4.64385778e-01 -7.33338669e-02 -1.16964912e+00 1.20359719e+00 2.08819270e-01 1.06484735e+00 1.03633508e-01 -4.91680682e-01 3.08831811e-01 -4.69646715e-02 6.06185913e-01 5.34374639e-02 -1.02223039e+00 -3.17657650e-01 1.94289252e-01 -3.58540177e-01 -6.73932135e-01 -3.55477065e-01 -7.05686569e-01 -7.29668856e-01 -3.11408520e-01 -8.79658610e-02 9.24072385e-01 9.42822278e-01 -1.39279351e-01 4.98447269e-01 1.57317802e-01 -8.07412505e-01 -2.00610355e-01 -8.47281933e-01 -6.93455875e-01 6.75414383e-01 -3.00560109e-02 -6.29670441e-01 -7.94388503e-02 5.83419763e-02]
[11.473394393920898, 0.3791159391403198]
588a92b1-842c-4748-96f3-988b9ce00dbd
multiple-people-tracking-by-lifted-multicut
null
null
http://openaccess.thecvf.com/content_cvpr_2017/html/Tang_Multiple_People_Tracking_CVPR_2017_paper.html
http://openaccess.thecvf.com/content_cvpr_2017/papers/Tang_Multiple_People_Tracking_CVPR_2017_paper.pdf
Multiple People Tracking by Lifted Multicut and Person Re-Identification
Tracking multiple persons in a monocular video of a crowded scene is a challenging task. Humans can master it even if they loose track of a person locally by re-identifying the same person based on their appearance. Care must be taken across long distances, as similar-looking persons need not be identical. In this work, we propose a novel graph-based formulation that links and clusters person hypotheses over time by solving an instance of a minimum cost lifted multicut problem. Our model generalizes previous works by introducing a mechanism for adding long-range attractive connections between nodes in the graph without modifying the original set of feasible solutions. This allows us to reward tracks that assign detections of similar appearance to the same person in a way that does not introduce implausible solutions. To effectively match hypotheses over longer temporal gaps we develop new deep architectures for re-identification of people. They combine holistic representations extracted with deep networks and body pose layout obtained with a state-of-the-art pose estimation model. We demonstrate the effectiveness of our formulation by reporting a new state-of-the-art for the MOT16 benchmark.
['Bernt Schiele', 'Bjoern Andres', 'Siyu Tang', 'Mykhaylo Andriluka']
2017-07-01
null
null
null
cvpr-2017-7
['multiple-people-tracking']
['computer-vision']
[-9.13393423e-02 3.26883569e-02 3.54119807e-01 -3.22797805e-01 -1.95517763e-01 -6.74569190e-01 2.32284650e-01 1.42451465e-01 -7.48290479e-01 6.23919487e-01 1.64537311e-01 3.76191616e-01 -2.39891768e-01 -5.12587547e-01 -9.66006875e-01 -2.65612006e-01 -3.51512879e-01 9.12541091e-01 4.65165615e-01 -1.96498394e-01 -2.19779983e-01 5.91662645e-01 -1.63790321e+00 6.97902590e-02 4.00551796e-01 4.60629433e-01 2.78346092e-02 6.88249171e-01 4.66708094e-01 3.34390789e-01 -7.38183737e-01 -8.28287303e-01 7.04672337e-01 -3.61442894e-01 -7.04662204e-01 3.24752271e-01 1.36252546e+00 -2.50715315e-01 -5.58993995e-01 9.58286941e-01 5.22906005e-01 4.18010473e-01 2.63169795e-01 -1.39482439e+00 -3.86986047e-01 3.42553884e-01 -7.37121463e-01 3.26472402e-01 6.47205710e-01 1.86020449e-01 8.29224706e-01 -3.12275290e-01 9.18196082e-01 1.44966757e+00 1.10954952e+00 8.02730680e-01 -1.16423607e+00 -4.18364882e-01 7.95277536e-01 1.91719294e-01 -1.52056980e+00 -2.30332911e-01 3.82155806e-01 -5.10890841e-01 8.99062455e-01 3.21604013e-01 1.16184580e+00 9.78435397e-01 1.06433213e-01 5.40395498e-01 5.56288779e-01 -1.64378941e-01 -1.64309651e-01 -1.31748557e-01 3.52510065e-02 9.90625978e-01 8.63407791e-01 1.14936560e-01 -7.45498657e-01 -1.23139471e-01 6.36099577e-01 2.66741186e-01 -2.19241053e-01 -1.00898135e+00 -1.31903291e+00 6.08344734e-01 9.28841949e-01 1.80804700e-01 -3.01635802e-01 3.77054453e-01 1.81868568e-01 1.09158538e-01 2.26466656e-01 6.86542273e-01 -2.64299035e-01 3.58549058e-01 -1.04675591e+00 7.42314875e-01 4.88739222e-01 8.88021708e-01 3.92934978e-01 -3.46444994e-01 -4.70417380e-01 2.39575416e-01 1.96914241e-01 4.12749082e-01 -3.22509035e-02 -8.86541724e-01 3.08623314e-01 7.13908792e-01 4.02341455e-01 -1.17940879e+00 -8.41099679e-01 -9.18728590e-01 -6.63454056e-01 2.79354453e-01 7.97279298e-01 -2.78990209e-01 -7.38217175e-01 2.08300877e+00 5.24814248e-01 2.55484343e-01 -3.19417506e-01 1.24042237e+00 6.47202492e-01 1.39191687e-01 -1.14392117e-01 1.26306325e-01 1.37929356e+00 -1.15736568e+00 -4.15116966e-01 -5.93283474e-01 2.54436314e-01 -4.64091480e-01 3.18252623e-01 2.34668225e-01 -1.19656885e+00 -7.14882910e-01 -8.55472863e-01 1.56871155e-01 -3.35496604e-01 1.50666729e-01 5.89465797e-01 6.67079866e-01 -1.25068974e+00 9.80877280e-01 -6.29823387e-01 -8.61047566e-01 3.23302805e-01 6.88491166e-01 -5.15461385e-01 -7.56298825e-02 -9.60938871e-01 9.42994833e-01 1.02340944e-01 4.19801801e-01 -6.58661842e-01 -3.84058326e-01 -8.66985202e-01 -5.10549657e-02 6.43462181e-01 -1.36747122e+00 1.00056541e+00 -1.01244438e+00 -9.84891951e-01 1.17485428e+00 -1.92153335e-01 -6.33235037e-01 9.80205417e-01 -5.74419737e-01 -3.07881653e-01 8.00434723e-02 2.63220400e-01 9.68487740e-01 8.15451741e-01 -1.15499806e+00 -6.77204490e-01 -6.06017411e-01 2.61858135e-01 4.07754689e-01 -4.37406264e-02 2.05154475e-02 -8.00019979e-01 -5.47439575e-01 1.10626686e-02 -1.27506793e+00 -4.58861023e-01 4.74395424e-01 -4.75682616e-01 -1.69264972e-01 3.08981001e-01 -5.75034678e-01 8.72130692e-01 -1.65143573e+00 6.69477642e-01 2.85646886e-01 6.08595610e-01 1.57837987e-01 -1.78254560e-01 2.24490270e-01 1.32951021e-01 -2.62036055e-01 8.85580480e-02 -7.47312784e-01 1.81576777e-02 -4.87723537e-02 3.18845004e-01 8.77838254e-01 7.69792721e-02 9.74817991e-01 -1.05475521e+00 -4.46733564e-01 8.94152522e-02 4.13229585e-01 -5.79910994e-01 6.26600310e-02 1.25549743e-02 4.77194399e-01 -3.91298495e-02 4.04542297e-01 6.05264127e-01 -3.94907653e-01 2.95509785e-01 -1.97556719e-01 -1.21724308e-02 -1.30292892e-01 -1.56387198e+00 1.97079062e+00 2.93318361e-01 4.63625789e-01 5.62547930e-02 -6.30850375e-01 4.72157776e-01 -8.77209473e-04 6.36117041e-01 -2.74815768e-01 1.05580166e-01 -1.91147253e-01 8.05264562e-02 -2.41671264e-01 6.33262515e-01 1.81906715e-01 -9.36932489e-02 2.04198241e-01 3.78321074e-02 5.77865541e-01 3.95615101e-01 2.90674001e-01 1.04374719e+00 3.98679018e-01 7.44530559e-02 -1.50659695e-01 3.61887634e-01 4.69194129e-02 6.24651790e-01 1.13423932e+00 -4.31167156e-01 6.08011067e-01 2.64332443e-02 -8.56606960e-01 -9.30640578e-01 -1.20486486e+00 4.23364043e-01 1.19122827e+00 5.39957643e-01 -3.79157156e-01 -5.84354758e-01 -7.85456181e-01 2.78560191e-01 -2.08863392e-02 -9.23489332e-01 -1.36736900e-01 -7.94837594e-01 -4.64848548e-01 4.77066636e-01 5.25508583e-01 3.02101701e-01 -8.13340962e-01 -1.03519320e+00 6.80134967e-02 -2.47284308e-01 -1.13213181e+00 -9.55501974e-01 -1.91374421e-01 -4.27062273e-01 -1.23967361e+00 -1.15551925e+00 -7.71714211e-01 8.95417392e-01 5.54120421e-01 1.42718005e+00 2.74376988e-01 -5.56358814e-01 6.36083364e-01 -4.02231812e-02 -1.77353814e-01 1.38331190e-01 3.53234895e-02 5.92434108e-01 1.36871979e-01 3.43337476e-01 -1.04619481e-01 -7.73987234e-01 3.68702888e-01 -1.54726371e-01 -8.42513219e-02 4.11237627e-01 4.69116628e-01 4.16985095e-01 -8.72455388e-02 6.53006360e-02 -4.34902966e-01 2.76905715e-01 -1.38000980e-01 -6.79653645e-01 3.77712727e-01 -3.87047119e-02 6.43542334e-02 2.98582911e-01 -6.02684617e-01 -6.62509143e-01 5.86686552e-01 4.21599150e-01 -7.02880442e-01 -1.08962707e-01 -1.65777326e-01 1.51838094e-01 -3.37935746e-01 6.58099413e-01 -1.11156099e-01 -7.74801746e-02 -3.80417079e-01 5.34309387e-01 -1.21490054e-01 8.52153182e-01 -2.82591909e-01 1.06202197e+00 7.49639571e-01 3.27523440e-01 -4.58022058e-01 -9.47714627e-01 -8.61881375e-01 -1.11506832e+00 -6.47329807e-01 9.90814090e-01 -1.12904084e+00 -1.29012275e+00 3.17969918e-01 -1.31534803e+00 -4.80803698e-02 -3.20005029e-01 5.02085686e-01 -1.76423267e-01 5.95435083e-01 -4.11930174e-01 -8.55381787e-01 -7.79995173e-02 -6.44130290e-01 1.14645243e+00 3.54503334e-01 -4.93163049e-01 -6.37143075e-01 2.21939042e-01 1.87049538e-01 8.82383361e-02 4.59143311e-01 -1.89300328e-01 -4.03752297e-01 -6.77406013e-01 -2.66074032e-01 -6.71395436e-02 -3.89914632e-01 -3.94003801e-02 -3.51108313e-01 -7.84545600e-01 -8.51122856e-01 -6.06013894e-01 -1.12801068e-01 1.07975292e+00 6.16404295e-01 7.11298466e-01 -2.54678607e-01 -9.15555298e-01 6.95228636e-01 1.06291890e+00 -4.40211654e-01 2.05217347e-01 4.26695824e-01 1.00581801e+00 8.83265615e-01 2.09045619e-01 3.04889768e-01 5.54981828e-01 1.11418426e+00 4.61092651e-01 -2.97061116e-01 -1.69789687e-01 -1.45652488e-01 1.99705452e-01 -4.73015197e-03 -4.11549181e-01 -2.39559308e-01 -6.91860974e-01 6.14936829e-01 -2.25628066e+00 -1.40929472e+00 -2.47790307e-01 2.47380209e+00 2.91526109e-01 1.38572797e-01 7.71617115e-01 -2.03644425e-01 1.04880428e+00 -2.05805987e-01 -4.73732412e-01 2.66963959e-01 -8.47735554e-02 -2.81172305e-01 7.35455215e-01 5.69545269e-01 -1.49327397e+00 7.80601799e-01 6.05808783e+00 8.95622149e-02 -6.27163172e-01 -2.21619252e-02 2.65988708e-01 -6.25740528e-01 3.56521249e-01 -3.21155846e-01 -1.10004807e+00 1.66051820e-01 2.67153978e-01 1.18268982e-01 5.08202314e-01 4.98934507e-01 -1.01196930e-01 -1.53327972e-01 -1.43142939e+00 1.27262139e+00 4.78477180e-01 -1.24326551e+00 -2.64133781e-01 6.10730909e-02 6.75447047e-01 -8.20007250e-02 4.01380248e-02 5.11326231e-02 3.59411895e-01 -9.43844259e-01 1.04686165e+00 7.59531617e-01 3.61029893e-01 -6.62618577e-01 6.17019355e-01 1.59191102e-01 -1.66065156e+00 -1.69277534e-01 -3.98917615e-01 -2.79156715e-01 2.59615630e-01 2.17115626e-01 -6.03944898e-01 7.41468668e-01 1.05353081e+00 5.70193470e-01 -9.40468729e-01 1.53147411e+00 7.36227483e-02 -3.53762329e-01 -4.57260549e-01 2.81599686e-02 5.38795218e-02 9.82738584e-02 7.75832951e-01 1.13150883e+00 2.82706678e-01 -2.37525538e-01 6.02199078e-01 8.53647470e-01 1.04156239e-02 -2.59135216e-01 -6.98856235e-01 4.97501343e-01 2.09208012e-01 1.18663907e+00 -8.13694775e-01 -2.68705904e-01 -2.52285153e-01 1.33930123e+00 5.99694967e-01 5.90749532e-02 -8.84022534e-01 1.55151695e-01 7.00110435e-01 3.64111245e-01 3.85177612e-01 -2.39454091e-01 2.30380103e-01 -1.10308528e+00 2.09300414e-01 -6.32056594e-01 6.85914040e-01 -6.32174313e-01 -1.54456890e+00 8.06845546e-01 4.80439849e-02 -1.26909912e+00 -1.59972727e-01 -5.06182849e-01 -3.33543092e-01 7.69725442e-01 -1.16767156e+00 -1.20613909e+00 -5.56623161e-01 7.21755266e-01 2.43764967e-01 8.91996454e-03 4.96752590e-01 2.94796973e-01 -4.91584510e-01 6.34467602e-01 -4.13253039e-01 2.48402148e-01 8.82026613e-01 -1.16159964e+00 6.70404673e-01 1.10520947e+00 2.54845321e-01 8.40541422e-01 8.37077022e-01 -9.96124268e-01 -1.08475304e+00 -1.17190075e+00 1.07702541e+00 -7.98239350e-01 1.69427320e-01 -5.38195193e-01 -5.78562558e-01 8.03879678e-01 9.76527259e-02 9.07753259e-02 3.41513067e-01 3.17179412e-01 -2.67636657e-01 -3.06182634e-03 -1.02401602e+00 6.42071724e-01 1.66431820e+00 -5.87900467e-02 -5.45322120e-01 5.64116478e-01 4.74625051e-01 -5.06278396e-01 -3.13303292e-01 1.06231168e-01 6.43125832e-01 -1.06566072e+00 1.35983908e+00 -7.77192235e-01 -2.11135864e-01 -6.10355496e-01 2.07554594e-01 -1.14171863e+00 -8.18750978e-01 -8.00646067e-01 -1.67882890e-01 7.78564572e-01 1.32771686e-01 -2.93304026e-01 1.04733944e+00 3.89483839e-01 1.13184378e-01 -4.00115639e-01 -9.75258887e-01 -1.10867381e+00 -3.45637321e-01 1.53201580e-01 5.03513098e-01 7.08518803e-01 -1.13989584e-01 2.55873859e-01 -8.04794312e-01 5.48132777e-01 1.04661453e+00 3.06474306e-02 9.96767104e-01 -1.70606709e+00 -5.10886252e-01 -3.98737580e-01 -5.66402793e-01 -9.75686550e-01 -4.71365191e-02 -7.77297139e-01 1.13486730e-01 -1.69132280e+00 4.71375525e-01 -2.28327051e-01 1.06539391e-02 4.31953013e-01 -2.59936452e-01 4.04586315e-01 6.53800964e-01 1.29842609e-01 -1.21705747e+00 2.56344616e-01 9.92681980e-01 -2.47893021e-01 -1.37529910e-01 1.47199616e-01 -5.48833787e-01 8.26783657e-01 6.15159690e-01 -5.62923193e-01 -1.03961118e-01 -5.94414294e-01 5.07182062e-01 -2.32565850e-01 8.95200610e-01 -1.60473359e+00 5.15019774e-01 2.36733392e-01 8.93451929e-01 -5.90182602e-01 7.24678576e-01 -9.02204692e-01 4.56225306e-01 8.19545925e-01 -1.55603290e-01 4.73405123e-01 2.80244619e-01 8.23091924e-01 4.23644960e-01 4.48450372e-02 6.44440114e-01 -3.83978575e-01 -6.94045067e-01 4.44328874e-01 -1.29412040e-02 -3.28783058e-02 1.17292762e+00 -3.13746095e-01 -2.37774625e-01 -4.20987517e-01 -9.70839024e-01 4.95145917e-01 6.46582961e-01 5.82655966e-01 4.75774527e-01 -1.26917207e+00 -8.17117453e-01 -9.36826319e-03 9.64158848e-02 -3.29753160e-01 3.29599202e-01 6.99921608e-01 -3.39874119e-01 2.96593696e-01 -4.75090444e-01 -7.08178103e-01 -1.65988135e+00 7.78829634e-01 7.43195534e-01 -4.86349255e-01 -7.79238522e-01 1.17319369e+00 2.55843759e-01 -4.75223362e-01 4.72264498e-01 2.40088943e-02 -1.97507441e-01 -2.47864239e-02 4.20106709e-01 4.09399927e-01 -1.84889004e-01 -9.66187835e-01 -8.01481724e-01 9.57275808e-01 6.44272640e-02 7.76051637e-03 1.13188910e+00 -3.06803405e-01 2.15919152e-01 1.21140949e-01 5.77685058e-01 4.07669619e-02 -1.51117110e+00 -2.45320112e-01 -4.05080825e-01 -5.65284073e-01 -5.25937319e-01 -8.19167137e-01 -1.05411506e+00 5.28724194e-01 7.62469530e-01 -3.46582867e-02 7.04031825e-01 1.74613014e-01 4.77264196e-01 5.49040675e-01 4.53193069e-01 -1.10882390e+00 2.21470535e-01 3.52351695e-01 7.52807319e-01 -1.26417255e+00 2.58878738e-01 -4.14339393e-01 -5.63105643e-01 8.74812245e-01 8.66109133e-01 -2.74059981e-01 1.87100962e-01 -1.56045049e-01 -1.54488251e-01 -4.59740460e-01 -2.26490453e-01 -6.66048646e-01 6.13948047e-01 8.91710579e-01 1.88355148e-01 -6.26510242e-04 2.99130529e-02 4.30473328e-01 -2.49835923e-01 -2.46308118e-01 2.75709093e-01 8.16205323e-01 -2.72947133e-01 -8.68396163e-01 -7.92928278e-01 1.45750374e-01 -2.08011732e-01 2.08791986e-01 -7.12123632e-01 7.48160779e-01 5.27518034e-01 9.08937693e-01 2.37858862e-01 -3.05716783e-01 5.13871074e-01 -1.27727479e-01 9.34305489e-01 -5.13985753e-01 -7.96126425e-01 -3.86034139e-02 1.32165492e-01 -6.53723300e-01 -6.74632132e-01 -8.63215804e-01 -8.60260189e-01 -4.24109250e-01 -1.34405881e-01 -1.16148368e-01 6.93358853e-02 8.59056532e-01 3.53775233e-01 8.39553535e-01 8.41080323e-02 -1.33724332e+00 -1.99292585e-01 -5.74897110e-01 -3.33977222e-01 6.98429048e-01 3.74360949e-01 -9.00982082e-01 1.21537589e-01 2.58494653e-02]
[6.902027130126953, -1.0674060583114624]
a6eee3b4-9183-4520-86a0-c8d3599e17e6
real-time-human-motion-capture-with-multiple
1605.08068
null
http://arxiv.org/abs/1605.08068v1
http://arxiv.org/pdf/1605.08068v1.pdf
Real-Time Human Motion Capture with Multiple Depth Cameras
Commonly used human motion capture systems require intrusive attachment of markers that are visually tracked with multiple cameras. In this work we present an efficient and inexpensive solution to markerless motion capture using only a few Kinect sensors. Unlike the previous work on 3d pose estimation using a single depth camera, we relax constraints on the camera location and do not assume a co-operative user. We apply recent image segmentation techniques to depth images and use curriculum learning to train our system on purely synthetic data. Our method accurately localizes body parts without requiring an explicit shape model. The body joint locations are then recovered by combining evidence from multiple views in real-time. We also introduce a dataset of ~6 million synthetic depth frames for pose estimation from multiple cameras and exceed state-of-the-art results on the Berkeley MHAD dataset.
['James J. Little', 'Alireza Shafaei']
2016-05-25
null
null
null
null
['markerless-motion-capture']
['computer-vision']
[-3.89718427e-03 7.59928599e-02 -2.15499759e-01 -1.22740977e-01 -9.44237769e-01 -6.46557629e-01 3.62415969e-01 -2.74702102e-01 -9.24251497e-01 6.48884535e-01 -1.70805439e-01 1.25896499e-01 6.36826217e-01 -2.71086544e-01 -8.40386093e-01 -3.22048962e-01 1.36200175e-01 8.01852703e-01 6.41179800e-01 2.14787960e-01 2.86558624e-02 5.67470610e-01 -1.06904769e+00 -4.82660502e-01 2.37316012e-01 6.18761480e-01 1.04401499e-01 1.31392276e+00 5.75724900e-01 3.76779586e-01 -4.61177468e-01 -2.30178609e-01 4.98985738e-01 -3.71553034e-01 -5.09926915e-01 6.32684171e-01 9.91780818e-01 -9.30885971e-01 -3.06137413e-01 7.01958418e-01 9.13987696e-01 1.35319993e-01 3.16524327e-01 -1.06827581e+00 2.84984857e-01 -3.14476848e-01 -9.88069296e-01 2.39521161e-01 1.06882715e+00 5.95780499e-02 2.19265997e-01 -7.55793631e-01 8.65383863e-01 8.97785425e-01 9.44983780e-01 9.99820769e-01 -1.14368165e+00 -3.29551637e-01 -2.75653321e-02 -3.68139476e-01 -1.40470588e+00 -6.11792624e-01 7.36717403e-01 -5.88142395e-01 9.14001286e-01 8.99686664e-02 1.06624520e+00 1.15165651e+00 1.55432541e-02 7.02191293e-01 7.08672464e-01 -5.30214310e-01 1.33437440e-01 -2.92998224e-01 -2.31640145e-01 1.08245003e+00 4.12762016e-01 -2.18702704e-02 -6.71587169e-01 -2.79089123e-01 1.61641741e+00 7.62995332e-02 -2.70521283e-01 -1.17035317e+00 -1.43324924e+00 5.72562277e-01 -2.55241748e-02 -2.72011131e-01 -2.28261009e-01 6.36914909e-01 -3.42959166e-02 -3.29327464e-01 2.77426481e-01 1.97345880e-03 -3.24790418e-01 -4.93283272e-01 -1.08009744e+00 3.09431225e-01 6.35269403e-01 1.07243967e+00 4.58935648e-01 -5.57653531e-02 5.99907279e-01 3.13769430e-01 5.77522397e-01 7.35457897e-01 4.67475116e-01 -1.65585291e+00 5.61541617e-01 2.36604169e-01 5.14838278e-01 -6.63541794e-01 -4.05820131e-01 4.06832874e-01 -3.30068409e-01 5.12794793e-01 6.94792569e-01 -5.39122581e-01 -8.88970971e-01 1.27178001e+00 9.50202227e-01 5.98271012e-01 -2.86798865e-01 1.12996793e+00 3.92954081e-01 1.05731543e-02 -3.12933028e-01 1.14631906e-01 9.56276596e-01 -9.81316268e-01 -4.58435416e-01 -7.49897063e-01 3.28783989e-01 -5.67888081e-01 5.91480017e-01 5.59594512e-01 -1.34957206e+00 -2.94618726e-01 -1.20905626e+00 -2.35706657e-01 1.53007284e-01 3.80110666e-02 6.46706223e-01 8.48971844e-01 -1.05183017e+00 4.72334892e-01 -1.65816963e+00 -5.22708118e-01 6.85561746e-02 8.19199622e-01 -6.69322670e-01 8.09374973e-02 -5.49013078e-01 9.82390821e-01 2.82978476e-03 2.84428913e-02 -9.48034883e-01 -2.84432232e-01 -1.27556193e+00 -8.38350773e-01 1.82908893e-01 -1.24615061e+00 1.58754134e+00 -3.72668266e-01 -1.84025812e+00 1.25929236e+00 -2.54603893e-01 -2.36462265e-01 9.43361938e-01 -8.87282014e-01 3.27700317e-01 7.70371854e-01 -1.24606285e-02 8.47862542e-01 6.46245480e-01 -1.15657902e+00 -3.31193119e-01 -4.23886389e-01 -8.51246789e-02 5.95092356e-01 2.39575967e-01 6.31386694e-03 -1.00148940e+00 -1.80408314e-01 4.16282296e-01 -1.33050263e+00 -3.81754965e-01 5.49572706e-01 -3.57768416e-01 5.78625858e-01 5.93309045e-01 -4.75532442e-01 5.75052261e-01 -1.65136981e+00 3.58150840e-01 -4.37826999e-02 2.88792104e-01 -4.35346104e-02 4.71104354e-01 -2.17111975e-01 3.63051027e-01 -2.68602610e-01 -1.17663862e-02 -9.86981213e-01 -1.65931448e-01 1.93780467e-01 1.82799265e-01 1.04033840e+00 -4.06654805e-01 6.81884050e-01 -9.14992630e-01 -8.96501303e-01 6.89916968e-01 6.14648819e-01 -4.74939585e-01 3.09540123e-01 1.93997756e-01 6.69679224e-01 -3.06256860e-01 8.67414176e-01 4.52821791e-01 -2.36079484e-01 6.84858486e-02 2.18645856e-01 2.26596415e-01 8.37640539e-02 -1.48531473e+00 2.40631986e+00 -8.15450773e-02 3.75835925e-01 4.02465969e-01 -3.57721448e-01 3.58964115e-01 5.20964384e-01 8.61834764e-01 4.86090370e-02 8.23249966e-02 6.47516847e-02 -4.61920917e-01 -2.71027774e-01 3.58095616e-01 -3.65028948e-01 -2.46310174e-01 4.16162610e-01 -6.29083207e-03 -3.18606883e-01 -3.32714230e-01 -5.38649783e-02 1.08905971e+00 6.64131463e-01 4.28981960e-01 4.07105416e-01 -3.50436382e-02 1.72620699e-01 4.93632764e-01 4.97800887e-01 -4.58289176e-01 1.36036956e+00 -1.38529703e-01 -3.93636674e-01 -9.73926783e-01 -1.27076137e+00 9.95879695e-02 6.05286896e-01 5.11868656e-01 -1.99332505e-01 -8.49385381e-01 -4.60682154e-01 2.81302501e-02 -1.01316340e-01 -4.30345982e-01 3.02393943e-01 -8.28683972e-01 -5.19453585e-01 7.93142915e-01 8.52993488e-01 4.32594448e-01 -4.46066141e-01 -1.40961921e+00 1.13184743e-01 -9.78109092e-02 -1.45659387e+00 -6.13032818e-01 -3.55867110e-02 -1.17124641e+00 -1.19876742e+00 -1.19207025e+00 -5.36141098e-01 8.23263347e-01 3.41144890e-01 1.06204641e+00 -1.04280770e-01 -2.80533463e-01 9.07545090e-01 1.37161106e-01 -2.03449160e-01 2.78446257e-01 -1.11266151e-01 3.86444509e-01 -4.95723575e-01 9.72166881e-02 -3.97811234e-01 -9.07123089e-01 3.55435252e-01 -1.53169006e-01 1.66684642e-01 2.00267673e-01 3.88247490e-01 7.51812696e-01 -7.73763239e-01 -4.95806456e-01 -5.48709393e-01 -1.71594974e-02 -1.63094431e-01 -6.53859556e-01 -2.26211458e-01 -1.87963285e-02 -4.41886991e-01 -1.53953269e-01 -7.91091204e-01 -8.06314111e-01 8.79827499e-01 8.94244686e-02 -9.54155266e-01 -3.18550289e-01 -2.12094113e-01 -3.92875858e-02 -3.42132330e-01 5.62382877e-01 -1.81191012e-01 -3.05143325e-03 -3.51729393e-01 2.79222637e-01 1.06304906e-01 1.14447260e+00 -5.94668388e-01 7.60791719e-01 8.79210532e-01 3.99423689e-02 -7.59746194e-01 -3.38344425e-01 -7.08072066e-01 -1.28231025e+00 -2.43804842e-01 1.10135961e+00 -1.25548720e+00 -8.43552113e-01 5.60422778e-01 -1.22857845e+00 -5.43126345e-01 -3.14405449e-02 9.48441684e-01 -9.14142132e-01 5.04682004e-01 -8.36088955e-01 -8.16352725e-01 -8.86622742e-02 -9.71935511e-01 1.78356051e+00 2.95765013e-01 -6.85186744e-01 -1.16652334e+00 5.19477785e-01 4.33917314e-01 -3.46832812e-01 7.99711466e-01 -3.13817531e-01 1.09586053e-01 -7.97736406e-01 -4.53854471e-01 5.70929170e-01 -2.80879438e-01 8.91410261e-02 -1.49314463e-01 -9.96380925e-01 -2.73960352e-01 -8.04797560e-03 -3.50259423e-01 3.22351098e-01 7.34070182e-01 4.85679358e-01 1.37511909e-01 -7.02395260e-01 9.61629868e-01 1.33091116e+00 8.48027021e-02 5.11634171e-01 5.64035177e-01 1.21267903e+00 2.52348483e-01 6.84289515e-01 3.61556947e-01 5.95996737e-01 8.19231987e-01 3.51683646e-01 -2.32393175e-01 1.59267575e-01 -3.44689965e-01 4.89802867e-01 5.70084691e-01 -3.64744842e-01 8.33063852e-03 -1.05438387e+00 5.61483800e-01 -1.69743824e+00 -7.65412092e-01 3.09586572e-03 2.32720089e+00 7.69306898e-01 1.15925930e-01 4.11886275e-01 -6.63940534e-02 5.94331920e-01 -6.88959807e-02 -5.42438388e-01 8.50741938e-02 3.41142029e-01 2.22386092e-01 7.81025887e-01 8.66360188e-01 -1.29845905e+00 8.28316927e-01 7.37129211e+00 -2.24955499e-01 -8.65582228e-01 1.21037148e-01 2.09263012e-01 -7.19258249e-01 1.79693803e-01 -1.47309631e-01 -9.20664489e-01 2.19553843e-01 6.89222634e-01 1.92063093e-01 -2.10161675e-02 9.94482219e-01 -1.44899655e-02 -7.52872169e-01 -1.27393723e+00 1.37019408e+00 3.45573336e-01 -9.70332146e-01 -7.47373581e-01 3.96990120e-01 6.88216805e-01 2.35123813e-01 -2.16763988e-01 -4.08302397e-01 4.31737781e-01 -7.99150765e-01 7.01292634e-01 6.92304015e-01 7.19138384e-01 -5.05030334e-01 3.12207818e-01 7.35671461e-01 -1.09456813e+00 6.50063634e-01 -3.14325452e-01 -2.75218219e-01 5.00774443e-01 9.61379930e-02 -9.20980990e-01 6.99813813e-02 5.68333209e-01 4.94853109e-01 -5.29027164e-01 1.19862902e+00 -3.38966459e-01 1.21921882e-01 -7.86920369e-01 3.01245540e-01 -2.73272187e-01 -1.96470246e-02 4.42809880e-01 9.96279418e-01 3.83217990e-01 1.67048216e-01 3.37240249e-01 3.26551944e-01 1.35645896e-01 -6.39121890e-01 -8.97295117e-01 5.35676539e-01 3.38828593e-01 1.09238756e+00 -9.05281186e-01 -3.91290784e-01 -3.80379379e-01 1.63912189e+00 1.56108811e-02 1.24242440e-01 -8.26511443e-01 -1.07072294e-01 6.45107090e-01 2.68863618e-01 2.72793770e-01 -8.40236783e-01 -1.93493485e-01 -1.49817467e+00 1.28545135e-01 -4.27989215e-01 1.52953774e-01 -9.68101203e-01 -7.84953475e-01 7.31891692e-02 3.10202092e-01 -1.36549115e+00 -9.42938089e-01 -4.79662567e-01 -3.85486871e-01 8.37293983e-01 -1.03457499e+00 -9.79473233e-01 -4.55140263e-01 8.45689356e-01 3.87001872e-01 3.92244965e-01 8.95901978e-01 1.91638663e-01 -3.60890388e-01 2.56970167e-01 -3.40057462e-01 4.39125717e-01 8.26154411e-01 -1.48936808e+00 7.51215935e-01 8.88074577e-01 2.08065212e-01 7.74996400e-01 5.90121627e-01 -8.70633602e-01 -1.61997175e+00 -5.42758405e-01 4.48128849e-01 -1.32387137e+00 1.67484760e-01 -5.37264049e-01 -3.92608553e-01 1.18198538e+00 -3.37983258e-02 6.03877306e-01 6.01251900e-01 -2.34562799e-01 2.27432713e-01 3.57840747e-01 -1.22782123e+00 3.02831709e-01 1.09859526e+00 -5.38392186e-01 -8.00204933e-01 2.65990682e-02 2.59878337e-01 -1.51802468e+00 -5.84309578e-01 4.65244502e-02 8.90152812e-01 -8.08022857e-01 1.33620918e+00 -1.37667611e-01 -1.49432216e-02 -5.08484483e-01 -8.10655206e-03 -8.88989210e-01 1.22639403e-01 -8.00383389e-01 -5.01433015e-01 7.82241642e-01 -1.15088083e-01 -8.73460993e-02 1.50866175e+00 1.14655340e+00 2.44913131e-01 -3.62407833e-01 -1.16454160e+00 -5.17404795e-01 -2.23204419e-01 -3.31277966e-01 -1.03475757e-01 7.91848958e-01 3.24945264e-02 1.19233117e-01 -5.52965939e-01 4.31481242e-01 1.00471675e+00 -7.25105554e-02 1.23997271e+00 -1.00955129e+00 -5.82105637e-01 1.70839921e-01 -6.13868058e-01 -1.37991750e+00 -7.08379596e-02 -1.79630116e-01 2.35834792e-01 -1.44108748e+00 1.38424292e-01 1.22172892e-01 4.13888782e-01 3.29731047e-01 -1.47511810e-01 7.25749373e-01 5.88211790e-03 1.69426501e-01 -9.01723862e-01 8.20170566e-02 1.04667962e+00 2.51511395e-01 -9.80229229e-02 5.76378219e-02 -1.25040859e-01 1.39550579e+00 4.71373558e-01 -5.32742441e-01 -4.60226119e-01 -7.17198074e-01 -9.47661139e-03 5.24299681e-01 7.31260955e-01 -1.20444226e+00 4.90286857e-01 -1.81459710e-01 1.12197340e+00 -6.76225603e-01 9.16986108e-01 -7.37472534e-01 3.58666629e-01 5.19880533e-01 2.34184980e-01 4.31158602e-01 2.64849931e-01 5.19830525e-01 2.51681477e-01 -1.92301184e-01 6.54569447e-01 -6.76860929e-01 -6.78618670e-01 3.05096447e-01 -4.32027876e-01 7.28509203e-02 1.08881235e+00 -7.24134624e-01 1.11636952e-01 -4.75629658e-01 -9.61166918e-01 6.54309988e-02 1.17638314e+00 7.29969069e-02 8.61252189e-01 -1.37697458e+00 -1.51829749e-01 9.51178595e-02 -3.09768587e-01 5.48920691e-01 -1.43819079e-01 6.28697217e-01 -1.08803570e+00 2.10580021e-01 -1.47986442e-01 -1.01390421e+00 -1.50116050e+00 1.26118347e-01 6.28094733e-01 1.65735394e-01 -7.70715535e-01 7.67637789e-01 -1.12635463e-01 -3.64293605e-01 2.19083279e-01 -3.09722960e-01 3.49280983e-01 -2.60774851e-01 3.97310644e-01 5.42082012e-01 -1.74757570e-01 -7.93358564e-01 -6.18580759e-01 1.14683080e+00 2.89999127e-01 -9.12733138e-01 1.08120906e+00 -3.99969995e-01 4.41664189e-01 6.07086360e-01 1.01017296e+00 3.17888498e-01 -1.85475969e+00 2.58932382e-01 -1.74619168e-01 -7.66734481e-01 -2.11157128e-01 -3.21491212e-01 -8.06185782e-01 8.96110713e-01 6.72681928e-01 -5.92437446e-01 6.70288920e-01 6.84163421e-02 7.98454404e-01 2.96947479e-01 7.60838687e-01 -1.08425605e+00 3.19717824e-01 2.52558708e-01 2.85915196e-01 -1.32851708e+00 4.47264016e-01 -4.10860270e-01 -5.78833759e-01 8.69790018e-01 7.89365828e-01 -3.88329893e-01 2.81843215e-01 6.45434916e-01 2.25034803e-01 -1.78214952e-01 -2.55478799e-01 -4.89972308e-02 2.02345148e-01 9.32718456e-01 3.47277850e-01 -2.07415685e-01 1.86320066e-01 3.17770727e-02 -2.26139203e-01 2.00959086e-01 6.39191806e-01 1.58182716e+00 -2.94617504e-01 -1.10019219e+00 -7.31163323e-01 -2.49006242e-01 -7.25801885e-01 2.90484816e-01 -5.66708505e-01 1.09831595e+00 4.14178409e-02 6.20954394e-01 1.08480416e-01 -1.92347407e-01 2.58105993e-01 7.32165873e-02 1.20266485e+00 -9.06551659e-01 -1.80236876e-01 6.23883247e-01 2.21139714e-02 -7.90669382e-01 -8.09111536e-01 -1.03542650e+00 -1.48501790e+00 -9.62591022e-02 -2.22927511e-01 -2.95483202e-01 6.66197598e-01 6.68132067e-01 8.15753564e-02 6.65792897e-02 9.99553204e-02 -1.70534611e+00 -6.81629032e-02 -5.69447160e-01 -5.07839918e-01 1.11987136e-01 7.22028613e-01 -7.10753798e-01 -1.64982915e-01 3.72595131e-01]
[7.1447954177856445, -1.0938127040863037]
00266f35-6678-4e16-8fe7-e22f280da80d
conda-continual-unsupervised-domain-1
2212.00621
null
https://arxiv.org/abs/2212.00621v1
https://arxiv.org/pdf/2212.00621v1.pdf
CONDA: Continual Unsupervised Domain Adaptation Learning in Visual Perception for Self-Driving Cars
Although unsupervised domain adaptation methods have achieved remarkable performance in semantic scene segmentation in visual perception for self-driving cars, these approaches remain impractical in real-world use cases. In practice, the segmentation models may encounter new data that have not been seen yet. Also, the previous data training of segmentation models may be inaccessible due to privacy problems. Therefore, to address these problems, in this work, we propose a Continual Unsupervised Domain Adaptation (CONDA) approach that allows the model to continuously learn and adapt with respect to the presence of the new data. Moreover, our proposed approach is designed without the requirement of accessing previous training data. To avoid the catastrophic forgetting problem and maintain the performance of the segmentation models, we present a novel Bijective Maximum Likelihood loss to impose the constraint of predicted segmentation distribution shifts. The experimental results on the benchmark of continual unsupervised domain adaptation have shown the advanced performance of the proposed CONDA method.
['Khoa Luu', 'Jackson David Cothren', 'Ahmed Moustafa', 'Pierce Helton', 'Thanh-Dat Truong']
2022-12-01
null
null
null
null
['scene-segmentation']
['computer-vision']
[ 4.37605411e-01 -1.27441408e-02 -5.04173972e-02 -5.97726464e-01 -2.41573259e-01 -3.91553789e-01 3.90466779e-01 1.61616534e-01 -7.23164678e-01 9.24721956e-01 -3.30081403e-01 -1.13580503e-01 8.71669874e-02 -8.04529428e-01 -6.55798197e-01 -7.62683749e-01 4.73859280e-01 4.90326732e-01 7.56658614e-01 8.81157815e-02 1.87561437e-01 3.52317661e-01 -1.60418439e+00 -1.50907412e-01 1.23978603e+00 7.72879660e-01 4.34624255e-01 2.06573680e-01 -2.88102716e-01 3.44078839e-01 -4.43736821e-01 -2.52033651e-01 3.34578067e-01 -3.68024647e-01 -7.14964271e-01 6.45813465e-01 2.06107870e-01 -2.97286898e-01 -2.81594723e-01 1.22285795e+00 1.81532994e-01 5.61196089e-01 5.01269281e-01 -1.36655223e+00 -4.48915303e-01 1.25471503e-01 -6.88329458e-01 1.80794880e-01 -1.18386403e-01 7.11101815e-02 4.46859330e-01 -6.15410805e-01 6.71554267e-01 9.40691352e-01 2.98450381e-01 6.95331872e-01 -1.13058484e+00 -7.34436333e-01 4.65284884e-01 4.29982066e-01 -1.48368752e+00 -4.74675715e-01 8.96987736e-01 -4.35073823e-01 4.09835160e-01 -1.12775825e-01 3.71151417e-01 8.44610929e-01 -7.71087483e-02 7.04151511e-01 1.10088289e+00 -3.18532348e-01 4.95703280e-01 6.49307668e-01 5.81197813e-03 3.89661700e-01 2.98376322e-01 8.77316222e-02 -2.85923123e-01 1.09892435e-01 5.43569803e-01 3.55071910e-02 2.00113297e-01 -9.07618701e-01 -8.58800650e-01 5.49482703e-01 3.59623551e-01 6.40768856e-02 -1.74527779e-01 -3.58881503e-01 4.33932424e-01 1.57787010e-01 3.02661538e-01 -2.79358607e-02 -3.67418200e-01 4.67509814e-02 -8.08419347e-01 6.64685667e-03 4.26282227e-01 1.26835835e+00 1.01995385e+00 8.85353237e-03 2.82354951e-01 8.54707897e-01 1.76916048e-01 3.41981232e-01 7.40628719e-01 -5.73239803e-01 3.40320230e-01 5.16535044e-01 1.04685158e-01 -9.25900400e-01 -1.60510957e-01 -2.42030010e-01 -8.60505044e-01 1.82369038e-01 3.86550814e-01 1.20619452e-02 -1.13233542e+00 1.65246069e+00 5.74246287e-01 3.09029549e-01 2.98255712e-01 7.34410226e-01 5.30680716e-01 6.89363480e-01 4.78680432e-01 -4.13608819e-01 7.53479421e-01 -8.55499089e-01 -6.99339867e-01 -3.50919217e-01 3.98783445e-01 -6.33682013e-01 1.00777769e+00 2.54487157e-01 -7.23303676e-01 -1.02170777e+00 -1.11917520e+00 1.55596763e-01 -4.47678655e-01 -8.65986869e-02 4.03779000e-01 6.54995620e-01 -5.43235540e-01 1.44689187e-01 -8.50249290e-01 -7.57758856e-01 6.27517164e-01 3.50868464e-01 -3.00415933e-01 -1.18755825e-01 -1.08312869e+00 6.24929309e-01 8.97697389e-01 -4.59173359e-02 -8.54408741e-01 -3.51530612e-01 -6.28059387e-01 -1.15570791e-01 7.30008602e-01 -4.10798490e-01 1.21935451e+00 -1.20030129e+00 -1.47973371e+00 7.57270157e-01 -3.10300529e-01 -7.45856762e-01 7.81730473e-01 -3.15615349e-02 -6.01673186e-01 -1.49022654e-01 1.83930144e-01 7.78751254e-01 8.34616959e-01 -1.33817792e+00 -1.00913441e+00 -3.06313008e-01 -1.65321097e-01 3.50903004e-01 -4.48515087e-01 -6.28741324e-01 -5.20987809e-01 -3.82474303e-01 2.69365042e-01 -7.82153964e-01 -4.25809264e-01 5.50000332e-02 -1.35967374e-01 -5.57587184e-02 1.16923964e+00 -3.33311111e-01 1.02784622e+00 -2.64451480e+00 -2.83786595e-01 2.61393458e-01 -1.56902790e-01 4.91265476e-01 8.63979310e-02 9.58685428e-02 1.93491340e-01 -6.66110516e-02 -6.96711063e-01 -2.52399057e-01 -2.88048416e-01 4.72432584e-01 -2.38564387e-01 1.80250645e-01 7.59705603e-02 3.64714205e-01 -6.61352634e-01 -8.07471454e-01 3.71033728e-01 2.72778012e-02 -5.40031254e-01 1.87869787e-01 -3.80915910e-01 7.36082077e-01 -6.10144794e-01 2.59794623e-01 1.03431046e+00 6.06571399e-02 8.50536972e-02 1.06805719e-01 3.72476466e-02 -1.67186290e-01 -1.19657862e+00 1.65547907e+00 -1.82581395e-01 2.52665401e-01 -1.86824471e-01 -1.20262551e+00 1.07032216e+00 9.62825492e-02 4.03526574e-01 -8.65008533e-01 -5.45898871e-03 1.46435322e-02 -8.74563009e-02 -6.10347688e-01 4.94677573e-01 -2.89335668e-01 -5.50518855e-02 4.14122716e-02 -1.75034225e-01 -3.83794233e-02 6.66569844e-02 8.04592967e-02 5.74250817e-01 6.92080930e-02 3.35154206e-01 -1.61537658e-02 8.50121021e-01 3.14833760e-01 1.00367606e+00 6.71667993e-01 -6.84117854e-01 4.15985793e-01 2.21035425e-02 -2.44522020e-01 -9.34592009e-01 -1.20040679e+00 -6.35726750e-02 8.53963733e-01 5.99709928e-01 8.68909806e-02 -7.77432501e-01 -9.23577011e-01 -1.14404798e-01 8.77738476e-01 -2.34315932e-01 -3.91362727e-01 -3.35775733e-01 -4.89245296e-01 2.67560542e-01 4.63711590e-01 1.17188227e+00 -9.69099641e-01 -6.29301369e-01 3.02962065e-01 -2.00809032e-01 -1.29502201e+00 -3.74204338e-01 -1.55426562e-01 -9.60004807e-01 -7.61294067e-01 -4.04780537e-01 -1.10728443e+00 9.35075283e-01 4.14892167e-01 4.56680566e-01 -8.27983841e-02 -1.15726404e-01 3.29471618e-01 -2.41715059e-01 -4.06494021e-01 -4.67662692e-01 2.11570799e-01 3.01969536e-02 3.90850037e-01 7.11625934e-01 -5.31848311e-01 -4.63533491e-01 4.23898965e-01 -1.02379692e+00 9.45527926e-02 3.70393962e-01 7.88351536e-01 9.00310338e-01 6.22407615e-01 9.87056673e-01 -1.22252262e+00 3.78995001e-01 -4.21192855e-01 -8.77632856e-01 1.31017879e-01 -9.68846858e-01 -2.51559734e-01 6.07500553e-01 -4.90043759e-01 -1.64629865e+00 4.86762762e-01 5.84948398e-02 -2.25351155e-01 -5.58749855e-01 3.47716093e-01 -4.94886070e-01 9.40517113e-02 5.26044250e-01 4.87338185e-01 1.35780171e-01 -4.72430646e-01 2.11927548e-01 7.82669365e-01 6.47313237e-01 -2.95975834e-01 8.98419201e-01 5.58092475e-01 -1.27082825e-01 -8.37382019e-01 -4.59996223e-01 -5.64779341e-01 -8.51886570e-01 -1.14953898e-01 9.42952216e-01 -8.92479360e-01 -3.33634168e-01 6.00246608e-01 -8.01327586e-01 -1.02346323e-01 -3.95726889e-01 4.18127775e-01 -5.91441691e-01 7.95599103e-01 1.23853341e-01 -7.96089351e-01 1.53534263e-01 -9.80631292e-01 2.44913906e-01 6.30121112e-01 6.39696792e-02 -8.41651976e-01 -1.88115045e-01 4.12727922e-01 3.26440662e-01 1.40832081e-01 8.43415976e-01 -7.72649765e-01 -7.98068345e-01 -1.76735967e-01 -1.52838990e-01 4.72257346e-01 4.72786784e-01 -1.82011276e-01 -8.23561668e-01 -3.86503935e-01 1.38418987e-01 -1.30995587e-01 7.50150204e-01 6.89201355e-02 1.24480450e+00 -1.68179318e-01 -5.42923510e-01 3.66763860e-01 1.36920822e+00 7.44780719e-01 6.84977591e-01 3.85032088e-01 5.76225281e-01 5.19243836e-01 1.05219769e+00 4.89858896e-01 3.51705939e-01 3.75023127e-01 2.12863326e-01 -1.71237260e-01 6.91154152e-02 -5.84636629e-01 4.72244062e-02 5.77805638e-01 3.14490944e-01 -1.63739741e-01 -8.26380193e-01 9.19686317e-01 -1.85990536e+00 -6.96857512e-01 3.67111675e-02 2.49871635e+00 6.16520941e-01 5.62586069e-01 3.52594368e-02 -3.65964584e-02 7.98035324e-01 -1.00427561e-01 -1.01975679e+00 -4.21555966e-01 2.59901602e-02 -6.63752630e-02 5.82086802e-01 4.02825534e-01 -1.17017233e+00 1.23317051e+00 5.83819532e+00 7.70419717e-01 -1.04531181e+00 2.48022258e-01 4.90123063e-01 2.75028914e-01 -1.53193682e-01 2.23055124e-01 -5.82706273e-01 5.02555549e-01 4.93093789e-01 -3.53828162e-01 3.06612670e-01 9.65050280e-01 9.77864489e-02 -5.31377316e-01 -8.36078227e-01 8.77953291e-01 -1.58874631e-01 -7.45660007e-01 9.71414447e-02 -1.37617260e-01 8.34493041e-01 -2.68713087e-01 2.06980899e-01 2.80543357e-01 8.62761438e-02 -5.50494134e-01 3.02294225e-01 4.40721363e-01 6.14147723e-01 -9.36702073e-01 5.08354545e-01 6.69448256e-01 -9.06037807e-01 -1.14850618e-01 -5.92896283e-01 1.10330641e-01 3.04076681e-03 3.69949907e-01 -1.11206710e+00 4.19878304e-01 6.55342519e-01 5.54843485e-01 -6.62009478e-01 1.21223927e+00 -2.56122146e-02 6.75878108e-01 -3.54118496e-01 1.33971930e-01 1.53516248e-01 -1.17940143e-01 6.86500788e-01 8.15763712e-01 1.16403356e-01 7.05287531e-02 2.90577531e-01 6.94447577e-01 6.82470649e-02 3.17042530e-01 -6.78531706e-01 -2.14284062e-02 5.06643414e-01 7.88601696e-01 -7.37076461e-01 -3.88196617e-01 -3.75705928e-01 1.08910906e+00 1.26118630e-01 5.25980532e-01 -7.72370338e-01 -2.70095050e-01 3.77610207e-01 2.86767930e-01 4.49338585e-01 -2.86758453e-01 -3.61885399e-01 -9.31805253e-01 1.17664680e-01 -6.53346896e-01 6.14906132e-01 -5.75341463e-01 -1.30030787e+00 4.28113550e-01 1.22062258e-01 -1.33749974e+00 -1.38562834e-02 -5.85162304e-02 -4.41186696e-01 5.17684758e-01 -1.48159158e+00 -8.23965311e-01 -2.60492980e-01 7.77379811e-01 7.25851893e-01 -4.05766487e-01 5.51532686e-01 2.64574945e-01 -5.28978467e-01 7.65439570e-01 3.47290814e-01 -2.51479428e-02 8.33761990e-01 -8.11498702e-01 1.32556349e-01 1.05342472e+00 -1.20598644e-01 2.80166149e-01 6.53030396e-01 -8.00894141e-01 -6.71185613e-01 -1.22191691e+00 6.62380338e-01 5.16805090e-02 1.98746458e-01 -2.66888440e-01 -1.23995125e+00 6.58452570e-01 -1.14397779e-02 -2.43948787e-01 5.39112389e-01 -5.51365316e-02 -1.81069180e-01 -6.92961037e-01 -1.55504441e+00 5.52135766e-01 8.57459068e-01 -3.06846648e-01 -5.69801509e-01 1.17082618e-01 5.31649590e-01 -2.16204107e-01 -4.08865988e-01 4.87013161e-01 1.74498111e-01 -7.62289643e-01 6.88487411e-01 -4.21223372e-01 -8.32616631e-03 -6.51255786e-01 3.76297235e-02 -9.88871217e-01 -2.33157188e-01 -3.66221458e-01 3.14163476e-01 1.47818172e+00 2.20003664e-01 -9.39729452e-01 8.27641547e-01 8.81966114e-01 1.52654767e-01 -1.20759793e-01 -1.11121750e+00 -8.13117743e-01 2.66308561e-02 -1.97747439e-01 8.10075045e-01 8.87819648e-01 -1.76422909e-01 1.70386538e-01 -2.86144793e-01 4.41801161e-01 6.44446790e-01 2.93063819e-02 9.35097516e-01 -1.04065347e+00 -5.42067662e-02 7.23233894e-02 -6.50042117e-01 -1.16067696e+00 -8.62548221e-03 -6.77289128e-01 3.67259860e-01 -1.33544111e+00 2.20002607e-01 -6.31671011e-01 -6.59098446e-01 2.60342807e-01 -9.59258005e-02 -1.35492429e-01 9.58835185e-02 2.93627471e-01 -6.81778610e-01 7.00135350e-01 1.24802387e+00 -1.85448483e-01 -3.69433045e-01 1.98890910e-01 -6.17015183e-01 7.22104669e-01 1.04662335e+00 -6.94195211e-01 -9.92403626e-01 -4.18668717e-01 -4.27664012e-01 -8.23079050e-02 2.72887051e-01 -1.32203877e+00 4.41295475e-01 -3.90554398e-01 3.47198933e-01 -7.59983480e-01 1.16706498e-01 -1.18857288e+00 1.89987913e-01 2.03837097e-01 -2.18309179e-01 -4.01219785e-01 3.94804955e-01 9.87014294e-01 -3.07265162e-01 -1.82661414e-01 1.01008081e+00 -2.73655616e-02 -1.26843345e+00 4.13252980e-01 -5.16548932e-01 -5.83621971e-02 1.42397273e+00 -6.16176665e-01 1.39086708e-01 -2.37763897e-01 -8.53782237e-01 4.60017353e-01 5.75478077e-01 4.94059622e-01 5.93676448e-01 -1.08673143e+00 -2.95521706e-01 4.99373019e-01 4.07588601e-01 3.49099278e-01 5.88081896e-01 3.44898671e-01 -2.21639559e-01 1.09627984e-01 -4.28131431e-01 -5.81422806e-01 -1.10319018e+00 1.04999745e+00 2.09575206e-01 8.51862356e-02 -5.15146613e-01 5.35361588e-01 4.85065907e-01 -3.61220062e-01 2.64771134e-01 -4.10568900e-02 -1.17870212e-01 -1.64099500e-01 1.48637667e-01 8.15504044e-02 -1.57746285e-01 -6.05174184e-01 -2.36245185e-01 3.72930557e-01 -5.35889864e-01 -1.41835764e-01 8.31397295e-01 -7.56763816e-01 2.70410240e-01 4.86995816e-01 8.25370073e-01 -2.37434834e-01 -1.58377457e+00 -5.80280244e-01 6.85823569e-03 -6.34860575e-01 -3.05817306e-01 -7.60655463e-01 -9.13863182e-01 8.71223629e-01 1.02238894e+00 -2.69247349e-02 1.35662615e+00 -3.13437343e-01 9.65799928e-01 4.49669540e-01 4.34879303e-01 -1.60308611e+00 -2.25599483e-01 1.98403776e-01 2.50131190e-01 -1.37633181e+00 -1.37446448e-01 -4.96997058e-01 -9.06220853e-01 7.32209206e-01 1.05209219e+00 1.76120311e-01 7.80622959e-01 -2.84906924e-01 2.33616292e-01 2.69997448e-01 -4.09311593e-01 -2.68225938e-01 -7.18556270e-02 9.28148627e-01 -2.87087500e-01 1.12880267e-01 -4.29005355e-01 4.35643882e-01 6.05204180e-02 2.03656361e-01 4.91685241e-01 9.54937100e-01 -4.79717702e-01 -1.06120086e+00 -1.54760942e-01 5.15047722e-02 -1.16799608e-01 3.70057791e-01 -2.15152144e-01 7.47456014e-01 4.98706937e-01 1.06227124e+00 8.40225145e-02 -2.20923722e-01 5.00790477e-01 2.28953749e-01 7.55341575e-02 -5.25187671e-01 1.24848217e-01 3.25593054e-02 -3.04756075e-01 -2.76989434e-02 -4.51518089e-01 -7.70134211e-01 -1.52784097e+00 -5.71101606e-02 -2.20457137e-01 1.05543517e-01 4.55930412e-01 9.55006242e-01 4.09747154e-01 2.90126532e-01 7.08418250e-01 1.62727218e-02 -3.80365878e-01 -6.50613487e-01 -5.13453424e-01 6.17937505e-01 1.21329635e-01 -6.89302862e-01 6.99678138e-02 5.23639977e-01]
[9.996642112731934, 2.298212766647339]
24855a54-eb33-4d23-aa0f-40a939006d01
model-reduction-of-consensus-network-systems
2203.14377
null
https://arxiv.org/abs/2203.14377v1
https://arxiv.org/pdf/2203.14377v1.pdf
Model Reduction of Consensus Network Systems via Selection of Optimal Edge Weights and Nodal Time-Scales
This paper proposes model reduction approaches for consensus network systems based on a given clustering of the underlying graph. Namely, given a consensus network system of time-scaled agents evolving over a weighted undirected graph and a graph clustering, a parameterized reduced consensus network system is constructed with its edge weights and nodal time-scales as the parameters to be optimized. H-infinity- and H-2-based optimization approaches are proposed to select the reduced network parameters such that the corresponding approximation errors, i.e., the H-infinity- and H-2-norms of the error system, are minimized. The effectiveness of the proposed model reduction methods is illustrated via a numerical example.
['Dany Abou Jaoude', 'Ralph Sabbagh']
2022-03-27
null
null
null
null
['graph-clustering']
['graphs']
[-1.40878424e-01 4.63975787e-01 1.45909578e-01 8.85708481e-02 -1.93365574e-01 -5.34972608e-01 2.10755527e-01 1.34995744e-01 1.33567452e-01 5.90897262e-01 -3.94719839e-01 -5.98471053e-02 -8.85894299e-01 -4.76985574e-01 -1.65646508e-01 -9.99149024e-01 -7.41879821e-01 2.19199494e-01 -1.26309022e-01 -3.80983412e-01 1.13972493e-01 3.04827929e-01 -7.08096623e-01 -8.68304431e-01 1.14569879e+00 8.24843824e-01 -2.63560176e-01 6.60655618e-01 7.29157388e-01 4.13461268e-01 -3.57278794e-01 5.24810292e-02 6.28403127e-01 -6.07384622e-01 -3.39668155e-01 8.27852249e-01 -1.48508042e-01 2.75231212e-01 -9.01591420e-01 1.45603621e+00 2.78381556e-01 4.19397652e-01 5.37671685e-01 -1.80146599e+00 -4.33847994e-01 6.30763829e-01 -3.84416580e-01 7.79533833e-02 -5.33314385e-02 2.78634876e-01 7.22579956e-01 -6.22941732e-01 6.77485764e-01 1.25686371e+00 6.94880664e-01 3.79271984e-01 -1.75285208e+00 -3.77561659e-01 3.39511007e-01 -5.40337488e-02 -1.98340976e+00 -4.46272902e-02 8.92678022e-01 -3.93128335e-01 5.58394372e-01 2.30504259e-01 1.09626222e+00 2.51248330e-01 6.73661649e-01 -2.74233729e-01 8.20915103e-01 -1.51041582e-01 5.12474179e-01 -2.70186126e-01 2.40217030e-01 7.50012636e-01 8.99668336e-01 2.21620202e-02 1.18570924e-01 -5.44749856e-01 6.77843392e-01 -1.06476070e-02 -4.63065207e-01 -6.33556008e-01 -1.15287089e+00 8.09684217e-01 5.33229828e-01 1.69755355e-01 -6.44745409e-01 2.10591793e-01 1.93379402e-01 5.54795325e-01 5.72995961e-01 2.07940444e-01 -4.28045057e-02 6.92756474e-01 -4.40667748e-01 1.59348294e-01 1.17485893e+00 1.24226761e+00 7.99164355e-01 4.81884092e-01 1.67113513e-01 2.75362164e-01 5.22766054e-01 8.17192256e-01 -3.87598097e-01 -1.28890228e+00 2.38756895e-01 9.77624536e-01 1.32169262e-01 -1.37717509e+00 -4.60857362e-01 -3.34771276e-01 -1.39792848e+00 5.46067394e-03 5.21312505e-02 -6.81606472e-01 -7.05391884e-01 1.76737010e+00 5.73437154e-01 4.25572783e-01 9.40978900e-02 1.07413530e+00 -1.25118196e-01 8.48475337e-01 -2.46272624e-01 -9.76914465e-01 6.21582031e-01 -5.88136077e-01 -9.25853431e-01 9.80584547e-02 4.22077894e-01 -3.24146420e-01 1.68688312e-01 -6.96574599e-02 -1.21799564e+00 6.32109470e-04 -1.13240528e+00 9.32905614e-01 1.46320993e-02 -3.55250180e-01 -9.06190202e-02 1.03745863e-01 -1.45615149e+00 5.45440912e-01 -7.80031145e-01 -6.94954634e-01 -4.41080242e-01 5.88698328e-01 -1.74191907e-01 1.88584745e-01 -1.17848420e+00 7.97244012e-01 3.61552298e-01 7.57487476e-01 -8.74289393e-01 -6.32393301e-01 -6.71533108e-01 -8.73165950e-02 4.29232329e-01 -7.56274939e-01 6.43595636e-01 -1.02620995e+00 -1.38591206e+00 2.59190857e-01 3.52293611e-01 -2.91963071e-01 4.08422291e-01 8.01319599e-01 -5.81585646e-01 2.54640251e-01 -2.58583855e-02 -6.60228059e-02 8.37541282e-01 -1.48233819e+00 -5.12751341e-01 -2.33278304e-01 -1.09243970e-02 2.50990868e-01 -2.66238391e-01 -4.17423576e-01 -2.13345334e-01 -4.29875880e-01 3.13131392e-01 -1.29467845e+00 -1.03003478e+00 -1.51961595e-01 -4.91993099e-01 -1.42460048e-01 1.01793420e+00 -6.81489170e-01 1.44934487e+00 -2.00482845e+00 7.86486447e-01 1.18656218e+00 5.79604924e-01 -1.10160217e-01 -3.30681294e-01 7.73783565e-01 6.02251068e-02 1.41443774e-01 -1.78189218e-01 2.19708711e-01 -1.42882437e-01 3.79216932e-02 2.75275379e-01 1.12154162e+00 -7.66197816e-02 2.70944774e-01 -9.41121578e-01 -2.82005936e-01 2.34108850e-01 5.09304404e-01 -3.05655032e-01 1.72649130e-01 4.43479382e-02 3.11043620e-01 -6.90699697e-01 9.31305289e-02 4.43013161e-01 -3.92950565e-01 7.88868248e-01 -3.05872768e-01 -2.52760470e-01 -8.27516258e-01 -1.34307814e+00 8.94003272e-01 -1.31611064e-01 3.48026603e-01 1.01080871e+00 -1.15897810e+00 1.02002943e+00 5.62397957e-01 9.91096079e-01 -6.66000247e-02 4.49774861e-01 9.60268974e-02 3.40723366e-01 -9.38718989e-02 1.57555901e-02 2.50267923e-01 -2.14964151e-01 2.43509427e-01 -2.55342335e-01 -3.97206962e-01 6.67673469e-01 6.63801372e-01 1.21775877e+00 -8.39743018e-01 3.02360088e-01 -1.05403101e+00 9.04175460e-01 7.79344216e-02 7.83971488e-01 2.09165335e-01 -2.92819560e-01 -1.59527421e-01 6.25135183e-01 -1.09847195e-01 -1.49538493e+00 -6.64044440e-01 2.62281835e-01 2.57591754e-01 5.21540403e-01 -2.56645322e-01 -8.85203719e-01 -1.98114797e-01 7.18501210e-02 1.94504350e-01 -5.68156123e-01 -4.64138836e-01 -5.51918149e-01 -2.64837027e-01 3.14045176e-02 -1.18376933e-01 2.04630569e-01 -3.52708638e-01 -1.00102484e-01 3.65419745e-01 1.47975639e-01 -8.21371436e-01 -1.08373344e+00 -1.58462390e-01 -9.04201090e-01 -1.43908989e+00 -4.35486913e-01 -1.15293002e+00 1.62452054e+00 3.50974798e-01 4.50423837e-01 4.29603636e-01 5.40219061e-02 7.90559471e-01 5.90702612e-03 1.63090691e-01 -4.11878347e-01 -6.31795973e-02 4.54031318e-01 4.96548861e-01 -5.80559075e-01 -5.79401195e-01 -5.97909510e-01 4.40252572e-01 -7.37825274e-01 -1.00747876e-01 2.73238391e-01 9.69278872e-01 5.74270725e-01 1.30612865e-01 6.81502938e-01 -3.86195481e-01 1.22106957e+00 -6.19948864e-01 -1.28357971e+00 4.84332621e-01 -1.01101887e+00 -2.08633125e-01 9.21553075e-01 -6.23635471e-01 -5.30850887e-01 1.24325350e-01 1.10836267e+00 -6.73337579e-01 6.59911156e-01 7.34372497e-01 -6.60337955e-02 -5.50496399e-01 3.00531387e-01 3.53936618e-03 4.62658376e-01 2.93165296e-01 4.89300311e-01 2.53663003e-01 2.46517152e-01 -3.75737011e-01 1.13792455e+00 3.19614321e-01 3.93620998e-01 -8.98810089e-01 -1.56219453e-01 -3.93466234e-01 -4.89864886e-01 -6.37402892e-01 5.08385122e-01 -4.77063090e-01 -1.09488499e+00 3.93812895e-01 -9.22995985e-01 -2.31831506e-01 -2.57894158e-01 2.91755170e-01 -3.58916759e-01 4.24494267e-01 -6.01819038e-01 -1.03927994e+00 -5.32695949e-01 -8.34568799e-01 2.47526273e-01 2.53310084e-01 -1.11924216e-01 -1.50305867e+00 3.98580819e-01 -5.09699523e-01 4.71400797e-01 6.56788707e-01 6.80936813e-01 -4.96021062e-01 -4.71786439e-01 -4.58362669e-01 -1.19087443e-01 2.20010132e-01 1.18164621e-01 4.44655448e-01 1.96572751e-01 -1.07096493e+00 -1.06899761e-01 1.46286353e-01 -1.05166614e-01 6.64872587e-01 3.54031473e-01 -8.92939568e-01 -6.29582167e-01 4.25528407e-01 1.64612174e+00 3.34094226e-01 -1.06798299e-01 -1.48546889e-01 6.46735907e-01 4.50685799e-01 1.72235399e-01 5.66251874e-01 3.80398065e-01 3.43364179e-01 4.93747741e-01 1.40301138e-01 3.34133297e-01 1.54629156e-01 1.05186306e-01 1.56468713e+00 -1.80625483e-01 -2.99976498e-01 -8.26714635e-01 6.61511540e-01 -2.29236126e+00 -6.61777914e-01 -1.47279859e-01 2.17914057e+00 4.31881577e-01 -2.09976599e-01 6.86622486e-02 -4.54292968e-02 1.57451379e+00 9.58832446e-03 -9.05659616e-01 -1.38566390e-01 -5.16140424e-02 -5.37415504e-01 7.36284375e-01 9.51151311e-01 -6.51011050e-01 2.96537012e-01 6.64377117e+00 2.45088786e-01 -1.03357255e+00 -1.92330673e-01 2.97671229e-01 1.89680383e-01 -1.62066996e-01 3.75812769e-01 -8.88789445e-02 3.70739907e-01 1.13315856e+00 -1.32524800e+00 8.94262910e-01 6.67524219e-01 7.20112503e-01 3.78079355e-01 -7.99078465e-01 8.32177281e-01 -1.19597882e-01 -1.14819562e+00 -3.76796603e-01 3.05523038e-01 1.27926123e+00 -2.71314561e-01 -1.96097493e-01 -7.13395849e-02 6.37672722e-01 -4.94926721e-01 4.82261479e-01 5.22686422e-01 5.14008462e-01 -7.09622025e-01 4.87823486e-01 -1.52973756e-02 -1.64990675e+00 -1.72494963e-01 -1.26593217e-01 9.87101793e-02 4.07137901e-01 4.95470285e-01 -2.55260319e-01 6.04565740e-01 2.28953630e-01 9.08618033e-01 -2.05225442e-02 1.27076364e+00 1.48944348e-01 4.00055885e-01 -5.56719542e-01 -2.39260867e-01 2.51538604e-01 -9.60089207e-01 9.58411574e-01 6.96442127e-01 2.66486108e-01 4.97767597e-01 7.46294260e-01 5.19892931e-01 -1.02048263e-01 -1.12531841e-01 -5.30393362e-01 9.04416591e-02 9.30240452e-01 1.57192791e+00 -8.65788102e-01 -2.55921036e-01 -1.23846956e-01 4.70472217e-01 2.48160750e-01 9.84231472e-01 -7.33459115e-01 -5.46224296e-01 7.95613885e-01 -7.21730515e-02 5.14783636e-02 -3.14395636e-01 1.03378281e-01 -1.03846884e+00 -1.84320480e-01 -7.60549545e-01 5.37179470e-01 -3.14854383e-01 -1.39514339e+00 5.83350301e-01 -6.59664869e-02 -1.29469430e+00 -1.53809533e-01 -3.96246184e-03 -8.39536786e-01 4.79404807e-01 -6.49455547e-01 -7.77033031e-01 -2.47061998e-01 7.70234644e-01 6.31570891e-02 5.45821153e-02 3.95011604e-01 6.96466938e-02 -1.07142556e+00 7.62356222e-02 5.10588646e-01 -1.61744766e-02 6.54993802e-02 -1.07498074e+00 -7.28688911e-02 1.11680162e+00 -6.74831092e-01 4.99876410e-01 9.94620860e-01 -6.05456233e-01 -2.02556729e+00 -1.36276090e+00 3.78607422e-01 2.18726680e-01 1.33226550e+00 -2.94459797e-02 -6.43377125e-01 9.43641603e-01 5.14533460e-01 2.65172362e-01 -1.26573086e-01 -4.80548829e-01 2.72068735e-02 -4.50273246e-01 -1.31830645e+00 8.02671552e-01 8.26429486e-01 -1.60967946e-01 5.49318045e-02 4.02925104e-01 6.01577044e-01 -1.80430159e-01 -1.37016511e+00 3.23060453e-01 9.91270021e-02 2.96756998e-02 5.76295316e-01 -5.23800135e-01 -5.09519815e-01 -5.49036205e-01 1.50242150e-01 -1.89168501e+00 -7.83470750e-01 -1.21265948e+00 -2.78971583e-01 7.75658190e-01 3.69489461e-01 -8.53298843e-01 3.87304008e-01 5.10259509e-01 6.73821419e-02 -5.09962499e-01 -1.00414205e+00 -8.89954805e-01 -2.90778250e-01 2.99813569e-01 1.30854964e-01 9.99832511e-01 6.93263113e-01 4.59209472e-01 -2.27172628e-01 6.90954089e-01 1.32202470e+00 -2.39948392e-01 4.38028961e-01 -1.26474500e+00 4.12884623e-01 -4.54910725e-01 -4.15444940e-01 -2.77696997e-01 3.24509859e-01 -5.39381921e-01 2.99912512e-01 -1.62819886e+00 -7.35655501e-02 -3.85514230e-01 -1.62570208e-01 -2.87786275e-02 3.82719934e-02 -1.15428545e-01 3.43793392e-01 3.05584222e-01 -8.50390971e-01 4.57777202e-01 1.13018835e+00 -1.99191257e-01 -3.50563854e-01 -8.45648944e-02 -9.49192792e-02 4.35160935e-01 7.47497320e-01 -2.74501890e-01 -7.61567116e-01 2.90321093e-02 -2.56926883e-02 7.43388891e-01 1.71786696e-01 -6.24309301e-01 5.66283643e-01 -5.24329960e-01 -5.41202426e-01 -3.14691842e-01 5.64451143e-02 -1.22761726e+00 8.71895373e-01 1.03545153e+00 -4.16325599e-01 4.00986969e-01 -3.24332565e-01 1.00603974e+00 -1.12725481e-01 3.27741832e-01 7.44733453e-01 4.13861215e-01 -2.85169601e-01 6.34188473e-01 -7.00935841e-01 -1.64113399e-02 1.40851033e+00 -6.16598316e-03 -3.79839689e-01 -7.78796375e-01 -8.79816830e-01 9.47713733e-01 4.27379698e-01 6.88925311e-02 4.80179638e-01 -1.31898403e+00 -7.84795582e-01 -4.29903567e-02 -9.81126279e-02 -2.02101409e-01 2.13443905e-01 1.13994300e+00 -4.43764418e-01 3.52391511e-01 -1.15545876e-01 -5.23168325e-01 -9.87023354e-01 5.67852914e-01 7.67130673e-01 -2.36286566e-01 -3.59563828e-01 7.95809701e-02 -1.95644319e-01 -4.05559301e-01 2.44517684e-01 -2.54667729e-01 2.83699155e-01 -4.05246764e-01 7.60122389e-02 9.31464612e-01 -1.63269415e-01 -5.78339636e-01 -2.83177555e-01 5.15643120e-01 4.81332332e-01 -8.27212706e-02 1.15742886e+00 -6.42207205e-01 -7.41721570e-01 7.35211596e-02 9.72691953e-01 -6.11327350e-01 -1.49141848e+00 -8.40299651e-02 -3.79105397e-02 5.14797084e-02 5.80130853e-02 -1.01231016e-01 -1.61527467e+00 -1.25457585e-01 2.95340925e-01 9.95166898e-01 1.01937664e+00 -2.76810467e-01 1.54183641e-01 3.40836197e-01 4.34557706e-01 -1.18337393e+00 -1.89830557e-01 5.71431577e-01 7.52427399e-01 -6.07055545e-01 -2.97810659e-02 -5.54237545e-01 -2.09683940e-01 1.05100608e+00 9.07335818e-01 -7.19014406e-01 1.06497717e+00 1.92669481e-01 -1.46891043e-01 -1.96761012e-01 -1.02815330e+00 1.33243874e-01 1.25954330e-01 6.13818645e-01 7.39074051e-02 2.62548596e-01 -8.59677672e-01 9.18223783e-02 3.92251313e-01 -4.09344018e-01 8.69563103e-01 6.70212448e-01 -3.18581671e-01 -4.65896964e-01 -5.94827950e-01 3.49260241e-01 3.20408911e-01 5.09488583e-01 -3.09298813e-01 8.19170117e-01 -5.38662016e-01 1.20807815e+00 -3.83603871e-02 -4.01245296e-01 6.56299293e-01 -4.36233252e-01 1.88662261e-01 -3.13342005e-01 -3.50668043e-01 9.64214802e-02 -4.78152148e-02 -3.88678551e-01 -4.37615246e-01 -2.87488163e-01 -1.24324954e+00 -9.72968638e-01 -7.00580597e-01 6.72383547e-01 3.54192734e-01 6.91130579e-01 5.29757917e-01 3.27897578e-01 1.23803127e+00 -9.21798050e-01 -8.52246046e-01 -8.59701037e-01 -8.78542662e-01 2.32685581e-01 2.23667681e-01 -2.51176834e-01 -8.57909620e-01 -3.28778960e-02]
[5.176692008972168, 2.697511672973633]
ba106849-1899-44ee-bfd0-8663aff1fe0c
how-well-do-self-supervised-methods-perform
2202.09014
null
https://arxiv.org/abs/2202.09014v1
https://arxiv.org/pdf/2202.09014v1.pdf
How Well Do Self-Supervised Methods Perform in Cross-Domain Few-Shot Learning?
Cross-domain few-shot learning (CDFSL) remains a largely unsolved problem in the area of computer vision, while self-supervised learning presents a promising solution. Both learning methods attempt to alleviate the dependency of deep networks on the requirement of large-scale labeled data. Although self-supervised methods have recently advanced dramatically, their utility on CDFSL is relatively unexplored. In this paper, we investigate the role of self-supervised representation learning in the context of CDFSL via a thorough evaluation of existing methods. It comes as a surprise that even with shallow architectures or small training datasets, self-supervised methods can perform favorably compared to the existing SOTA methods. Nevertheless, no single self-supervised approach dominates all datasets indicating that existing self-supervised methods are not universally applicable. In addition, we find that representations extracted from self-supervised methods exhibit stronger robustness than the supervised method. Intriguingly, whether self-supervised representations perform well on the source domain has little correlation with their applicability on the target domain. As part of our study, we conduct an objective measurement of the performance for six kinds of representative classifiers. The results suggest Prototypical Classifier as the standard evaluation recipe for CDFSL.
['Jun Wang', 'Xiaogang Xu', 'Ying Zheng', 'Yiyi Zhang']
2022-02-18
null
null
null
null
['cross-domain-few-shot', 'cross-domain-few-shot-learning']
['computer-vision', 'computer-vision']
[ 7.78529886e-03 3.44438702e-02 -5.01120627e-01 -6.72999859e-01 -8.72096181e-01 -4.68604475e-01 9.20729637e-01 1.66226208e-01 -4.22914505e-01 7.89040983e-01 1.98038548e-01 -2.59548016e-02 -2.58918196e-01 -7.03110814e-01 -5.14236867e-01 -6.79036736e-01 1.33396819e-01 4.62357938e-01 3.26196551e-01 -3.62564415e-01 7.74810240e-02 2.58956879e-01 -1.70384610e+00 2.36833557e-01 6.54929698e-01 1.07664990e+00 -1.22327283e-01 1.06561117e-01 -4.05855417e-01 9.25852478e-01 -6.10458553e-01 -2.92765290e-01 1.52785748e-01 -5.07952273e-01 -7.12357104e-01 -3.59356659e-03 6.33109391e-01 9.45377126e-02 -2.92928249e-01 9.47743595e-01 6.01614118e-01 9.03886929e-02 9.02738869e-01 -1.38955629e+00 -7.51475751e-01 5.93486905e-01 -3.30060035e-01 3.95561129e-01 -4.27105278e-02 1.61285937e-01 1.26851332e+00 -8.82716119e-01 7.82491148e-01 9.58219290e-01 8.61916125e-01 5.97397685e-01 -1.28918171e+00 -6.11364007e-01 -1.64079163e-02 2.23863989e-01 -1.17019665e+00 -6.52316630e-01 8.51183116e-01 -3.58673215e-01 7.62187421e-01 -2.08170846e-01 3.44366312e-01 1.47180474e+00 1.16834626e-01 8.94628763e-01 1.19757473e+00 -5.63983262e-01 7.47383058e-01 4.62866515e-01 5.61581612e-01 4.27308202e-01 3.23206127e-01 2.83031017e-01 -6.87697172e-01 -1.48201957e-01 1.30567208e-01 -6.19808212e-02 1.22764997e-01 -8.45320940e-01 -9.53396499e-01 1.12779856e+00 3.62933308e-01 6.80325806e-01 2.99721789e-02 -1.18826225e-01 7.76506066e-01 4.60557789e-01 8.25557768e-01 6.48517311e-01 -5.07701874e-01 -4.90304045e-02 -1.04576147e+00 1.95999458e-01 6.94646895e-01 9.04699266e-01 8.85103822e-01 3.09136212e-01 2.46234704e-02 9.71432924e-01 -5.79569153e-02 1.57591537e-01 7.75299966e-01 -5.58048904e-01 7.43809193e-02 6.42955959e-01 -2.34657824e-01 -7.47917712e-01 -3.57450634e-01 -6.45090997e-01 -6.43198133e-01 2.08119050e-01 5.74689806e-01 -1.15099639e-01 -6.75580621e-01 1.69915211e+00 5.41170277e-02 -7.61056244e-02 3.60278547e-01 6.39588892e-01 9.37011242e-01 4.76563305e-01 2.37855360e-01 -2.14225024e-01 8.83952200e-01 -9.89597082e-01 -4.71251428e-01 -4.54582304e-01 7.19505429e-01 -4.69686121e-01 1.01564324e+00 2.82142371e-01 -5.73712528e-01 -5.94809651e-01 -1.13862026e+00 1.89087823e-01 -7.01091766e-01 4.20136079e-02 9.11072254e-01 7.63422847e-01 -7.10973084e-01 8.00806165e-01 -3.76405746e-01 -8.24241996e-01 6.42042398e-01 -4.52980474e-02 -4.86765623e-01 -2.35224158e-01 -1.20902467e+00 1.10614491e+00 4.13299382e-01 -4.90273863e-01 -9.05539453e-01 -7.14110255e-01 -8.56211841e-01 4.25263867e-02 3.33150923e-01 -9.48317647e-02 1.30634630e+00 -1.20520473e+00 -1.13791656e+00 1.18670678e+00 5.02386615e-02 -7.97269344e-01 3.71475816e-01 -5.86611964e-02 -5.63028097e-01 -7.79697951e-03 2.79121876e-01 6.88525200e-01 8.85568976e-01 -1.18176866e+00 -6.00380003e-01 -2.71238625e-01 -1.40311569e-01 -6.42994642e-02 -5.60318112e-01 -8.94796550e-02 8.62465501e-02 -5.45045614e-01 -8.29942301e-02 -6.86858714e-01 -3.67842428e-02 3.04428348e-03 -1.48596585e-01 -5.74557781e-01 9.79235947e-01 1.50514960e-01 1.03486764e+00 -2.47961283e+00 -3.56172591e-01 -2.71606386e-01 1.32002637e-01 4.82139617e-01 -1.62208438e-01 6.35623395e-01 -3.86420369e-01 -1.12148203e-01 -3.97922724e-01 -6.85115233e-02 5.38317077e-02 2.75066406e-01 -4.17061388e-01 6.20330811e-01 3.91916364e-01 8.88924420e-01 -1.05208123e+00 -4.48407114e-01 1.06074765e-01 1.52460054e-01 -1.67729005e-01 8.14592168e-02 -2.97454715e-01 1.13441274e-01 -2.77348846e-01 6.14758909e-01 4.92688000e-01 -3.57767045e-01 1.90405950e-01 -1.18161961e-01 -7.41652176e-02 1.51864797e-01 -8.93613398e-01 1.74694395e+00 -1.21349856e-01 9.20167744e-01 -4.39163595e-01 -1.54238153e+00 1.23943186e+00 1.80641651e-01 5.81586003e-01 -8.04131329e-01 2.05905586e-01 2.93694407e-01 1.68237478e-01 -3.04852217e-01 1.75823554e-01 -4.98620182e-01 -1.60185128e-01 6.04784548e-01 6.03715241e-01 2.50162989e-01 3.08163315e-01 2.49729514e-01 1.18529046e+00 1.95676461e-01 6.19344950e-01 -4.93087173e-01 1.03988022e-01 5.09896696e-01 5.48607171e-01 7.45938838e-01 -5.50163090e-01 6.12216651e-01 4.26150978e-01 -6.73310459e-01 -9.89638805e-01 -1.06370711e+00 -4.08915937e-01 1.30777669e+00 -5.84553257e-02 -3.53956372e-01 -5.83010554e-01 -1.09675300e+00 1.51985645e-01 7.45270431e-01 -6.75997257e-01 -3.94404262e-01 -9.74177644e-02 -9.01902378e-01 6.50543272e-01 6.40925050e-01 4.84460115e-01 -9.49974358e-01 -7.04716921e-01 1.71101198e-01 4.10474241e-01 -1.03519773e+00 6.11558743e-02 6.90611541e-01 -8.03868175e-01 -1.23877633e+00 -8.24033499e-01 -8.47370207e-01 3.51150066e-01 6.10183001e-01 1.29444373e+00 -3.34475189e-01 -2.98173517e-01 4.93211508e-01 -4.79021907e-01 -5.01430273e-01 -4.26760674e-01 2.48406306e-01 1.78595677e-01 -3.80544923e-02 8.56510580e-01 -5.95862269e-01 -3.33506107e-01 3.01175535e-01 -6.21788383e-01 -4.52437252e-01 4.46462750e-01 9.53563452e-01 2.31344163e-01 4.12647091e-02 1.12765193e+00 -1.32223022e+00 5.67913830e-01 -6.96934104e-01 -2.90197879e-01 2.10799068e-01 -9.09371138e-01 8.26557007e-05 6.59638882e-01 -3.74153733e-01 -1.18686903e+00 1.35663420e-01 6.81286231e-02 -3.50466400e-01 -4.75635558e-01 4.57200527e-01 1.14480378e-02 3.65854166e-02 1.25141978e+00 1.40380263e-01 9.97592956e-02 -4.72820193e-01 3.69201571e-01 6.51934803e-01 3.36370409e-01 -4.23413008e-01 6.77753270e-01 5.05936205e-01 -2.77304053e-01 -9.86742198e-01 -1.38692403e+00 -5.66572428e-01 -7.67040968e-01 -1.42378449e-01 5.96330762e-01 -9.46745217e-01 1.09986903e-03 2.06238449e-01 -7.77493000e-01 -2.39904448e-01 -7.13236630e-01 2.79146612e-01 -5.35440743e-01 2.87024111e-01 -2.42447466e-01 -7.33722508e-01 -5.24385870e-02 -6.77875161e-01 7.11857378e-01 1.84974924e-01 -4.52979714e-01 -1.16589928e+00 1.84821323e-01 2.59311795e-01 4.60709274e-01 1.19431943e-01 8.90755713e-01 -1.28517079e+00 -1.23112038e-01 -4.64319289e-01 -2.46487617e-01 5.35406113e-01 2.44521812e-01 -1.86292171e-01 -1.39494717e+00 -2.38214716e-01 -8.88109282e-02 -8.46069813e-01 1.05895376e+00 1.73715785e-01 8.41325939e-01 2.41636306e-01 -1.71892181e-01 2.16325283e-01 1.39033675e+00 5.68413176e-02 4.22432482e-01 3.56393695e-01 3.08738679e-01 7.18554020e-01 7.49574840e-01 2.11447284e-01 1.96280643e-01 3.37470174e-01 1.16222322e-01 3.40674520e-02 -2.36680835e-01 -2.90839195e-01 2.77209878e-01 6.16683304e-01 3.62427145e-01 -7.45865479e-02 -1.18772292e+00 5.79025447e-01 -1.86253154e+00 -1.03031123e+00 2.28929915e-03 2.14449811e+00 6.96057498e-01 4.06623751e-01 1.99697122e-01 3.27607840e-01 6.61181688e-01 3.64544451e-01 -6.91698253e-01 -1.69595450e-01 -3.65155756e-01 1.93314418e-01 2.61989743e-01 -1.40253320e-01 -1.39174604e+00 9.25414383e-01 7.23878813e+00 9.92429137e-01 -1.05336308e+00 3.08685273e-01 6.61406696e-01 -2.36040168e-02 1.20731173e-02 6.92785978e-02 -9.19303834e-01 3.82454634e-01 9.58701193e-01 -2.88039088e-01 -1.68122843e-01 1.34176576e+00 -1.61038145e-01 -2.86554962e-01 -1.33869624e+00 9.30940747e-01 2.35089496e-01 -1.43402421e+00 -2.63418794e-01 -1.52211323e-01 8.54076266e-01 2.35698804e-01 -9.71085355e-02 7.22371161e-01 3.55999053e-01 -7.91364074e-01 4.06466782e-01 1.54727235e-01 7.09912419e-01 -6.58724666e-01 5.74017167e-01 4.31977391e-01 -7.77598858e-01 -2.02452138e-01 -6.45588696e-01 -2.06212178e-01 -1.92288935e-01 7.09492207e-01 -6.41245782e-01 2.07041278e-01 5.50198913e-01 1.00532043e+00 -7.82884538e-01 9.48481739e-01 -7.79664218e-02 8.55726838e-01 6.70424700e-02 -1.73912421e-02 5.27361095e-01 1.50868446e-01 3.32508385e-01 1.26841998e+00 1.03268796e-03 -3.42342556e-01 1.73574001e-01 6.87254846e-01 4.27638181e-02 1.57149673e-01 -1.08113182e+00 -4.95840251e-01 2.20012799e-01 1.11638975e+00 -7.62252092e-01 -3.83727163e-01 -6.96960151e-01 6.87339246e-01 6.17348850e-01 2.78517932e-01 -5.06279647e-01 -1.88604280e-01 5.43116271e-01 7.69728348e-02 1.41356006e-01 -5.50642610e-02 -4.06680822e-01 -1.25778162e+00 -2.01908678e-01 -7.42222905e-01 5.74368536e-01 -5.05845785e-01 -1.95618618e+00 4.62046236e-01 -1.56492978e-01 -1.44296551e+00 -3.28353822e-01 -6.48177624e-01 -7.35765874e-01 3.16882044e-01 -1.55610991e+00 -9.65328157e-01 -2.36613289e-01 4.93606627e-01 7.97339976e-01 -8.20696115e-01 9.96551394e-01 2.13827163e-01 -5.08817852e-01 5.55454612e-01 4.43010896e-01 1.52695864e-01 1.05710983e+00 -1.12887406e+00 2.96093643e-01 6.01632178e-01 4.37741756e-01 5.38888931e-01 6.97412491e-01 -5.52222967e-01 -1.22152174e+00 -9.91237879e-01 7.62766540e-01 -2.75998175e-01 7.99309254e-01 -3.31166208e-01 -9.70574439e-01 4.71268147e-01 2.59803087e-01 4.58623409e-01 8.57090771e-01 3.86559784e-01 -7.39952683e-01 -2.90929586e-01 -1.04795623e+00 2.30936781e-01 1.07110822e+00 -5.97106576e-01 -7.71544635e-01 4.24814522e-01 4.05883104e-01 2.02499151e-01 -6.47187769e-01 3.01837802e-01 3.71183634e-01 -1.16455424e+00 7.54345000e-01 -8.51966083e-01 6.63242400e-01 1.18706927e-01 -3.21807414e-01 -1.46383238e+00 -2.73993611e-01 -1.86568961e-01 -4.50719669e-02 1.43431044e+00 4.29647863e-01 -5.96010149e-01 1.01331639e+00 3.64133030e-01 -6.78468822e-03 -6.33557081e-01 -9.15972650e-01 -1.26378834e+00 3.97907197e-01 -4.51285928e-01 4.01759371e-02 1.38083315e+00 2.59022266e-01 6.83066547e-01 -3.37940574e-01 -5.08073986e-01 8.51371706e-01 1.47452682e-01 7.48100400e-01 -1.56599402e+00 -8.75678286e-02 -2.90144295e-01 -5.94191611e-01 -4.34952050e-01 5.52330852e-01 -1.06606424e+00 -1.85088944e-02 -1.36905742e+00 3.97609919e-01 -5.07020235e-01 -4.20464784e-01 4.78597194e-01 1.13257930e-01 2.32930467e-01 1.20395310e-01 1.95607364e-01 -7.76888430e-01 5.81609428e-01 7.41759598e-01 -2.93661326e-01 3.37608233e-02 -3.48026343e-02 -8.06587636e-01 7.75383234e-01 9.56345916e-01 -4.95331764e-01 -6.93969071e-01 -1.20552070e-01 -5.21736778e-02 -3.11508864e-01 2.25200236e-01 -1.24480867e+00 2.02933177e-01 -6.70509487e-02 3.36808443e-01 -3.00998300e-01 1.08477011e-01 -6.45852268e-01 -3.40225339e-01 1.67223364e-01 -4.56008345e-01 -3.52604955e-01 3.09800617e-02 7.88972199e-01 -4.45298374e-01 -5.60512364e-01 1.13929224e+00 -3.99181545e-01 -1.14807904e+00 1.06118895e-01 -4.40717041e-01 4.84968901e-01 1.12255228e+00 -2.27315232e-01 -3.87252361e-01 -3.03728551e-01 -7.43995190e-01 -3.13364826e-02 3.49352717e-01 5.57216227e-01 5.03668904e-01 -1.31373906e+00 -4.86575902e-01 7.32022673e-02 6.33863151e-01 -4.12635267e-01 3.88647504e-02 5.17624021e-01 6.95087835e-02 4.70189363e-01 -3.53792548e-01 -6.30431831e-01 -8.53312612e-01 7.11470246e-01 2.18275264e-01 -1.27059326e-01 -6.67665303e-01 6.16202772e-01 1.97977260e-01 -4.42497075e-01 4.14278388e-01 3.15571547e-01 -1.98273242e-01 5.91122210e-01 4.66735005e-01 3.26642364e-01 1.56882674e-01 -5.09381473e-01 -2.64179498e-01 1.51875764e-01 -3.98562849e-01 1.85561165e-01 1.57954264e+00 1.11371942e-01 3.90141368e-01 8.88536215e-01 1.31729639e+00 -5.79947293e-01 -1.20558393e+00 -5.50200284e-01 4.63682115e-01 -1.76305786e-01 3.46116461e-02 -6.46834612e-01 -9.68313456e-01 1.02125227e+00 6.39584243e-01 1.93197146e-01 8.60317767e-01 1.37001887e-01 4.50879812e-01 6.08615696e-01 4.58938360e-01 -1.33468258e+00 3.62672150e-01 6.47447526e-01 5.42352796e-01 -1.69040573e+00 1.37212142e-01 -2.04268754e-01 -8.65484297e-01 1.11663210e+00 7.33088136e-01 -3.52902323e-01 6.95405662e-01 1.37578264e-01 7.96981454e-02 -2.38576233e-01 -8.76520693e-01 -5.04952013e-01 1.01284280e-01 8.04989219e-01 5.88712752e-01 -2.92129755e-01 -3.28102380e-01 6.47230089e-01 1.03781134e-01 4.75286730e-02 4.39213067e-01 1.05184317e+00 -6.36993289e-01 -1.02132559e+00 -5.43749444e-02 5.46387792e-01 -2.08765313e-01 3.79061364e-02 -4.81284410e-01 8.08336318e-01 9.03339460e-02 8.55782866e-01 3.53755169e-02 -2.48821571e-01 2.45544523e-01 5.26020288e-01 2.49703541e-01 -8.99944723e-01 -5.37837267e-01 -2.16502354e-01 2.40768164e-01 -3.23084921e-01 -5.56160271e-01 -7.04276681e-01 -9.23197448e-01 -3.21386755e-02 -2.20778599e-01 4.93869260e-02 4.49463159e-01 1.10206735e+00 1.98752597e-01 1.68820739e-01 6.30082846e-01 -5.87306321e-01 -8.36788714e-01 -8.53179693e-01 -8.04730833e-01 5.47974288e-01 5.59875071e-02 -9.56098378e-01 -3.49060386e-01 -1.28529981e-01]
[9.931191444396973, 2.9676365852355957]
f6af1f0a-554e-48ba-86cf-edeeaf31e771
headline-generation-learning-from-decomposed
1904.08455
null
https://arxiv.org/abs/1904.08455v3
https://arxiv.org/pdf/1904.08455v3.pdf
Headline Generation: Learning from Decomposable Document Titles
We propose a novel method for generating titles for unstructured text documents. We reframe the problem as a sequential question-answering task. A deep neural network is trained on document-title pairs with decomposable titles, meaning that the vocabulary of the title is a subset of the vocabulary of the document. To train the model we use a corpus of millions of publicly available document-title pairs: news articles and headlines. We present the results of a randomized double-blind trial in which subjects were unaware of which titles were human or machine-generated. When trained on approximately 1.5 million news articles, the model generates headlines that humans judge to be as good or better than the original human-written headlines in the majority of cases.
['Oleg Vasilyev', 'John Bohannon', 'Tom Grek']
2019-04-17
null
null
null
null
['headline-generation']
['natural-language-processing']
[ 1.86399326e-01 5.66442847e-01 -3.37314129e-01 -4.98766452e-01 -1.05694020e+00 -6.40935838e-01 1.04595983e+00 2.92806298e-01 -6.61334157e-01 9.64078009e-01 9.05260324e-01 -1.79098800e-01 1.67952761e-01 -8.64356041e-01 -1.06512010e+00 -2.05193162e-02 2.87565231e-01 8.91283631e-01 -1.67261641e-02 -2.77780563e-01 3.16671580e-01 -2.45032459e-01 -1.35931945e+00 6.31591976e-01 5.80896318e-01 8.54389906e-01 -8.05940405e-02 1.02100062e+00 -7.02590719e-02 1.05489564e+00 -1.09389031e+00 -7.61769891e-01 1.79965764e-01 -6.77733779e-01 -1.12781286e+00 2.29365364e-01 1.00536478e+00 -8.43154013e-01 -6.00364327e-01 8.48951757e-01 2.47388020e-01 -1.31404251e-01 7.72196770e-01 -1.00644672e+00 -1.02562845e+00 9.11544919e-01 -2.48633578e-01 2.71276116e-01 6.16375268e-01 -1.58814386e-01 1.59948027e+00 -7.54275918e-01 9.59990263e-01 9.65094388e-01 3.84963393e-01 5.75914800e-01 -1.16074443e+00 -2.89020449e-01 4.20637392e-02 -2.86995113e-01 -1.02025032e+00 -5.28100610e-01 2.60339439e-01 -6.69903398e-01 6.62667632e-01 3.34247500e-01 3.20107579e-01 1.60451400e+00 5.07872820e-01 7.42359281e-01 6.89871609e-01 -3.69055212e-01 2.46221781e-01 3.06138247e-01 3.98901522e-01 2.39383802e-01 4.42501694e-01 -3.70842814e-02 -6.27458572e-01 -3.07935297e-01 3.37163061e-01 -3.50525320e-01 -4.32891548e-01 -4.92677325e-03 -1.46510029e+00 1.14220273e+00 3.37526441e-01 1.32466748e-01 -5.55143654e-01 2.34190114e-02 4.25407708e-01 3.80052924e-01 6.26498580e-01 9.51300204e-01 -2.85016090e-01 2.80396521e-01 -1.18150401e+00 9.26507592e-01 1.19720078e+00 1.06264091e+00 2.31303517e-02 -3.80142987e-01 -5.38175583e-01 4.68240708e-01 1.09147877e-01 7.63194621e-01 7.22651422e-01 -6.84461892e-01 8.28774333e-01 2.82663941e-01 6.84500873e-01 -1.10985804e+00 -2.74239749e-01 -5.69819629e-01 -7.28028476e-01 -4.73803997e-01 2.62078762e-01 -3.32894444e-01 -7.71937788e-01 1.30353415e+00 -7.83381984e-02 -7.70029247e-01 -1.91213619e-02 7.54822373e-01 1.18991935e+00 9.37330246e-01 -1.70966223e-01 -2.08278552e-01 1.57569194e+00 -1.05422747e+00 -1.00583375e+00 -3.45651180e-01 3.50477099e-01 -7.19844282e-01 9.27185655e-01 3.01398307e-01 -1.21946466e+00 -5.94853461e-01 -1.11167479e+00 -1.71244413e-01 -3.14713538e-01 1.24337681e-01 -5.90767264e-02 6.46730512e-02 -1.04067826e+00 4.54661995e-01 3.70260552e-02 -1.00856014e-01 3.61049175e-01 -8.17386284e-02 -2.98953503e-01 1.22865587e-01 -1.55099583e+00 6.50036216e-01 1.78646848e-01 -3.16220939e-01 -1.15761101e+00 -5.82356393e-01 -8.06327999e-01 1.62079558e-01 1.17961556e-01 -9.23792362e-01 2.12784219e+00 -9.77366865e-01 -7.95779586e-01 1.26025641e+00 -3.08364809e-01 -8.72747362e-01 7.33231306e-01 -5.73940456e-01 -4.51075256e-01 4.92649615e-01 5.79219639e-01 5.69738925e-01 1.02202618e+00 -1.02407205e+00 -7.15439379e-01 -1.03073210e-01 9.99476016e-02 1.40955344e-01 -2.74536759e-01 1.30062222e-01 -2.28080347e-01 -9.68230069e-01 -4.30619895e-01 -6.49762869e-01 -4.32238467e-02 -2.19527140e-01 -7.88160324e-01 -3.67180914e-01 2.04142317e-01 -1.00349653e+00 1.33603454e+00 -1.94388366e+00 -2.02691168e-01 -1.78740621e-01 6.21843219e-01 -1.55490085e-01 -3.46982479e-01 6.95489764e-01 4.13400754e-02 4.73067015e-01 1.53391808e-01 -2.09943995e-01 1.03702746e-01 -3.74959111e-01 -1.14364791e+00 2.86206871e-01 5.37964329e-02 9.53151882e-01 -8.85848105e-01 -2.90729612e-01 -6.07735336e-01 -9.34024453e-02 -3.20260167e-01 5.45849860e-01 -8.71531427e-01 -1.45527357e-02 -4.40546662e-01 -6.71784207e-02 2.38123968e-01 -7.33455241e-01 -3.23424190e-01 6.90965354e-02 1.86101928e-01 9.04337287e-01 -4.82483774e-01 9.28463757e-01 -2.50269055e-01 1.04141319e+00 -1.28962532e-01 -3.37196201e-01 6.70050502e-01 6.95714056e-01 -6.78305700e-02 -9.46711063e-01 1.39741212e-01 7.96284303e-02 -4.70568329e-01 -7.20343769e-01 7.86991060e-01 -1.42686039e-01 -2.83752382e-01 8.49269927e-01 -2.02528499e-02 -1.78937018e-01 6.42688870e-01 6.41887724e-01 1.16751623e+00 -3.90835226e-01 1.78499430e-01 1.06729809e-02 1.49183184e-01 2.90186435e-01 5.87675795e-02 1.38233161e+00 1.13176286e-01 8.77487063e-01 8.14525723e-01 -4.93689030e-01 -1.55919302e+00 -9.41558838e-01 3.24221440e-02 1.18033624e+00 -2.34521374e-01 -5.41346252e-01 -9.04293656e-01 -7.73484945e-01 -1.50746763e-01 9.64249671e-01 -8.59501541e-01 1.62133649e-01 -3.55958849e-01 -1.17419012e-01 2.33905718e-01 2.59677261e-01 4.46493477e-01 -1.15578282e+00 -4.64804590e-01 1.68655500e-01 -4.45356369e-01 -1.16324008e+00 -8.54127407e-01 -2.62934238e-01 -5.52822948e-01 -9.15955424e-01 -1.16189742e+00 -1.00566459e+00 6.05022311e-01 8.66578966e-02 1.60145295e+00 1.31939560e-01 2.41333216e-01 2.01195642e-01 -3.95084381e-01 -4.79195088e-01 -7.44654953e-01 3.63417864e-01 -3.26856151e-02 -1.29473075e-01 3.43462348e-01 7.12256134e-02 -6.46801651e-01 -1.01914434e-02 -1.12152398e+00 1.51408389e-01 4.03921455e-01 8.89026761e-01 1.47918001e-01 -6.54256418e-02 6.97999418e-01 -1.27970326e+00 1.28616917e+00 -7.01688826e-01 -6.29096746e-01 3.10429096e-01 -5.06744146e-01 6.99293539e-02 8.19711566e-01 -3.50482881e-01 -6.67963922e-01 -3.27884614e-01 -1.07466489e-01 7.22946748e-02 -1.35117099e-01 7.07934737e-01 4.18277055e-01 7.59865761e-01 1.04956162e+00 1.84545562e-01 -2.19001085e-01 -4.10096228e-01 3.50665808e-01 9.27153111e-01 6.16054714e-01 -4.71155532e-02 8.24240267e-01 1.48689702e-01 -6.91782534e-01 -5.80170095e-01 -1.48722303e+00 -4.23874617e-01 -1.62742332e-01 -1.19422227e-01 9.06609237e-01 -9.92567122e-01 -3.88257116e-01 3.92505467e-01 -1.51399827e+00 -1.99157611e-01 -3.67203832e-01 1.99754983e-01 -2.72558957e-01 -3.53968665e-02 -6.33386374e-01 -2.51176924e-01 -7.39159167e-01 -7.24547863e-01 1.15285099e+00 1.84667706e-01 -7.64325142e-01 -7.87541091e-01 3.16814274e-01 5.45301199e-01 3.93441468e-01 -9.95166413e-03 9.35221672e-01 -1.32167721e+00 -1.62206665e-01 -8.88758779e-01 -1.56299710e-01 3.50074530e-01 -6.11281097e-02 -4.30148914e-02 -8.13516319e-01 -1.55973226e-01 1.34886116e-01 -6.90267384e-01 7.24566698e-01 1.58354566e-01 8.87585163e-01 -1.18397260e+00 -2.13079333e-01 -2.07804754e-01 1.01710427e+00 5.04671745e-02 5.00670552e-01 4.41165358e-01 3.08407724e-01 7.57378221e-01 2.52110004e-01 4.82520908e-01 4.74556923e-01 3.04859728e-01 2.86310464e-02 -1.91541724e-02 6.89701065e-02 -7.35693991e-01 2.89455503e-01 5.76341331e-01 8.39385569e-01 -9.61399496e-01 -9.21296358e-01 8.29577327e-01 -1.34181118e+00 -1.17945933e+00 -1.13431457e-02 1.94488430e+00 1.18515408e+00 6.55131578e-01 1.92460552e-01 -1.07522964e-01 8.06358159e-01 4.72012848e-01 -2.58248091e-01 -3.46262664e-01 -1.21963173e-01 5.80136254e-02 2.87324816e-01 5.63250899e-01 -1.21820390e+00 5.02689004e-01 7.78542614e+00 4.06242877e-01 -8.45264792e-01 -1.90169066e-01 7.57772207e-01 -1.89645812e-01 -6.38642550e-01 -2.45606333e-01 -9.86777306e-01 5.89203477e-01 1.29563916e+00 -5.30053854e-01 8.23404714e-02 8.37884545e-01 2.94457525e-01 -2.28561759e-01 -1.32635486e+00 4.37222779e-01 3.02020609e-01 -1.72892070e+00 3.96814793e-01 -5.53846620e-02 9.94038880e-01 -5.76408543e-02 1.48170471e-01 3.18904281e-01 4.63892996e-01 -1.20458174e+00 1.05075145e+00 5.77116728e-01 7.77435660e-01 -7.49232322e-02 9.52184021e-01 5.77191293e-01 -1.99502409e-01 2.33809844e-01 2.19427031e-02 -1.36666343e-01 4.24978197e-01 7.05634594e-01 -9.89121497e-01 -6.33499622e-02 5.76998174e-01 5.03949344e-01 -8.43245447e-01 9.31457520e-01 -4.04403031e-01 7.68939793e-01 6.78387955e-02 -5.41662395e-01 2.85202682e-01 3.22760582e-01 4.00745809e-01 9.69167948e-01 -1.80282727e-01 9.16200429e-02 -2.61474401e-01 8.21090817e-01 -7.82432616e-01 4.13217880e-02 -7.54369140e-01 -6.65413857e-01 4.41036105e-01 9.48958755e-01 -4.25975680e-01 -6.70105457e-01 -4.80645031e-01 8.07456493e-01 2.62117416e-01 4.92785841e-01 -4.79094446e-01 -8.41807723e-01 -1.14194348e-01 4.13887173e-01 4.16628808e-01 1.23060241e-01 -2.19647691e-01 -1.12105763e+00 2.77238339e-01 -1.38931084e+00 2.84105122e-01 -1.27174485e+00 -1.55613923e+00 9.78428900e-01 -2.05896646e-01 -9.78536427e-01 -3.49191368e-01 -4.43942785e-01 -5.38248181e-01 7.68578410e-01 -1.01152742e+00 -4.86339688e-01 -2.20514387e-01 3.59088890e-02 7.14665115e-01 -2.03791082e-01 6.54536545e-01 -8.94142911e-02 -3.02969962e-01 4.18186605e-01 2.03176796e-01 5.40423572e-01 7.33225942e-01 -1.13225472e+00 8.81824315e-01 7.13709772e-01 1.67573839e-01 8.85009944e-01 1.13503826e+00 -8.50779176e-01 -6.72201097e-01 -1.05455327e+00 1.68555641e+00 -6.06110811e-01 8.81822586e-01 -5.03588557e-01 -7.72296846e-01 8.42491627e-01 9.98468637e-01 -4.77415383e-01 7.41630197e-01 -1.91996485e-01 -3.91378969e-01 1.43320948e-01 -8.02393377e-01 6.79704189e-01 4.52902377e-01 -6.37028694e-01 -1.32741976e+00 8.58796358e-01 1.31182432e+00 -4.29328352e-01 -6.63036406e-01 6.26899675e-02 2.68854231e-01 -5.75410247e-01 3.88413042e-01 -1.04797566e+00 1.27887511e+00 8.02325830e-02 7.79428706e-02 -1.08343387e+00 -1.55718774e-01 -5.36991239e-01 -1.48595870e-01 1.03278029e+00 7.02887416e-01 -3.61143202e-01 5.33726990e-01 7.05758274e-01 2.01064855e-01 -6.77811503e-01 -6.54343843e-01 -3.76600146e-01 1.19520910e-01 2.18491063e-01 4.88507599e-01 5.21766424e-01 5.32318093e-02 1.06703877e+00 -2.84894615e-01 -3.81572843e-01 2.99833894e-01 1.83884576e-01 7.34456897e-01 -1.08140087e+00 -2.11342961e-01 -4.00811344e-01 2.21648127e-01 -1.34370196e+00 8.84836316e-02 -6.23356462e-01 2.95682728e-01 -1.85000873e+00 4.44145650e-01 1.44028321e-01 3.15938890e-01 1.37875304e-01 -2.23953575e-01 -1.59716643e-02 -1.10673122e-01 3.90059948e-01 -8.67446065e-01 6.03670418e-01 1.23459554e+00 -3.61725777e-01 1.84299752e-01 2.13788211e-01 -9.52020884e-01 2.78558344e-01 6.05507851e-01 -5.57123423e-01 -2.70799041e-01 -5.57886183e-01 9.38032329e-01 2.71882445e-01 3.57629955e-01 -7.46046960e-01 1.38342440e-01 -4.36622687e-02 4.05168295e-01 -8.61170471e-01 -2.19880968e-01 -3.83914113e-01 -2.39728004e-01 2.67537445e-01 -1.41010535e+00 2.44824082e-01 -6.37285709e-02 6.45097077e-01 -3.49809647e-01 -5.84234357e-01 2.99333602e-01 -2.97726899e-01 -8.02322701e-02 2.20043153e-01 -7.68700480e-01 7.05464542e-01 6.02220893e-01 3.54183167e-01 -8.17039669e-01 -9.88118172e-01 -3.75496626e-01 3.48613709e-01 3.23830068e-01 6.58185303e-01 6.30402088e-01 -1.21050680e+00 -9.53590035e-01 -9.28051770e-02 2.65202612e-01 -1.34783372e-01 -1.01288006e-01 3.44584882e-01 -7.03059375e-01 9.04729605e-01 2.32336998e-01 -7.36323968e-02 -6.81692898e-01 8.04570496e-01 1.98124930e-01 -4.31923151e-01 -5.39284527e-01 8.97175252e-01 3.63360256e-01 -6.58620149e-02 5.10874033e-01 -1.81302741e-01 -4.00395364e-01 3.35757256e-01 1.01046383e+00 2.88091302e-02 3.61862481e-02 -5.35336077e-01 7.61284381e-02 -1.38646560e-02 -5.83829463e-01 -5.78157067e-01 1.01111770e+00 -7.18782544e-02 -1.31381437e-01 4.97391254e-01 1.50676763e+00 1.15832373e-01 -8.33915055e-01 -3.87895226e-01 -5.22588529e-02 -7.13203773e-02 -2.72400267e-02 -7.49131620e-01 -5.60511589e-01 5.21607041e-01 -6.73264116e-02 6.17900610e-01 6.47547722e-01 1.45075083e-01 1.08362162e+00 5.95477045e-01 -3.84664070e-03 -1.16666436e+00 4.28259194e-01 5.99851489e-01 1.35991788e+00 -1.31475282e+00 2.45975573e-02 3.44013482e-01 -7.87031949e-01 1.00948250e+00 4.71325308e-01 -2.12102979e-01 3.69160950e-01 -1.34644821e-01 3.94816697e-02 -5.29220819e-01 -9.50202405e-01 3.70793581e-01 5.86294413e-01 2.33030811e-01 5.81088245e-01 -2.11462721e-01 -3.71454507e-01 8.09798896e-01 -7.68892467e-01 2.98236813e-02 8.40210140e-01 8.05433214e-01 -5.96392572e-01 -3.43470961e-01 -2.44767189e-01 9.37685311e-01 -6.69925690e-01 -1.76796153e-01 -6.39007211e-01 4.01516289e-01 -5.45118928e-01 1.15359163e+00 3.10368240e-01 -2.85330206e-01 3.58543754e-01 -1.72972549e-02 -6.47639036e-02 -8.18059862e-01 -6.33865774e-01 -3.57602358e-01 4.07432109e-01 -2.34074011e-01 -5.29131703e-02 -4.38524008e-01 -8.93528998e-01 -2.68750310e-01 -6.54280111e-02 4.80135083e-01 3.50259751e-01 9.28298652e-01 3.48950505e-01 2.39019737e-01 8.48316133e-01 -2.30062172e-01 -9.94151115e-01 -1.05738842e+00 -3.11814934e-01 6.33285999e-01 9.73910689e-01 -3.82598042e-02 -6.26750946e-01 4.91917491e-01]
[12.295337677001953, 9.359468460083008]
adda9c4c-4635-4fb4-9bd1-704d3bc36e95
improving-road-signs-detection-performance-by
2010.06453
null
https://arxiv.org/abs/2010.06453v1
https://arxiv.org/pdf/2010.06453v1.pdf
Improving Road Signs Detection performance by Combining the Features of Hough Transform and Texture
With the large uses of the intelligent systems in different domains, and in order to increase the drivers and pedestrians safety, the road and traffic sign recognition system has been a challenging issue and an important task for many years. But studies, done in this field of detection and recognition of traffic signs in an image, which are interested in the Arab context, are still insufficient. Detection of the road signs present in the scene is the one of the main stages of the traffic sign detection and recognition. In this paper, an efficient solution to enhance road signs detection, including Arabic context, performance based on color segmentation, Randomized Hough Transform and the combination of Zernike moments and Haralick features has been made. Segmentation stage is useful to determine the Region of Interest (ROI) in the image. The Randomized Hough Transform (RHT) is used to detect the circular and octagonal shapes. This stage is improved by the extraction of the Haralick features and Zernike moments. Furthermore, we use it as input of a classifier based on SVM. Experimental results show that the proposed approach allows us to perform the measurements precision.
['Abdellah Amghar', 'Karim Afdel', 'Mourad Boussaid', 'Tarik Ayaou']
2020-10-13
null
null
null
null
['traffic-sign-recognition', 'traffic-sign-detection']
['computer-vision', 'computer-vision']
[ 1.60761610e-01 -4.26661670e-01 2.22886335e-02 -3.40678394e-01 -7.64560550e-02 -3.37739825e-01 7.73083627e-01 1.57045685e-02 -5.53257465e-01 5.99683166e-01 -2.22930416e-01 -5.38356304e-01 -4.43791837e-01 -7.91920424e-01 -1.42964751e-01 -8.66860271e-01 1.55853540e-01 4.27392244e-01 6.46521389e-01 -2.89520591e-01 8.17771971e-01 9.83819723e-01 -1.86861563e+00 2.33093016e-02 9.60592389e-01 8.57161820e-01 7.36881569e-02 6.52673483e-01 -3.95279914e-01 6.41554177e-01 -3.00866872e-01 -8.56358334e-02 1.92620471e-01 -4.90466863e-01 -4.31302935e-01 1.53808564e-01 1.89753592e-01 -3.14957917e-01 1.86628729e-01 1.00041628e+00 3.11600119e-01 4.70496267e-02 1.00597382e+00 -1.22678697e+00 -1.90209463e-01 1.05492800e-01 -5.50346494e-01 3.00742775e-01 -3.36156860e-02 7.37212747e-02 3.37554872e-01 -8.67194355e-01 4.61546302e-01 1.02491462e+00 1.60859913e-01 -8.96006636e-03 -5.05865037e-01 -7.06046581e-01 -5.44555426e-01 1.03560531e+00 -1.17819357e+00 -1.67081445e-01 8.82636428e-01 -5.18249989e-01 2.65529662e-01 3.80983800e-01 7.02287674e-01 8.83758664e-02 -1.06360272e-01 5.92063189e-01 1.72238553e+00 -7.89392710e-01 -7.07054809e-02 4.23490435e-01 6.05864704e-01 5.61643183e-01 3.63028109e-01 5.56287095e-02 2.20612571e-01 3.17747086e-01 5.02508759e-01 3.57595682e-02 1.23226427e-01 -2.09508941e-01 -7.30818510e-01 6.80325985e-01 1.80334732e-01 9.10540223e-01 -5.51687896e-01 -9.05967802e-02 2.17816859e-01 -1.56400934e-01 -1.23352416e-01 -2.06253640e-02 -1.25860125e-01 -4.27823067e-01 -1.02564943e+00 -1.07103378e-01 5.92521608e-01 3.66041422e-01 6.35025442e-01 -7.84459710e-02 6.57127658e-03 4.70509857e-01 3.71598810e-01 9.76682663e-01 4.05429542e-01 -3.35899681e-01 2.58911937e-01 1.03258312e+00 -1.23794693e-02 -1.25617540e+00 -2.12703675e-01 -5.89301735e-02 -5.47212005e-01 6.59383476e-01 7.57544696e-01 -7.02883676e-02 -1.16605222e+00 6.20751262e-01 6.33129299e-01 5.91481738e-02 -1.34195596e-01 8.69773030e-01 7.16812611e-01 8.94470751e-01 5.29856142e-03 8.10680240e-02 1.41597080e+00 -7.54683852e-01 -7.56006718e-01 1.30619019e-01 4.78221059e-01 -1.25674331e+00 6.28302217e-01 4.22334641e-01 -4.57870156e-01 -4.26230311e-01 -8.78871620e-01 1.75365582e-01 -9.06920910e-01 8.48594427e-01 3.84841710e-01 9.72928405e-01 -4.82082397e-01 1.42873436e-01 -3.85710686e-01 -6.63765788e-01 -1.29042640e-02 2.49780685e-01 -3.65604788e-01 -1.88321859e-01 -8.12869012e-01 1.43890607e+00 4.98035431e-01 6.27770483e-01 -4.09418344e-02 2.78355241e-01 -4.47713107e-01 -1.00266926e-01 3.16475511e-01 4.00877446e-01 5.61693132e-01 -9.84921753e-01 -1.18797505e+00 8.02491248e-01 -1.17342733e-01 -2.77713865e-01 6.63486719e-01 1.94148615e-01 -6.01508260e-01 3.36199313e-01 -1.39365315e-01 1.22366801e-01 6.83321655e-01 -9.39664245e-01 -9.56928134e-01 -2.50031948e-01 -4.65028793e-01 9.73679200e-02 1.24149680e-01 5.28510451e-01 -1.64804399e-01 -3.86937000e-02 2.93169528e-01 -8.45368624e-01 1.34296000e-01 -3.19797158e-01 -1.70348823e-01 -3.78514498e-01 1.42256641e+00 -1.24291253e+00 1.01920223e+00 -2.05123901e+00 -5.18382728e-01 9.74726021e-01 -3.15145552e-01 8.99991691e-01 3.67950708e-01 1.41998097e-01 4.17822897e-02 -9.25477073e-02 -2.31226414e-01 5.90501010e-01 -6.74815029e-02 7.01775998e-02 8.46045092e-02 3.99181008e-01 2.78898001e-01 4.49235857e-01 -3.54877174e-01 -9.05011833e-01 7.85233140e-01 3.12187880e-01 2.17959389e-01 -1.98497504e-01 4.48806494e-01 3.14230055e-01 -6.60547435e-01 7.08182275e-01 1.09515095e+00 4.80392188e-01 -2.62470454e-01 -4.41682398e-01 -6.56094909e-01 -3.31294358e-01 -1.48330188e+00 1.45731583e-01 -9.43727791e-02 8.64773810e-01 -5.32248877e-02 -1.29785872e+00 1.30625844e+00 2.28693187e-01 4.55013037e-01 -1.00493848e+00 5.03201544e-01 6.96549833e-01 2.48859510e-01 -9.43630695e-01 3.97020221e-01 2.62843877e-01 4.78638500e-01 1.80905059e-01 -4.18929130e-01 -1.02430552e-01 8.82057428e-01 -2.96662778e-01 2.26044670e-01 1.89242154e-01 3.49892944e-01 1.93266124e-01 1.08517098e+00 2.72943914e-01 7.35389963e-02 3.85286659e-02 -2.54785568e-01 1.18394509e-01 4.27429676e-01 -2.92216599e-01 -1.17498112e+00 -3.30119789e-01 -4.23486918e-01 4.95218903e-01 1.78459436e-01 5.40803254e-01 -6.66520238e-01 -4.13657606e-01 -4.41972315e-02 5.23075521e-01 -1.80401564e-01 2.67574757e-01 -8.62140179e-01 -6.76247180e-01 3.60403389e-01 2.65253019e-02 1.08014131e+00 -1.08077657e+00 -1.04627669e+00 4.33570109e-02 1.19908907e-01 -8.56314242e-01 8.54244083e-02 -1.68113574e-01 -8.49075556e-01 -1.54644108e+00 -1.01618052e+00 -8.55815113e-01 7.32147813e-01 2.60282189e-01 1.72783002e-01 2.44958714e-01 -5.44473290e-01 1.49500174e-02 -8.10251772e-01 -3.48743200e-01 -5.65849781e-01 -1.33156374e-01 -6.50326908e-01 4.29859996e-01 5.47102571e-01 -9.23556760e-02 -3.61759752e-01 7.22521842e-01 -6.90647304e-01 -1.34556651e-01 1.12664056e+00 4.58367079e-01 7.47274011e-02 -3.10297254e-02 3.24984282e-01 -6.25606477e-01 4.03569400e-01 -8.90590698e-02 -8.44325960e-01 3.86371791e-01 -4.98920381e-01 3.74135077e-02 2.24915087e-01 -4.55567762e-02 -8.93196583e-01 2.74443626e-03 -1.71623498e-01 4.04416829e-01 -6.22637451e-01 2.28935331e-01 -7.32103288e-02 -5.62403202e-01 4.87940907e-01 5.02981782e-01 1.96929008e-01 -3.84294003e-01 3.66996944e-01 1.14805984e+00 1.69391170e-01 -2.68081538e-02 8.69717002e-01 2.14553237e-01 4.60656047e-01 -1.32424903e+00 1.27731308e-01 -8.94580483e-01 -7.45977223e-01 -7.19388902e-01 1.01813781e+00 -5.77207990e-02 -8.67712379e-01 7.15868890e-01 -1.16802633e+00 3.06142390e-01 7.05688819e-02 8.13779473e-01 -8.97291079e-02 6.88652217e-01 -5.94381467e-02 -1.28519738e+00 -3.04469645e-01 -1.17985606e+00 4.39704686e-01 7.17886448e-01 3.31338435e-01 -5.42491496e-01 -4.36842293e-01 5.47042906e-01 4.76769865e-01 3.67759585e-01 8.80688846e-01 -5.77364981e-01 -8.39227319e-01 -5.92534602e-01 -7.27370262e-01 4.83274907e-01 4.70209792e-02 6.53546691e-01 -5.06847560e-01 5.98460317e-01 -2.74629384e-01 9.08308178e-02 7.87574053e-01 4.22209501e-01 5.15640795e-01 2.20807735e-02 -2.39577711e-01 2.18330726e-01 1.25390089e+00 1.00512135e+00 9.34840620e-01 7.38863885e-01 3.54717940e-01 5.76782584e-01 1.25161147e+00 2.16916695e-01 1.39084980e-01 3.70396793e-01 3.16592544e-01 -3.38614017e-01 -1.84942260e-01 4.48632538e-02 2.99173295e-01 5.25799453e-01 -4.89824325e-01 2.75057226e-01 -9.50378835e-01 4.86196548e-01 -1.56907427e+00 -1.18274558e+00 -8.33305776e-01 2.10010552e+00 3.52921337e-01 6.82380572e-02 2.20544904e-01 9.84805942e-01 1.04935956e+00 -2.00755209e-01 6.48842147e-03 -7.72210419e-01 -1.34802938e-01 3.27088058e-01 9.30343866e-01 4.75559145e-01 -1.11209238e+00 7.22613335e-01 4.56724691e+00 1.00259995e+00 -1.60878074e+00 -2.31473193e-01 3.69280010e-01 9.32526290e-01 2.29840770e-01 1.05362587e-01 -7.76610792e-01 6.23301148e-01 3.16430241e-01 2.43560985e-01 2.18544886e-01 7.35337257e-01 4.95133281e-01 -9.26262021e-01 -2.25585878e-01 8.15184474e-01 1.66794524e-01 -7.23986149e-01 -1.76281899e-01 -5.75990900e-02 4.56493884e-01 -4.25913990e-01 3.07250619e-02 -6.74004257e-02 -2.69036114e-01 -6.90062225e-01 4.93138164e-01 7.44654894e-01 3.09762001e-01 -7.59804368e-01 1.17196226e+00 2.86486149e-01 -1.16974545e+00 -1.67231396e-01 -1.95569128e-01 4.87131663e-02 2.49065682e-01 4.11828816e-01 -1.39162064e+00 2.90068120e-01 2.35817954e-01 3.19051594e-01 -8.30674946e-01 1.87360179e+00 -4.43780333e-01 6.72699690e-01 -5.92075884e-01 -8.48151088e-01 3.61641824e-01 -8.55222642e-01 4.05971169e-01 1.20944107e+00 5.65092564e-01 -1.66632608e-01 -1.96655080e-01 6.60007119e-01 7.09383309e-01 8.12409401e-01 -4.94171262e-01 -1.06858626e-01 1.97778553e-01 1.23363459e+00 -1.07842195e+00 -4.83227104e-01 1.02048963e-02 5.04693270e-01 -5.34686267e-01 2.83313364e-01 -8.77744138e-01 -9.93618667e-01 -2.51359850e-01 2.60131121e-01 2.22350761e-01 -4.79944766e-01 -4.00303841e-01 -4.98676538e-01 1.67864144e-01 -6.64435983e-01 2.03732803e-01 -6.70711756e-01 -2.94069350e-01 6.22678399e-02 5.59859611e-02 -1.29494953e+00 -1.23085484e-01 -9.88427579e-01 -7.34730721e-01 1.12695742e+00 -1.59488165e+00 -1.40136123e+00 -3.91436100e-01 4.45461780e-01 3.25028181e-01 -1.26595601e-01 2.08072200e-01 6.04527950e-01 -4.57569540e-01 9.98155475e-02 4.42165196e-01 4.09968287e-01 5.58915019e-01 -9.31961477e-01 -2.70711154e-01 1.06673062e+00 -1.69568241e-01 1.77489474e-01 6.42079949e-01 -6.84365869e-01 -9.16781485e-01 -3.87784004e-01 1.20498610e+00 1.69197246e-01 3.71316850e-01 3.32950681e-01 -3.93783838e-01 1.74720407e-01 -9.03467089e-02 -3.67138445e-01 1.84902936e-01 -4.01883274e-01 3.15455973e-01 -4.19344276e-01 -1.14933264e+00 4.79195625e-01 1.73751579e-03 -7.83409774e-02 -7.32259333e-01 1.00695141e-01 -4.18750256e-01 -8.10957700e-02 -3.34662348e-01 2.16464385e-01 8.03077221e-01 -1.00074017e+00 7.81557024e-01 -2.08765462e-01 -3.25165093e-02 -6.89846933e-01 3.14600140e-01 -7.89883375e-01 1.04210198e-01 1.76679000e-01 6.92157984e-01 1.28191936e+00 4.87301797e-01 -6.77433193e-01 7.51927853e-01 6.02943063e-01 1.19824834e-01 -2.60599464e-01 -8.03008795e-01 -3.91977400e-01 -5.55156529e-01 -1.23995773e-01 2.16512650e-01 6.35038018e-01 -1.22539207e-01 -1.01058461e-01 -2.83407599e-01 -3.49242473e-03 5.80278337e-01 1.02510057e-01 8.70929658e-01 -1.26313651e+00 2.72852629e-01 -6.08116746e-01 -9.52624559e-01 -6.50482953e-01 -4.33026880e-01 -5.20499647e-01 -1.10141769e-01 -1.80945492e+00 -2.16303647e-01 -5.47984719e-01 1.74904570e-01 2.27888584e-01 -4.25282679e-02 1.26052693e-01 4.07828629e-01 4.89840768e-02 -4.38648313e-02 1.12344138e-01 1.38518322e+00 -1.55413106e-01 -1.39982700e-01 3.04620296e-01 1.71018109e-01 6.48637116e-01 6.81073070e-01 -1.48389488e-01 -1.72412321e-02 1.72507539e-01 -1.84988715e-02 -2.19435602e-01 3.45595926e-01 -1.07034636e+00 4.22764599e-01 -3.77079189e-01 3.10899019e-01 -1.15559030e+00 2.25596458e-01 -1.16741574e+00 -1.38406277e-01 7.51199424e-01 1.21241435e-01 3.83978919e-03 -7.82800987e-02 6.59912527e-02 -3.22924674e-01 -6.45788193e-01 8.13348413e-01 9.95639935e-02 -1.23456943e+00 -2.63560027e-01 -6.07591152e-01 -4.34513718e-01 1.48531878e+00 -9.13189709e-01 -2.45812610e-01 -2.29556903e-01 -4.25331950e-01 6.11211956e-02 -4.58698720e-02 -8.16697031e-02 6.14972532e-01 -9.97236550e-01 -5.74983120e-01 1.11365005e-01 -9.47833881e-02 -4.02866930e-01 2.80087739e-01 1.25232375e+00 -1.42815125e+00 5.11776090e-01 -8.31989348e-01 -3.48453432e-01 -1.83612144e+00 2.73512840e-01 1.98526800e-01 -3.85067500e-02 2.83542462e-02 1.66593596e-01 -7.07368314e-01 1.41068831e-01 3.62884887e-02 -6.64939165e-01 -1.22385371e+00 3.55458260e-01 2.75266469e-01 9.92681205e-01 -4.02038060e-02 -1.16224897e+00 -2.19862044e-01 1.27783251e+00 4.50125933e-01 -2.33597741e-01 8.22134495e-01 3.32291201e-02 -3.47627580e-01 1.19600438e-01 8.60028863e-01 2.44276777e-01 -5.15276074e-01 8.95411968e-02 3.07494253e-01 -6.14228606e-01 -1.50476709e-01 -9.52047527e-01 -6.17986858e-01 1.23180580e+00 9.38290536e-01 4.01187330e-01 9.85927522e-01 -5.71507335e-01 5.69881141e-01 4.32235748e-01 1.50927201e-01 -1.51838851e+00 -5.52889049e-01 4.78263587e-01 5.48100650e-01 -1.25309527e+00 -1.50472835e-01 -6.58466518e-01 -6.62376463e-01 1.68245268e+00 3.31471533e-01 -3.85488085e-02 6.10102117e-01 -2.78957281e-02 7.24618062e-02 1.41356885e-01 2.54839301e-01 -1.00799882e+00 4.81544018e-01 4.92854983e-01 3.55072051e-01 1.92768887e-01 -1.29563010e+00 1.08204015e-01 7.01267570e-02 3.16767246e-01 4.96347249e-01 9.23497736e-01 -1.26161206e+00 -1.11406922e+00 -1.12913299e+00 5.18353045e-01 -1.63173586e-01 1.98515087e-01 -4.09217179e-01 9.89051998e-01 6.79424644e-01 1.02031970e+00 -2.28499800e-01 -8.67210850e-02 3.70809615e-01 2.13991046e-01 3.34496796e-01 6.59116507e-02 -1.74229413e-01 -1.58665270e-01 2.47716621e-01 8.18251669e-02 -4.80247825e-01 -6.90472722e-01 -1.42185712e+00 -5.14686517e-02 -4.23583061e-01 4.06154543e-01 1.38292015e+00 1.17754591e+00 -3.02704185e-01 -2.28574723e-02 6.84025526e-01 -3.90129477e-01 -4.00434226e-01 -9.94273305e-01 -5.93294382e-01 6.09638512e-01 5.12733832e-02 -7.08026409e-01 -2.20277905e-01 -3.71342599e-02]
[9.789229393005371, -4.990457057952881]
016e57b7-8b26-4615-8065-16c099b40884
low-resource-cross-lingual-adaptive-training
2307.00382
null
https://arxiv.org/abs/2307.00382v1
https://arxiv.org/pdf/2307.00382v1.pdf
Low-Resource Cross-Lingual Adaptive Training for Nigerian Pidgin
Developing effective spoken language processing systems for low-resource languages poses several challenges due to the lack of parallel data and limited resources for fine-tuning models. In this work, we target on improving upon both text classification and translation of Nigerian Pidgin (Naija) by collecting a large-scale parallel English-Pidgin corpus and further propose a framework of cross-lingual adaptive training that includes both continual and task adaptive training so as to adapt a base pre-trained model to low-resource languages. Our studies show that English pre-trained language models serve as a stronger prior than multilingual language models on English-Pidgin tasks with up to 2.38 BLEU improvements; and demonstrate that augmenting orthographic data and using task adaptive training with back-translation can have a significant impact on model performance.
['Merel Scholman', 'Ernie Chang', 'Muhammed Saeed', 'Pin-Jie Lin']
2023-07-01
null
null
null
null
['text-classification']
['natural-language-processing']
[-3.43152136e-02 -2.18002692e-01 -1.06273517e-01 -7.97822118e-01 -1.33635378e+00 -7.24125862e-01 6.21947348e-01 1.50612488e-01 -8.83391142e-01 7.42938638e-01 6.00646377e-01 -6.09656036e-01 2.15924129e-01 -4.39083248e-01 -5.64697802e-01 -1.18132299e-02 1.97362334e-01 9.98345673e-01 3.95343406e-03 -6.00564122e-01 1.44551218e-01 2.95979261e-01 -8.32755804e-01 5.79122186e-01 9.83999848e-01 -2.12875810e-02 5.88878810e-01 8.26657414e-01 -3.18831682e-01 4.47855234e-01 -5.01991689e-01 -1.58818841e-01 2.95812398e-01 -4.18162674e-01 -1.04846680e+00 -1.14448823e-01 4.22599286e-01 -2.76168466e-01 2.78448351e-02 5.42645454e-01 6.33678734e-01 2.53379732e-01 6.78615272e-01 -6.46051764e-01 -8.78336668e-01 9.48498368e-01 -2.74445027e-01 2.52056062e-01 2.11139604e-01 -8.98000691e-03 7.38611937e-01 -1.16820264e+00 4.30225492e-01 1.49406981e+00 8.81848991e-01 5.79168200e-01 -9.60061014e-01 -7.69534290e-01 8.36499855e-02 -2.53805425e-02 -1.32641470e+00 -9.39490616e-01 2.14763716e-01 -1.88787237e-01 1.73808575e+00 -8.09206441e-02 1.17284797e-01 7.57615805e-01 1.88732788e-01 2.69696116e-01 1.32291174e+00 -1.24974740e+00 -1.88398197e-01 5.16256094e-01 -5.05741127e-02 6.10686421e-01 1.95055373e-03 -2.84856111e-01 -6.27596259e-01 1.59409717e-02 4.36499119e-01 -7.31241643e-01 2.06536338e-01 4.51415002e-01 -1.01160085e+00 9.25560594e-01 -4.62967381e-02 2.77484804e-01 -3.94916236e-01 -2.65415817e-01 6.36946559e-01 4.73879099e-01 6.25227392e-01 6.79176331e-01 -1.05288100e+00 -2.85581708e-01 -6.27829850e-01 -2.19851211e-01 8.59856963e-01 9.95638907e-01 6.44035399e-01 3.08490127e-01 1.20660350e-01 1.59537566e+00 3.90503287e-01 8.84972751e-01 9.07475114e-01 -5.99792004e-01 7.49183476e-01 4.02212232e-01 -2.21421838e-01 -2.96680748e-01 -3.42990160e-01 1.10639058e-01 -1.60615250e-01 -1.58297524e-01 5.35396218e-01 -6.09854519e-01 -1.09009862e+00 1.64759064e+00 2.55743653e-01 -5.46501935e-01 7.60131896e-01 4.05171096e-01 4.57569867e-01 1.16356552e+00 5.61130285e-01 -1.42882153e-01 1.44972110e+00 -1.20981514e+00 -4.94679391e-01 -7.34590828e-01 9.68471050e-01 -1.31661057e+00 1.68737388e+00 3.10115337e-01 -1.13935089e+00 -5.01939774e-01 -9.44462240e-01 -2.55438060e-01 -6.92459226e-01 1.73265189e-01 3.46506745e-01 7.63387203e-01 -1.12140083e+00 -9.65728238e-03 -7.76626885e-01 -1.06659925e+00 -2.44778395e-01 3.74914706e-01 -4.89006042e-01 -4.90619689e-01 -1.19073391e+00 1.28683448e+00 6.79085255e-01 -2.72862375e-01 -6.08632505e-01 -2.86981374e-01 -7.29355752e-01 -2.49726757e-01 6.49755895e-02 -2.03505471e-01 1.45356262e+00 -1.15484035e+00 -1.72132063e+00 8.15893114e-01 -1.64265320e-01 -1.38640046e-01 2.18549505e-01 -1.77636594e-01 -5.14979422e-01 -1.79840550e-01 1.14610545e-01 6.73715353e-01 3.85609597e-01 -8.79862368e-01 -9.60496724e-01 -4.83786523e-01 -3.11471671e-01 9.32884336e-01 -7.82854736e-01 7.60521710e-01 -6.62045419e-01 -6.39437377e-01 -5.18860891e-02 -1.10823250e+00 -1.68249719e-02 -7.43780255e-01 3.65195811e-01 -3.42806280e-01 5.31552732e-01 -1.34960759e+00 1.25374067e+00 -1.63938415e+00 -2.05343097e-01 1.25634402e-01 -6.59969628e-01 4.41898644e-01 -4.72355008e-01 7.69479811e-01 3.97529185e-01 1.17351465e-01 -4.11678329e-02 -3.37043583e-01 -1.57044515e-01 7.13069499e-01 1.26033068e-01 7.56462589e-02 3.10011715e-01 5.82624853e-01 -6.44280910e-01 -3.73018354e-01 -2.33063451e-03 4.40292090e-01 -4.05272812e-01 -2.82621160e-02 -8.18410441e-02 1.55613706e-01 -4.37616296e-02 4.42705452e-01 3.79778922e-01 4.06688362e-01 5.31735718e-01 4.04938370e-01 -4.08792704e-01 8.78058732e-01 -8.40498149e-01 1.35412550e+00 -1.00670671e+00 3.86333346e-01 1.09840997e-01 -6.44245744e-01 8.50108445e-01 3.53121191e-01 -2.15203036e-02 -9.57158685e-01 5.80081306e-02 6.56164050e-01 3.57324630e-01 -4.82054293e-01 8.12749803e-01 -2.39060864e-01 -2.28416577e-01 7.03822494e-01 9.21327099e-02 -2.33350486e-01 2.52250992e-02 -1.36299208e-01 6.77340865e-01 1.15576521e-01 3.70597631e-01 -6.50992870e-01 5.88446796e-01 5.97763538e-01 4.41240877e-01 6.34052098e-01 -1.48789557e-02 2.14260906e-01 -2.84100085e-01 -7.91319832e-03 -1.29909217e+00 -6.05970800e-01 -7.28422180e-02 2.06802130e+00 -7.48482168e-01 -2.57695228e-01 -8.24985921e-01 -5.72559953e-01 -1.58811837e-01 1.21225381e+00 -1.77103415e-01 7.88812786e-02 -9.57651675e-01 -8.38026583e-01 1.17673457e+00 5.40047288e-01 3.84696215e-01 -1.18325496e+00 3.88715006e-02 5.46004295e-01 -1.39030114e-01 -1.10710013e+00 -7.37949550e-01 4.37987745e-01 -6.91501617e-01 -5.04392445e-01 -5.56305230e-01 -1.44248664e+00 3.56569916e-01 7.28956610e-02 1.05861306e+00 -2.12918788e-01 1.93272278e-01 3.14145446e-01 -3.69135231e-01 -8.53413284e-01 -1.08635283e+00 5.72299361e-01 3.85916382e-01 -6.56446099e-01 7.98390269e-01 1.81779951e-01 2.98666596e-01 1.23143405e-01 -5.72860241e-01 3.45561989e-02 7.76976466e-01 8.55141938e-01 4.49089408e-01 -2.42763475e-01 1.02505326e+00 -1.18022013e+00 9.01248395e-01 -3.36188316e-01 -4.54782009e-01 6.42731786e-01 -8.50060582e-01 1.15497455e-01 8.08928549e-01 -4.94191349e-01 -1.60119188e+00 8.12771842e-02 -3.31661165e-01 3.05620790e-01 -1.77338980e-02 8.51526499e-01 8.48798722e-04 -7.46418769e-03 8.17887008e-01 1.48472056e-01 1.96425361e-03 -6.37713313e-01 3.63903165e-01 1.18945110e+00 5.22430897e-01 -7.84725547e-01 4.63088065e-01 -1.57280833e-01 -6.04458928e-01 -1.19690800e+00 -3.60929847e-01 -5.69377959e-01 -8.92788827e-01 1.61480054e-01 8.58864069e-01 -1.16821456e+00 9.51861218e-02 6.41945183e-01 -9.89958882e-01 -9.72936749e-01 2.20465750e-01 7.85188437e-01 -4.15877178e-02 6.27357811e-02 -8.00040066e-01 -5.47667205e-01 -6.01318955e-01 -8.94080639e-01 6.78747833e-01 -5.48363570e-03 -3.14201027e-01 -1.47107995e+00 3.68824244e-01 3.52621317e-01 6.84304357e-01 -8.32631946e-01 1.32956088e+00 -1.08369327e+00 1.37385711e-01 1.94323082e-02 -1.68171711e-02 6.41305804e-01 2.76945561e-01 -1.59019426e-01 -8.57166946e-01 -4.63599831e-01 -3.03696275e-01 -6.50924385e-01 7.09868921e-03 5.58306761e-02 9.52592865e-02 -4.59350407e-01 1.28810510e-01 3.77080828e-01 1.34826326e+00 3.41635376e-01 5.60577586e-02 5.20292938e-01 6.38732016e-01 7.10399330e-01 5.55571973e-01 5.03392294e-02 8.68250668e-01 5.31506240e-01 -7.12249935e-01 2.38875719e-03 -3.22793424e-01 -2.57764250e-01 8.10440361e-01 1.60601461e+00 3.87687236e-01 -1.87340975e-01 -1.55503356e+00 6.38879418e-01 -1.49176550e+00 -2.21067190e-01 -9.31040943e-02 2.21930432e+00 1.24405873e+00 -1.91577271e-01 -8.93193632e-02 -5.30080914e-01 4.66124743e-01 -5.30846119e-01 -1.74186856e-01 -1.01065588e+00 -1.86603963e-01 3.91408950e-01 7.28304982e-01 1.01837492e+00 -8.29816282e-01 1.80664682e+00 6.82491446e+00 5.96024632e-01 -1.31329405e+00 3.76469761e-01 3.23455393e-01 2.50432312e-01 -1.21511608e-01 -1.14680968e-01 -1.11517000e+00 -4.19627726e-02 1.65724003e+00 -4.34311688e-01 6.16586387e-01 5.64210474e-01 4.01075810e-01 2.59814486e-02 -8.40758741e-01 6.19867861e-01 3.41033041e-01 -7.33571112e-01 1.19758882e-01 -2.03261822e-01 8.82182777e-01 7.05604494e-01 -3.90311837e-01 7.08117068e-01 7.79994130e-01 -9.38506186e-01 5.17644525e-01 -4.56207283e-02 8.94022882e-01 -6.81000054e-01 7.29641378e-01 5.16846657e-01 -8.94781411e-01 3.09313416e-01 -5.95733523e-01 -8.62238482e-02 -3.27184424e-02 -5.50984442e-01 -1.58731019e+00 1.75260499e-01 6.09018385e-01 1.10620610e-01 -7.78447151e-01 4.51902568e-01 -1.17173821e-01 1.13229585e+00 -6.78277910e-01 -1.41736493e-01 5.20380080e-01 -1.66186571e-01 1.97287142e-01 1.72856462e+00 3.33567828e-01 2.39107326e-01 4.47824001e-01 -1.37850091e-01 -9.89031643e-02 9.45444047e-01 -6.04818225e-01 -1.66587576e-01 5.57551205e-01 1.01755917e+00 -6.38737977e-01 -5.60048997e-01 -7.07738936e-01 1.10201025e+00 4.94272918e-01 2.08847895e-01 -2.34821960e-01 -3.61944884e-01 3.96031320e-01 -3.13400924e-02 -8.78155380e-02 -5.75499892e-01 -5.37798882e-01 -8.94696832e-01 -3.43935996e-01 -1.40803337e+00 5.48785031e-01 -6.03697360e-01 -1.29806316e+00 9.32909012e-01 4.71776398e-03 -4.86698985e-01 -6.27721190e-01 -7.80847967e-01 -3.14341635e-01 1.45274985e+00 -1.27063549e+00 -1.63214874e+00 2.62865815e-02 5.54396629e-01 1.16389155e+00 -6.52058601e-01 1.24227047e+00 3.80326658e-01 -5.06402850e-01 9.00757730e-01 4.68225360e-01 1.50015533e-01 1.21986461e+00 -1.11291957e+00 4.62741196e-01 9.55738008e-01 1.08660489e-01 7.18661666e-01 4.43603516e-01 -8.32094491e-01 -1.35547626e+00 -1.11341524e+00 1.48143017e+00 -5.32938540e-01 7.31187284e-01 -5.38806677e-01 -1.12407100e+00 9.99737740e-01 6.95559144e-01 -7.81647801e-01 7.45546699e-01 4.63656306e-01 -1.82095692e-02 -2.70744711e-02 -9.97594774e-01 6.37088478e-01 5.98208845e-01 -7.61347532e-01 -7.30306268e-01 5.01240373e-01 4.52083647e-01 -1.02376148e-01 -8.92713070e-01 -1.87499046e-01 5.54785609e-01 -1.46282077e-01 4.23465759e-01 -8.97444129e-01 -6.36772141e-02 1.71635255e-01 -3.71998250e-01 -1.66378546e+00 -2.37018928e-01 -3.96610141e-01 1.01178479e+00 1.58798456e+00 8.31816137e-01 -8.55215490e-01 3.62408757e-01 5.88107347e-01 -5.63491523e-01 -1.86856568e-01 -6.11641645e-01 -6.15769327e-01 6.24295056e-01 -5.59524000e-01 2.52621710e-01 1.04765022e+00 9.99969244e-02 9.51423764e-01 -4.42093387e-02 3.07852089e-01 1.13266252e-01 -7.59852111e-01 7.45145261e-01 -7.65791357e-01 -2.33419687e-01 -2.17961729e-01 1.67306110e-01 -7.13343084e-01 3.75090241e-01 -1.06466281e+00 5.38612425e-01 -1.45750678e+00 -1.98828168e-02 -9.45181489e-01 -3.53413704e-03 9.69351053e-01 -2.03303173e-01 2.71191835e-01 6.65605143e-02 3.32241923e-01 -1.43138662e-01 3.07174921e-01 5.77646255e-01 3.08636785e-01 -5.45737088e-01 -2.68504262e-01 -7.32631624e-01 6.68020546e-01 8.44144702e-01 -5.99696517e-01 -4.66872811e-01 -1.07128882e+00 -5.08688502e-02 -2.42112488e-01 -6.26945078e-01 -8.19944441e-01 2.62829095e-01 -1.84433460e-01 3.29842925e-01 -1.26449555e-01 1.12761125e-01 -4.85623300e-01 -2.51805097e-01 2.58738667e-01 -3.18993628e-01 7.96394944e-01 8.96227419e-01 -9.24835056e-02 -7.31599554e-02 -4.90649790e-01 8.95510554e-01 -3.21127653e-01 -1.08179975e+00 -1.47388577e-01 -8.34554732e-01 2.84567177e-01 6.51166022e-01 2.91715208e-02 -1.90944940e-01 -3.52832824e-01 -3.46468836e-01 2.96511441e-01 3.84816051e-01 7.08637893e-01 9.61033907e-03 -9.79011953e-01 -1.14181769e+00 5.03167987e-01 1.25632688e-01 -3.28322053e-01 -2.59511948e-01 3.51666123e-01 -1.05761552e+00 7.43685842e-01 -4.02319640e-01 -2.70269513e-01 -1.37101054e+00 -2.09209844e-02 1.27919495e-01 -2.64523536e-01 -1.77096695e-01 7.72831738e-01 -1.72346830e-01 -1.20560694e+00 7.75747672e-02 -6.12033159e-02 -2.86994159e-01 -1.11106917e-01 1.81285545e-01 2.69512653e-01 4.16391432e-01 -9.58955586e-01 -3.77923965e-01 1.96523055e-01 -4.01108384e-01 -8.32533956e-01 1.18885934e+00 -4.98045951e-01 -3.60562839e-03 6.36123478e-01 8.54400814e-01 4.65578914e-01 -7.91903257e-01 -4.26331490e-01 3.93006742e-01 1.04771644e-01 1.32749662e-01 -1.18669617e+00 -1.70062199e-01 7.10431933e-01 4.26594913e-01 -4.19708699e-01 9.92381752e-01 -1.86603919e-01 6.07576966e-01 9.80005383e-01 2.05461234e-01 -1.55774677e+00 -1.86924830e-01 1.26244640e+00 7.16394842e-01 -1.36132729e+00 -1.87910408e-01 -6.33815397e-03 -1.08423388e+00 1.01868320e+00 8.04416955e-01 1.39800251e-01 5.54641902e-01 3.65591407e-01 9.03308272e-01 1.13055713e-01 -8.44210684e-01 7.68561438e-02 3.56642425e-01 4.78957117e-01 9.66735303e-01 1.88407391e-01 -5.73372543e-01 6.19905531e-01 -4.83739287e-01 -1.06512740e-01 4.77967322e-01 8.89951587e-01 -4.97440398e-01 -1.45403421e+00 -6.04659736e-01 3.60190600e-01 -6.77682936e-01 -6.91513717e-01 -3.46192539e-01 1.16652691e+00 -1.33534119e-01 9.65737283e-01 2.03444973e-01 -1.08315386e-01 3.25103998e-01 7.03215897e-01 3.54197443e-01 -1.16272688e+00 -8.36228132e-01 4.16552842e-01 5.71817279e-01 1.82391435e-01 -3.19288559e-02 -7.44507492e-01 -1.31215036e+00 -2.27970645e-01 -1.71532959e-01 3.62470686e-01 9.38635886e-01 1.02439904e+00 7.13132843e-02 1.27455905e-01 2.80932337e-01 -5.92105448e-01 -8.63584220e-01 -1.51143408e+00 -2.12894469e-01 -1.45280575e-02 -3.05690527e-01 4.11373787e-02 3.88866626e-02 3.18843663e-01]
[10.957368850708008, 10.002338409423828]
0cde69ed-26a9-4ebe-a53e-91a1937d7e3b
adaptive-event-detection-for-representative
2107.11287
null
https://arxiv.org/abs/2107.11287v1
https://arxiv.org/pdf/2107.11287v1.pdf
Adaptive Event Detection for Representative Load Signature Extraction
Event detection is the first step in event-based non-intrusive load monitoring (NILM) and it can provide useful transient information to identify appliances. However, existing event detection methods with fixed parameters may fail in case of unpredictable and complicated residential load changes such as high fluctuation, long transition, and near simultaneity. This paper proposes a dynamic time-window approach to deal with these highly complex load variations. Specifically, a window with adaptive margins, multi-timescale window screening, and adaptive threshold (WAMMA) method is proposed to detect events in aggregated home appliance load data with high sampling rate (>1Hz). The proposed method accurately captures the transient process by adaptively tuning parameters including window width, margin width, and change threshold. Furthermore, representative transient and steady-state load signatures are extracted and, for the first time, quantified from transient and steady periods segmented by detected events. Case studies on a 20Hz dataset, the 50Hz LIFTED dataset, and the 60Hz BLUED dataset show that the proposed method can robustly outperform other state-of-art event detection methods. This paper also shows that the extracted load signatures can improve NILM accuracy and help develop other applications such as load reconstruction to generate realistic load data for NILM research.
['Zuyi Li', 'Jiayu Han', 'Wei Tian', 'Lei Yan']
2021-07-23
null
null
null
null
['non-intrusive-load-monitoring', 'non-intrusive-load-monitoring', 'non-intrusive-load-monitoring']
['knowledge-base', 'miscellaneous', 'time-series']
[ 2.18244359e-01 -4.76763248e-01 -2.09291965e-01 -1.99189439e-01 -5.96885562e-01 -2.50225574e-01 2.86704302e-01 4.06882703e-01 1.65188178e-01 8.33723903e-01 1.21520974e-01 -9.20173973e-02 -3.82970273e-01 -7.98176050e-01 1.33801447e-02 -6.14328265e-01 -3.88676941e-01 3.40661466e-01 4.81467426e-01 -1.86065212e-03 -1.80664927e-01 5.53077638e-01 -1.77069056e+00 3.97914052e-02 8.27762246e-01 1.08964825e+00 2.59140849e-01 3.35314095e-01 1.33957773e-01 6.05789304e-01 -1.26561987e+00 6.22384846e-01 -1.33529454e-01 -3.92627716e-01 -3.71982694e-01 1.07378416e-01 -6.53205991e-01 -4.15114880e-01 4.62094396e-02 5.49561262e-01 7.71378756e-01 9.37553123e-02 3.61159414e-01 -1.82606590e+00 2.17113614e-01 8.04226756e-01 -7.72260070e-01 8.87713730e-01 8.23828459e-01 1.16402999e-01 3.83774579e-01 -1.71907485e-01 2.56300151e-01 9.07841802e-01 7.25105584e-01 -1.29640892e-01 -1.25638604e+00 -9.02407289e-01 1.61620766e-01 5.42507410e-01 -1.50009239e+00 -7.40094781e-02 9.11137521e-01 -2.21137449e-01 1.29836023e+00 6.45270586e-01 5.03131032e-01 1.03575146e+00 2.76666969e-01 6.76710844e-01 7.57741153e-01 -3.46133113e-01 3.88513625e-01 -2.22850055e-01 2.59982944e-01 -1.31498277e-01 2.15812966e-01 -1.08876890e-02 -3.22906613e-01 -5.77634156e-01 2.62405187e-01 4.79412258e-01 -3.39720666e-01 6.52295232e-01 -9.09848332e-01 4.51734900e-01 -2.33538181e-01 4.71054256e-01 -8.18300664e-01 -3.83091241e-01 7.33144999e-01 1.96538687e-01 4.88785148e-01 -5.57271168e-02 -5.59729815e-01 -5.11251211e-01 -1.15032911e+00 -1.21065769e-02 8.76652837e-01 8.24078739e-01 3.20142746e-01 3.85648310e-01 -5.31307697e-01 4.43252027e-01 2.08230525e-01 4.39704597e-01 7.16622353e-01 -4.84398514e-01 2.50943899e-01 7.68734455e-01 1.57504618e-01 -5.90067744e-01 -9.73832905e-01 -9.99694988e-02 -8.46065760e-01 -1.02593124e-01 -3.53332609e-02 -2.21939266e-01 -4.54408824e-01 1.56082988e+00 5.61990023e-01 5.73871255e-01 -4.10154223e-01 2.78023988e-01 2.99482703e-01 8.29185188e-01 2.28613824e-01 -1.12013435e+00 1.52254856e+00 1.24831079e-02 -1.36361217e+00 3.70247096e-01 1.47304535e-01 -9.52150524e-01 9.59436774e-01 3.42155904e-01 -8.73262882e-01 -3.62408519e-01 -8.95135939e-01 8.60459089e-01 -2.92306662e-01 -7.97801465e-02 3.28882307e-01 5.61472416e-01 -2.20175773e-01 4.11014289e-01 -1.01009512e+00 -6.62551999e-01 1.65740743e-01 2.26167291e-01 4.74283278e-01 3.58767033e-01 -1.35789907e+00 6.07487679e-01 5.07083893e-01 -1.60234630e-01 -5.86996973e-01 -9.49340701e-01 -5.56453824e-01 3.32399905e-02 2.09222525e-01 -2.47060470e-02 1.27213240e+00 -1.04202870e-02 -1.30730128e+00 1.12572417e-01 -1.90828159e-01 -6.10250831e-01 3.09037209e-01 -2.98350807e-02 -1.14638352e+00 1.86643630e-01 9.97983217e-02 -4.91582334e-01 6.43409491e-01 -6.85131073e-01 -7.55369782e-01 -3.50641787e-01 -4.92885053e-01 -1.34632558e-01 -1.06019780e-01 3.06743562e-01 2.56092072e-01 -7.10756183e-01 -5.70532531e-02 -4.56692517e-01 1.08241916e-01 -8.52391541e-01 -5.43041825e-01 -4.43065763e-01 1.55424201e+00 -5.98683476e-01 1.94301367e+00 -2.14635730e+00 -9.43162143e-01 3.56402189e-01 -3.60493720e-01 1.88548878e-01 5.07944226e-01 9.39192832e-01 -2.27436095e-01 -4.37226415e-01 3.43167707e-02 -5.62135279e-02 2.34380320e-01 2.48775661e-01 -3.90393287e-01 4.64152038e-01 5.21905720e-03 6.16150022e-01 -8.01282525e-01 -3.57127011e-01 9.06798363e-01 3.86618435e-01 8.89840871e-02 3.39943349e-01 -5.70215732e-02 3.34969252e-01 -2.95045197e-01 7.70127833e-01 4.75560397e-01 -2.44573459e-01 6.60254359e-02 -4.80826437e-01 -2.17452615e-01 3.07924151e-01 -1.84150195e+00 7.11015463e-01 -5.56667924e-01 4.84923661e-01 -1.16890989e-01 -9.22590852e-01 9.11652923e-01 8.79334986e-01 1.12294734e+00 -7.29731798e-01 3.01590353e-01 -5.75115792e-02 -2.81494766e-01 -6.17801547e-01 2.32405439e-01 2.57302612e-01 1.67035423e-02 7.44696140e-01 -2.70527542e-01 3.40479404e-01 4.65505809e-01 -7.50383288e-02 1.43213999e+00 -2.42408827e-01 6.56299710e-01 -2.09255308e-01 3.29629749e-01 -3.97154987e-01 7.87198782e-01 3.77917647e-01 -5.12009859e-01 3.53184715e-02 1.17106788e-01 -2.86863625e-01 -5.56166887e-01 -1.11736369e+00 -3.49136144e-01 1.07999682e+00 7.82091767e-02 -3.83541852e-01 -4.64439392e-01 -3.78432162e-02 -1.50602423e-02 1.03508186e+00 -1.91905126e-01 -2.83235788e-01 -7.25437045e-01 -1.49877131e+00 2.18673378e-01 4.96344894e-01 6.34711981e-01 -1.36848068e+00 -9.84827280e-01 1.07757354e+00 -3.37948263e-01 -1.13894641e+00 -4.37122554e-01 5.41931570e-01 -5.66272616e-01 -1.28728330e+00 1.18454002e-01 -5.57783842e-01 2.87379861e-01 8.27544853e-02 1.19373727e+00 -3.13096106e-01 -8.00454617e-01 3.97568822e-01 -3.41850638e-01 -7.25589156e-01 -2.55292833e-01 -8.65801722e-02 3.44158448e-02 5.59041351e-02 6.56330764e-01 -1.12576771e+00 -7.42516458e-01 6.69739664e-01 -9.58438158e-01 -5.79698086e-01 6.88876957e-04 3.36510181e-01 7.01230228e-01 7.45813489e-01 1.30208611e+00 -5.95825255e-01 8.44946146e-01 -8.06845307e-01 -6.25061750e-01 2.02809542e-01 -9.38950300e-01 -6.34835482e-01 6.66604102e-01 -1.02673590e+00 -1.05478883e+00 -1.16095319e-02 1.18524872e-01 -5.97224496e-02 -3.49785686e-01 2.06113178e-02 -3.50165427e-01 6.57870471e-01 5.90795338e-01 8.75423104e-02 -5.75405717e-01 -4.57976818e-01 -1.08888119e-01 8.26227903e-01 6.30751014e-01 -4.33884680e-01 4.29256856e-01 5.19344568e-01 -1.85109109e-01 -7.03909814e-01 -3.04009020e-01 -7.16520190e-01 -1.42182469e-01 -3.07982594e-01 5.69768071e-01 -8.66302371e-01 -1.14224660e+00 5.56760192e-01 -7.90870070e-01 -2.91753799e-01 -5.82994461e-01 4.73542929e-01 -3.77854645e-01 6.16937205e-02 -8.92584383e-01 -1.30422628e+00 -6.76175535e-01 -7.07119644e-01 1.05682588e+00 5.28063297e-01 -8.31264019e-01 -8.39894831e-01 -9.89943147e-02 -1.92626476e-01 5.93777180e-01 8.44795763e-01 6.47882044e-01 -6.53683066e-01 -1.18063189e-01 -5.47415257e-01 2.03444257e-01 -2.42995515e-01 6.65314734e-01 1.54888660e-01 -8.79306316e-01 -3.42795372e-01 2.80950785e-01 4.58078861e-01 6.69902097e-03 5.49720764e-01 1.10952127e+00 -3.24753284e-01 -5.76520026e-01 2.43728966e-01 1.38437450e+00 7.57678747e-01 6.86837852e-01 1.84899658e-01 1.93509817e-01 2.97021419e-02 5.55833936e-01 1.24559915e+00 4.50854063e-01 6.67421103e-01 4.36693013e-01 -1.21066831e-01 2.50440508e-01 1.52276028e-02 4.06488538e-01 7.10296154e-01 2.28949383e-01 -3.18623364e-01 -4.15387809e-01 5.19767642e-01 -1.80390036e+00 -1.38338041e+00 -1.94711462e-01 2.11293674e+00 8.94489408e-01 4.14782345e-01 5.40320039e-01 9.56063747e-01 1.07216704e+00 3.54355276e-02 -7.88517058e-01 -2.53282964e-01 -1.63406804e-01 2.87331998e-01 3.21538210e-01 6.40551448e-02 -9.87508178e-01 5.04906029e-02 6.40170097e+00 6.93885028e-01 -8.93708348e-01 2.49536157e-01 1.98932886e-01 -1.95559025e-01 3.07723641e-01 -4.42746103e-01 -9.18341279e-01 1.12893474e+00 1.49792588e+00 -5.08816600e-01 3.57830822e-01 7.53876448e-01 1.02567029e+00 -4.89418983e-01 -9.44321632e-01 1.13461411e+00 -4.11847174e-01 -7.41318047e-01 -4.69093263e-01 1.08594462e-01 6.26565099e-01 -9.84826311e-03 -5.94501257e-01 2.93123543e-01 2.62051344e-01 -3.40905398e-01 1.55646771e-01 5.60308635e-01 2.43842781e-01 -8.32700133e-01 5.89117944e-01 3.13972235e-01 -1.84669304e+00 -2.02013612e-01 2.03802481e-01 -1.50616840e-01 7.63798356e-01 1.29201436e+00 -7.35246658e-01 5.10626018e-01 7.71137834e-01 2.68032700e-01 -7.58452937e-02 1.05874181e+00 1.65091187e-01 1.15510023e+00 -1.02110946e+00 2.23048672e-01 -5.81658065e-01 1.84469640e-01 5.27304351e-01 1.10230267e+00 3.92360508e-01 1.79633498e-01 5.53066134e-01 5.35609782e-01 3.66878301e-01 -1.37296394e-01 -1.35186836e-01 5.78035831e-01 1.07450128e+00 1.25251889e+00 -9.81609225e-01 -4.66859102e-01 -3.03418696e-01 6.76828861e-01 -6.30052924e-01 3.77994239e-01 -1.17386174e+00 -5.62004387e-01 6.31602824e-01 2.77167499e-01 1.31923869e-01 -4.27961536e-02 -9.91017371e-02 -6.81543469e-01 1.31066144e-01 -4.62534338e-01 7.97720194e-01 -4.51109320e-01 -1.45003486e+00 2.14415058e-01 3.72568071e-01 -1.31892824e+00 -5.90523422e-01 2.18758434e-01 -1.53381121e+00 6.08151555e-01 -1.17499077e+00 -6.83181345e-01 -4.32266593e-01 1.05612111e+00 6.58824325e-01 3.27686012e-01 8.50349724e-01 7.10623205e-01 -8.77936721e-01 4.43398774e-01 2.20383331e-01 -2.05033958e-01 2.68563777e-01 -1.19577813e+00 1.57213330e-01 7.42109776e-01 -3.99006873e-01 7.53491670e-02 9.37175632e-01 -7.84153938e-01 -1.14317441e+00 -1.35039079e+00 6.00396991e-01 -2.96245925e-02 6.44289136e-01 -3.52812111e-01 -1.16809201e+00 4.89812464e-01 4.35808487e-02 -6.10027276e-03 6.98845327e-01 -2.80699372e-01 3.24840367e-01 -4.62568045e-01 -1.49123323e+00 9.34251249e-02 5.35720646e-01 -3.71640354e-01 -4.44068909e-01 5.56912780e-01 5.03511906e-01 -1.95975825e-01 -1.36872959e+00 4.86858726e-01 2.15292558e-01 -7.17024565e-01 7.51746655e-01 1.22142360e-01 -8.41632485e-01 -5.39669037e-01 -2.35895216e-01 -1.01976097e+00 -1.61157355e-01 -1.20514095e+00 -8.24387431e-01 1.86084938e+00 6.47358224e-03 -8.14113796e-01 1.81147039e-01 4.06783760e-01 -2.07387898e-02 -5.24029970e-01 -1.04585004e+00 -9.66362357e-01 -1.06665754e+00 -6.62688017e-01 1.33967924e+00 7.96443760e-01 2.20133364e-01 1.66779101e-01 -1.42685205e-01 3.38036448e-01 5.57622552e-01 6.88254535e-02 3.31085384e-01 -1.33905220e+00 1.18164152e-01 -2.20416903e-01 -4.76731300e-01 -1.88507825e-01 -4.06066298e-01 -1.96975216e-01 -5.24814054e-02 -1.36591887e+00 -1.22676045e-01 -8.61854479e-03 -6.50622845e-01 4.82467920e-01 -2.85227746e-01 1.76537573e-01 -1.81849182e-01 2.28852227e-01 -4.97022688e-01 3.45380634e-01 2.90412843e-01 1.21829219e-01 -7.81250596e-01 4.07341361e-01 1.07564583e-01 5.83039641e-01 1.21292782e+00 -3.00392002e-01 -5.87383509e-01 5.89406967e-01 -2.38691494e-01 -3.56849469e-02 2.51102626e-01 -1.23248243e+00 -7.71192554e-03 -2.72837818e-01 5.56786180e-01 -1.21479750e+00 -4.20028791e-02 -9.26829576e-01 8.06150317e-01 5.64926147e-01 2.23590612e-01 5.17928064e-01 3.43268543e-01 4.31471169e-01 1.72943771e-02 2.41111070e-01 5.69982231e-01 1.69831082e-01 -5.51324546e-01 2.25206271e-01 -5.78286529e-01 1.04929563e-02 1.32948899e+00 -1.52141184e-01 -3.60680163e-01 -2.35240102e-01 -7.80432463e-01 3.82482558e-01 -9.69589204e-02 4.41242129e-01 4.17834967e-02 -1.60721898e+00 -4.28496480e-01 2.73538113e-01 -6.71788976e-02 -2.98526078e-01 2.75160253e-01 9.13113952e-01 1.96024626e-01 1.04250185e-01 2.29101866e-01 -8.01910281e-01 -1.08628798e+00 5.73870361e-01 1.84136853e-01 -4.37679052e-01 -8.82776022e-01 -1.44208714e-01 -3.92078549e-01 4.50592220e-01 4.79636461e-01 -6.87641144e-01 -1.81386948e-01 5.35209298e-01 9.73858058e-01 1.00336385e+00 4.23021853e-01 -4.67642218e-01 -4.92136896e-01 2.69856811e-01 3.35984468e-01 3.81956041e-01 1.12393248e+00 -5.23572624e-01 1.08791694e-01 7.72158802e-01 9.45738316e-01 -3.88688803e-01 -9.74128485e-01 -1.32443458e-01 3.32709402e-01 7.85131454e-02 -5.99592626e-02 -6.78636014e-01 -8.70008111e-01 1.68881699e-01 1.06833792e+00 1.08440995e+00 1.76579356e+00 -5.98882623e-02 1.26350641e+00 -2.70027429e-01 4.38770413e-01 -1.26410413e+00 -4.12816256e-01 -4.95312363e-02 4.93009716e-01 -9.67131853e-01 -1.21415770e-02 -1.20759889e-01 -6.31581917e-02 8.00161242e-01 5.02254248e-01 1.37892038e-01 1.05554676e+00 7.47666717e-01 -2.35157132e-01 3.21382172e-02 -8.40030849e-01 -3.15704465e-01 -2.21158028e-01 7.53901482e-01 7.35361502e-02 4.15406376e-01 -4.58085597e-01 6.04745507e-01 -9.10694003e-02 1.50555551e-01 2.65138000e-01 1.18125010e+00 -4.40466046e-01 -7.27551043e-01 -7.37955332e-01 8.68411958e-01 -6.40109062e-01 4.19586480e-01 4.20444101e-01 7.83045352e-01 2.24419102e-01 1.41898370e+00 2.91596472e-01 -2.00014919e-01 8.24127376e-01 2.93648958e-01 4.83615100e-02 -2.05090523e-01 -8.03455472e-01 3.48754764e-01 -2.28371024e-01 -7.31435776e-01 -3.89188498e-01 -7.53295660e-01 -1.53322887e+00 -3.75250936e-01 -6.19617701e-01 2.85312794e-02 4.91270840e-01 9.68852758e-01 4.25794661e-01 9.11850095e-01 1.09216046e+00 -7.94022083e-01 -2.70831287e-01 -1.28076434e+00 -8.21076035e-01 6.35060847e-01 2.04491958e-01 -7.52751231e-01 -6.17245317e-01 2.06696600e-01]
[6.021787166595459, 2.5990068912506104]
5344accc-e36f-4fb6-88d6-0a244f053111
compressing-low-precision-deep-neural
1709.06262
null
http://arxiv.org/abs/1709.06262v2
http://arxiv.org/pdf/1709.06262v2.pdf
Compressing Low Precision Deep Neural Networks Using Sparsity-Induced Regularization in Ternary Networks
A low precision deep neural network training technique for producing sparse, ternary neural networks is presented. The technique incorporates hard- ware implementation costs during training to achieve significant model compression for inference. Training involves three stages: network training using L2 regularization and a quantization threshold regularizer, quantization pruning, and finally retraining. Resulting networks achieve improved accuracy, reduced memory footprint and reduced computational complexity compared with conventional methods, on MNIST and CIFAR10 datasets. Our networks are up to 98% sparse and 5 & 11 times smaller than equivalent binary and ternary models, translating to significant resource and speed benefits for hardware implementations.
['Giulio Gambardella', 'Michaela Blott', 'Julian Faraone', 'Nicholas Fraser', 'Philip H. W. Leong']
2017-09-19
null
null
null
null
['l2-regularization']
['methodology']
[ 4.75701421e-01 3.33782107e-01 -6.53325737e-01 -7.94982553e-01 -3.43027800e-01 -3.75919268e-02 1.61450937e-01 2.47828677e-01 -7.67956376e-01 8.97395194e-01 -7.64848143e-02 -6.92115843e-01 -1.17779955e-01 -9.13186908e-01 -8.22392702e-01 -3.21078241e-01 -8.18861574e-02 3.52211565e-01 1.49278715e-02 3.12118709e-01 1.09062508e-01 5.76387465e-01 -1.58283389e+00 6.27977490e-01 4.03345734e-01 1.60205901e+00 -1.39233902e-01 7.58354008e-01 1.37542874e-01 1.30744159e+00 -6.31621599e-01 -4.62723345e-01 4.22131091e-01 1.15446016e-01 -5.38688600e-01 -4.60468441e-01 8.66142094e-01 -5.36763310e-01 -7.77611256e-01 1.11337137e+00 4.57435280e-01 1.22687861e-01 6.00044668e-01 -8.10491323e-01 -6.87209785e-01 1.05896604e+00 -1.53046772e-01 5.37521839e-01 -3.16176683e-01 -1.53819337e-01 9.00786102e-01 -7.90494680e-01 3.13204378e-02 1.33610964e+00 1.19488788e+00 5.37656009e-01 -1.13127685e+00 -1.04433906e+00 -3.36364686e-01 -2.35339832e-02 -1.66413367e+00 -9.37998235e-01 1.64720193e-01 1.43112332e-01 1.72470784e+00 -1.12606240e-02 7.02776611e-01 7.84231544e-01 1.76226422e-01 5.33175290e-01 5.02624810e-01 -5.59051633e-01 5.13429344e-01 5.29788919e-02 4.84461039e-01 9.46283996e-01 7.86117315e-01 2.51454860e-02 -5.35036802e-01 9.18582231e-02 1.08971250e+00 -5.52220456e-02 1.92346871e-01 4.48611319e-01 -6.22466028e-01 8.90664220e-01 6.53811932e-01 4.58479784e-02 -4.96438622e-01 7.84483254e-01 5.54225624e-01 3.50776434e-01 6.75233901e-02 2.02862531e-01 -3.52880627e-01 -1.92223310e-01 -1.40805745e+00 2.71630168e-01 8.08625996e-01 1.20148587e+00 4.67150867e-01 1.02543640e+00 -1.35637838e-02 8.11440051e-01 3.98772150e-01 4.89533842e-01 8.18264246e-01 -1.16139615e+00 6.75102413e-01 4.03163731e-01 -5.82662165e-01 -1.05784953e+00 -3.17029625e-01 -6.02161825e-01 -1.29295933e+00 -2.29501370e-02 -1.05631284e-01 -3.35974216e-01 -1.37471938e+00 1.31293893e+00 -2.71039546e-01 3.23772281e-01 3.30971807e-01 3.19580436e-01 9.64013219e-01 7.33361721e-01 2.53749579e-01 9.70806852e-02 1.14072800e+00 -7.48665810e-01 -7.94305086e-01 -4.60328072e-01 6.73196197e-01 -3.52835119e-01 6.70298219e-01 4.87799734e-01 -1.53872800e+00 -6.74156308e-01 -1.65655208e+00 -5.97273052e-01 -4.28349316e-01 4.13013309e-01 1.17884660e+00 9.89614964e-01 -1.07529378e+00 8.53782177e-01 -1.02238572e+00 2.86010265e-01 1.05940914e+00 1.13205552e+00 9.99269262e-02 -1.00760370e-01 -1.23070693e+00 8.79519224e-01 9.77229714e-01 3.74086350e-02 -4.62293416e-01 -7.68248379e-01 -1.12618709e+00 4.90960091e-01 -5.30987263e-01 -3.94143581e-01 1.48225081e+00 -6.26359403e-01 -1.50960732e+00 2.75723159e-01 -3.09246480e-01 -1.56818056e+00 -2.58720636e-01 3.70170772e-02 -5.86070359e-01 2.16322154e-01 -4.80820179e-01 1.14943266e+00 6.12655401e-01 -2.52064019e-01 -7.40878940e-01 -1.05160370e-01 -3.96261692e-01 1.33574568e-02 -5.70548058e-01 -3.55318815e-01 -2.39188522e-01 -6.38909698e-01 4.33124036e-01 -4.91966248e-01 -2.88120031e-01 -1.65123399e-02 -6.78040609e-02 -1.26004010e-01 8.37133408e-01 -5.08811951e-01 1.42542458e+00 -1.86605322e+00 -6.60085559e-01 5.34457564e-01 2.46210694e-01 7.32376516e-01 7.19963834e-02 -4.07665133e-01 8.27384144e-02 1.81477353e-01 9.45056677e-02 -4.12378192e-01 -1.13376498e-01 5.60921788e-01 -2.81994104e-01 3.12319815e-01 3.45866621e-01 1.02779222e+00 -3.60591829e-01 -5.07490814e-01 2.92872220e-01 8.31971705e-01 -8.91263485e-01 -4.96375769e-01 -4.37071882e-02 -5.01192331e-01 8.05643871e-02 9.15418029e-01 3.96810174e-01 -3.89709711e-01 1.49851471e-01 -3.59215736e-01 2.45937407e-01 8.52445662e-01 -9.12633657e-01 1.28746343e+00 -3.76330256e-01 1.05199420e+00 -8.66369307e-02 -1.18704188e+00 1.12058008e+00 3.81438613e-01 -1.13264732e-01 -9.83673692e-01 6.94488883e-01 2.60935932e-01 1.94213897e-01 5.24769761e-02 5.96272528e-01 3.72616109e-03 1.22123711e-01 -1.00390606e-01 5.04446208e-01 -2.09610593e-02 3.29378158e-01 5.81680797e-03 8.21248531e-01 -5.69404662e-01 4.45389077e-02 -1.45592615e-01 1.95536222e-02 -4.96739484e-02 5.27587414e-01 7.42052376e-01 -9.23998132e-02 3.00602794e-01 3.75516444e-01 -7.04484403e-01 -1.23046100e+00 -7.16421306e-01 -4.85929906e-01 9.31182623e-01 -4.80123937e-01 -4.24207538e-01 -7.78478086e-01 1.74794942e-01 1.90132156e-01 6.11910462e-01 -3.97447318e-01 -2.68693060e-01 -7.30891168e-01 -7.17295110e-01 1.23632562e+00 8.86075914e-01 8.81584287e-01 -7.98407137e-01 -6.45645618e-01 2.75078386e-01 4.47136253e-01 -1.30592740e+00 3.23651657e-02 1.01382565e+00 -1.61676872e+00 -4.98932362e-01 -1.26034021e-01 -1.17063713e+00 5.69156170e-01 -3.64771873e-01 1.00023842e+00 3.03500835e-02 -3.14370573e-01 -7.47399211e-01 2.51990855e-01 -5.80136657e-01 -1.33587345e-01 1.28253415e-01 3.85200143e-01 -7.37816989e-01 7.30532169e-01 -7.81788409e-01 -3.97226602e-01 -4.87847567e-01 -6.28874242e-01 -8.99734497e-02 1.12662685e+00 9.86878395e-01 9.27926540e-01 2.80727208e-01 5.57099640e-01 -8.95742655e-01 6.68790221e-01 -1.73451439e-01 -7.23796785e-01 -3.81998718e-02 -9.43742871e-01 3.10265094e-01 7.44249165e-01 -5.22736490e-01 -5.99067926e-01 3.03354770e-01 -3.15727085e-01 -6.16335988e-01 2.11420238e-01 5.93593776e-01 3.12629342e-01 -3.17336798e-01 9.77983773e-01 4.92457561e-02 7.29294540e-03 -2.83602148e-01 1.59761727e-01 7.45288134e-01 8.45749140e-01 -1.19039737e-01 2.62722790e-01 -1.13048125e-02 1.59731641e-01 -7.81587422e-01 -6.38478100e-01 1.76253557e-01 -6.75746024e-01 5.10793269e-01 5.00050843e-01 -1.22423625e+00 -6.58321738e-01 8.86876807e-02 -9.43340361e-01 -3.59834433e-01 -4.11547840e-01 6.80405080e-01 6.82125147e-03 -2.62901157e-01 -1.09170163e+00 -4.95225102e-01 -1.01486325e+00 -1.02367115e+00 5.39679408e-01 3.62470835e-01 -3.15259367e-01 -9.69581008e-01 -6.22621298e-01 -8.23434517e-02 7.19688475e-01 -1.64198756e-01 1.01150632e+00 -6.30843222e-01 -4.08174604e-01 -6.26666665e-01 -6.07164860e-01 6.45555794e-01 -2.32427999e-01 -1.17069587e-01 -9.60487485e-01 -1.04459181e-01 -2.46897161e-01 -6.62743986e-01 9.31747019e-01 7.51142979e-01 1.45405364e+00 -7.31347919e-01 -1.72243223e-01 1.05760920e+00 1.29173243e+00 3.14376384e-01 6.84275091e-01 -8.08523670e-02 4.47262913e-01 -2.98571020e-01 -1.75985899e-02 3.78559411e-01 1.77777105e-03 -1.38167944e-02 1.12204969e-01 1.21351471e-03 -2.85962731e-01 -2.22282857e-01 -1.13772810e-01 1.00516236e+00 1.76798970e-01 9.90766808e-02 -9.47307289e-01 4.74590421e-01 -1.27369690e+00 -9.26781774e-01 1.46607026e-01 1.89483738e+00 1.19107187e+00 6.80007100e-01 -4.39744204e-01 7.67730534e-01 2.21422017e-01 -1.10398717e-01 -5.93902171e-01 -9.54955041e-01 -4.84667532e-02 1.02690077e+00 1.14183009e+00 4.76965696e-01 -1.02828777e+00 1.00197542e+00 7.81156588e+00 8.03865910e-01 -1.23091757e+00 -4.03246582e-02 1.13028991e+00 -5.63777387e-01 1.97751105e-01 -6.22852027e-01 -1.41162229e+00 1.37119532e-01 1.88954604e+00 5.19484840e-02 3.45832556e-01 1.21086752e+00 -2.00936213e-01 1.03613019e-01 -1.05506706e+00 1.29066396e+00 -1.60033852e-01 -1.88976455e+00 3.51864755e-01 -2.14315757e-01 7.66497374e-01 3.19780171e-01 3.16080332e-01 6.01157427e-01 4.30749178e-01 -1.62955809e+00 5.56359410e-01 1.18146308e-01 1.19849062e+00 -1.31645894e+00 7.35439599e-01 8.33110884e-02 -1.08226609e+00 -3.56907308e-01 -1.00618303e+00 -4.67917353e-01 -3.21754903e-01 6.13532007e-01 -1.01093054e+00 -3.96518975e-01 7.64011264e-01 5.56119978e-01 -3.95387679e-01 6.74717128e-01 1.71087608e-01 8.42990458e-01 -6.97075069e-01 -3.01039815e-01 3.79543453e-01 2.65990257e-01 -4.75411385e-01 1.39859962e+00 1.75543904e-01 2.87815005e-01 -2.45955572e-01 6.17837489e-01 -5.55090725e-01 -4.76181984e-01 -4.61068183e-01 -1.64963365e-01 1.11366236e+00 8.00449431e-01 -5.80895782e-01 -7.99476564e-01 -1.24272302e-01 5.82908213e-01 4.91514683e-01 2.12797746e-01 -7.99076617e-01 -1.00714636e+00 4.63407874e-01 -1.61981449e-01 5.40870368e-01 -1.36617824e-01 -1.18260789e+00 -6.05735123e-01 -2.58787334e-01 -8.21027040e-01 1.82017069e-02 -3.20455313e-01 -4.76304531e-01 7.62907863e-01 -2.99240977e-01 -6.76621735e-01 -4.35107321e-01 -8.87354493e-01 -2.00988039e-01 9.57737923e-01 -1.36880565e+00 -6.77593708e-01 -1.57676876e-01 4.26617563e-01 3.03494722e-01 -4.67021346e-01 1.00710940e+00 7.46638060e-01 -7.47103989e-01 1.21013248e+00 6.26163185e-02 3.31375360e-01 -9.56944674e-02 -8.89515638e-01 6.22000337e-01 7.28866339e-01 9.27391723e-02 1.03771067e+00 1.16213992e-01 -4.14484233e-01 -1.48177981e+00 -1.55822182e+00 1.06589496e+00 1.76447332e-01 3.68199766e-01 -4.50842381e-01 -8.59731674e-01 9.21047330e-01 -1.37658685e-01 2.87294954e-01 6.65765762e-01 1.31980712e-02 -3.90911311e-01 -3.28776717e-01 -1.41674232e+00 4.39724118e-01 6.73357069e-01 -5.35043836e-01 -4.33421254e-01 3.87919873e-01 7.58344591e-01 -7.37069964e-01 -1.04803491e+00 5.10705173e-01 7.53200114e-01 -4.94972706e-01 1.13451135e+00 -3.09650630e-01 4.90236551e-01 2.96169549e-01 -4.29557443e-01 -5.34334004e-01 -3.27307045e-01 -3.45937967e-01 -8.22259784e-01 6.07462525e-01 8.36806238e-01 -3.83368105e-01 1.25548446e+00 7.75643706e-01 -1.54508069e-01 -1.02657676e+00 -9.98006940e-01 -5.09465933e-01 3.56094018e-02 -6.31233931e-01 4.60225224e-01 6.38913155e-01 9.46505461e-03 4.09115314e-01 -1.46265119e-01 8.14957917e-02 5.86879671e-01 -5.35970688e-01 -1.60100292e-02 -1.15140438e+00 -8.51373300e-02 -6.14791930e-01 -5.86043596e-01 -1.34071732e+00 2.37558503e-02 -8.95347893e-01 -1.70965776e-01 -1.36583233e+00 -4.35063869e-01 -7.59727895e-01 -3.18292469e-01 9.01472390e-01 4.48976994e-01 8.23202133e-01 -3.01178813e-01 9.83435810e-02 -3.93046558e-01 2.23488614e-01 6.70454383e-01 -1.30304649e-01 -1.37827426e-01 -1.61732137e-01 -5.64401686e-01 7.02157259e-01 1.03124917e+00 -3.70834470e-01 -6.85244441e-01 -8.59601557e-01 -8.72757435e-02 -2.17007287e-02 6.64486587e-02 -1.83322072e+00 6.71361208e-01 3.51875544e-01 1.07229710e+00 -8.79648983e-01 6.99295938e-01 -6.94720864e-01 -2.03281298e-01 9.07823503e-01 -6.32839322e-01 2.06881836e-01 6.76566422e-01 1.89390287e-01 -3.84470999e-01 -4.14732665e-01 1.11136830e+00 6.27491847e-02 -5.33172607e-01 3.15580904e-01 -4.80965763e-01 -2.78899670e-01 4.71744210e-01 -3.90385002e-01 -2.01112539e-01 -5.07306866e-02 -6.21198237e-01 -1.49170920e-01 -3.14350992e-01 -1.29910767e-01 1.04230237e+00 -1.33040500e+00 -3.66376013e-01 7.10341811e-01 -7.65957952e-01 3.55953872e-01 -3.00885051e-01 4.03151721e-01 -1.07947516e+00 1.21144581e+00 -2.20269769e-01 -4.19741184e-01 -1.23711979e+00 -4.76053767e-02 4.94369328e-01 -7.61858234e-03 -6.69737279e-01 1.47676516e+00 -8.05274785e-01 -7.65050426e-02 9.87318218e-01 -1.12873459e+00 1.25641124e-02 -3.15292805e-01 8.65234017e-01 4.24365520e-01 4.98754621e-01 -2.16508940e-01 -9.64788124e-02 3.96387018e-02 -2.50846177e-01 -7.41746556e-03 1.13291407e+00 3.69814426e-01 9.40476954e-02 1.62149131e-01 1.53047788e+00 -7.54036367e-01 -8.86445403e-01 -4.47756648e-01 7.12208599e-02 -1.14155596e-03 8.12872112e-01 -7.05314934e-01 -1.33572567e+00 9.40302908e-01 7.69614220e-01 -1.99696392e-01 1.10053587e+00 -6.34833753e-01 1.04712391e+00 1.32585502e+00 3.05581875e-02 -1.10644829e+00 -3.47227484e-01 1.05958962e+00 2.98003495e-01 -9.87301707e-01 3.76320928e-01 -9.36155990e-02 -1.23189420e-01 1.18129754e+00 5.67801118e-01 -4.96577889e-01 8.81685972e-01 9.08065200e-01 -1.94472238e-01 -2.60084085e-02 -1.07552934e+00 3.31029326e-01 2.20730379e-01 5.97891092e-01 6.21154308e-01 -6.21511191e-02 1.74813837e-01 5.62876880e-01 -7.58950770e-01 3.77675265e-01 -2.84367818e-02 1.05533051e+00 -8.15450430e-01 -5.14980614e-01 6.49034306e-02 1.12145197e+00 -7.25888491e-01 -7.43584394e-01 1.30955145e-01 6.77355111e-01 1.02083251e-01 6.87860727e-01 6.30600095e-01 -5.71109533e-01 2.67589223e-02 1.41551450e-01 4.79921192e-01 -4.76229280e-01 -7.05160320e-01 -2.58357942e-01 1.27787143e-01 -4.13159996e-01 -2.68438049e-02 -1.23096220e-01 -1.63746560e+00 -7.00967908e-01 -3.79444391e-01 -5.52229695e-02 9.27945435e-01 7.29620278e-01 5.43184757e-01 8.45205784e-01 -6.21639974e-02 -8.25233161e-01 -6.97106004e-01 -1.18700314e+00 -3.13782573e-01 -2.68971413e-01 3.14223647e-01 -2.95722723e-01 -1.69409841e-01 1.64692074e-01]
[8.533129692077637, 3.041383981704712]
abcb38d5-c9e9-4753-9942-712327298dff
dual-stream-computer-generated-image
2207.03205
null
https://arxiv.org/abs/2207.03205v1
https://arxiv.org/pdf/2207.03205v1.pdf
Dual Stream Computer-Generated Image Detection Network Based On Channel Joint And Softpool
With the development of computer graphics technology, the images synthesized by computer software become more and more closer to the photographs. While computer graphics technology brings us a grand visual feast in the field of games and movies, it may also be utilized by someone with bad intentions to guide public opinions and cause political crisis or social unrest. Therefore, how to distinguish the computer-generated graphics (CG) from the photographs (PG) has become an important topic in the field of digital image forensics. This paper proposes a dual stream convolutional neural network based on channel joint and softpool. The proposed network architecture includes a residual module for extracting image noise information and a joint channel information extraction module for capturing the shallow semantic information of image. In addition, we also design a residual structure to enhance feature extraction and reduce the loss of information in residual flow. The joint channel information extraction module can obtain the shallow semantic information of the input image which can be used as the information supplement block of the residual module. The whole network uses SoftPool to reduce the information loss of down-sampling for image. Finally, we fuse the two flows to get the classification results. Experiments on SPL2018 and DsTok show that the proposed method outperforms existing methods, especially on the DsTok dataset. For example, the performance of our model surpasses the state-of-the-art by a large margin of 3%.
['Weiqi Luo', 'Hao Lin', 'Ziyi Xi']
2022-07-07
null
null
null
null
['image-forensics']
['computer-vision']
[ 2.14425191e-01 -3.56513023e-01 1.44945653e-02 -1.38780624e-01 -5.05793691e-01 -2.15473324e-01 3.67357254e-01 -1.74205303e-01 -4.68559295e-01 4.03295875e-01 2.44389161e-01 -4.09774512e-01 3.85749549e-01 -1.08594823e+00 -4.12878036e-01 -8.20078194e-01 4.14317071e-01 -5.54977000e-01 5.91400206e-01 -6.98696589e-03 5.93975067e-01 4.83804762e-01 -1.34477103e+00 6.82378292e-01 6.88539565e-01 1.40378225e+00 2.50066757e-01 4.00512993e-01 -5.65419257e-01 1.19287264e+00 -7.09729493e-01 -5.05252063e-01 4.47881460e-01 -5.32823741e-01 -4.33590204e-01 8.27484503e-02 2.33845532e-01 -7.82937825e-01 -7.23568082e-01 1.42679191e+00 7.61116147e-01 -1.92151591e-01 1.40022486e-01 -1.21594751e+00 -4.66937065e-01 4.82941270e-01 -1.01200223e+00 5.32481551e-01 7.54342899e-02 2.77469873e-01 2.37369120e-01 -6.46108508e-01 4.85169172e-01 1.44459963e+00 5.78827143e-01 1.71513081e-01 -6.30834341e-01 -1.07291687e+00 -1.59790814e-01 6.82935834e-01 -1.25522840e+00 -4.65911448e-01 1.02862012e+00 -2.98925281e-01 4.60026056e-01 8.43898654e-02 5.73066592e-01 6.15754545e-01 2.89746940e-01 9.76503193e-01 1.05099237e+00 -1.68128833e-01 -5.48692159e-02 3.45187336e-01 1.70048311e-01 7.46526062e-01 2.85825312e-01 -8.01386833e-02 -5.48667490e-01 1.08148277e-01 7.78634608e-01 2.21205670e-02 -6.50684357e-01 3.89504582e-01 -6.62688792e-01 6.96474314e-01 3.24061692e-01 1.12585992e-01 -2.45994538e-01 1.34202629e-01 4.84301031e-01 1.87085256e-01 2.99288034e-01 -1.42823219e-01 -7.42148003e-03 -2.00774178e-01 -9.46144998e-01 -1.01078950e-01 5.40364861e-01 6.82375968e-01 7.51391590e-01 3.34790021e-01 -1.01065956e-01 6.51519060e-01 2.06884801e-01 4.82504308e-01 4.57840353e-01 -8.89587402e-01 6.28783882e-01 7.43325412e-01 -2.49230057e-01 -1.54799652e+00 -6.52353466e-02 -3.15935612e-01 -9.93448853e-01 3.48659188e-01 4.16318059e-01 -1.87868580e-01 -6.84063375e-01 1.12537134e+00 2.35248819e-01 3.37927878e-01 -7.68720955e-02 9.82636034e-01 1.03322983e+00 1.12314296e+00 -1.78883344e-01 -1.58043876e-01 1.49126375e+00 -8.23799908e-01 -8.36190939e-01 -3.01141411e-01 2.19736934e-01 -9.57852125e-01 7.38505125e-01 7.08355367e-01 -8.71471703e-01 -7.17725575e-01 -1.19261074e+00 -8.20717439e-02 -9.78646576e-02 3.38314265e-01 4.22420382e-01 9.17637706e-01 -5.23015201e-01 4.14072096e-01 -5.26149154e-01 -5.63311838e-02 8.67486119e-01 8.15437660e-02 -2.47757658e-01 -3.31503987e-01 -1.15093565e+00 5.22088647e-01 2.86083639e-01 3.12447637e-01 -5.41668892e-01 -6.22172415e-01 -7.45739102e-01 1.25920430e-01 4.07531887e-01 -6.05216809e-02 7.15391099e-01 -9.75297570e-01 -1.31997621e+00 5.27237475e-01 9.35291648e-02 -2.48608425e-01 5.40824234e-01 2.52531301e-02 -5.86355448e-01 5.08390903e-01 2.40565874e-02 4.18540210e-01 9.38637674e-01 -8.80881488e-01 -9.46280539e-01 -2.66775250e-01 -1.27542630e-01 5.33589348e-02 -4.51852947e-01 1.97992623e-01 -8.78467441e-01 -5.82354486e-01 -6.39476404e-02 -4.21350420e-01 3.37138190e-03 1.79808512e-01 -4.80564356e-01 1.62991956e-01 1.28676391e+00 -1.00749457e+00 1.31498909e+00 -2.37918448e+00 -5.07609308e-01 5.76888658e-02 1.40375406e-01 5.46490312e-01 1.62252754e-01 1.75327942e-01 -3.81151102e-02 1.12022229e-01 -1.98815078e-01 1.93711575e-02 -4.29454267e-01 -2.48904392e-01 -3.41640234e-01 6.55980170e-01 1.00620136e-01 5.19118369e-01 -6.30669713e-01 -5.32227516e-01 3.61103773e-01 4.49850798e-01 -3.23365599e-01 1.56267464e-01 4.36514646e-01 1.96196586e-01 -5.45664191e-01 4.87550557e-01 1.24923754e+00 -1.18771069e-01 -1.88726813e-01 -5.64450979e-01 -2.10923791e-01 -2.48227715e-02 -1.42954683e+00 1.39224970e+00 -2.73924589e-01 9.10087764e-01 1.76393196e-01 -9.13540602e-01 1.03342295e+00 2.80295461e-02 2.19088301e-01 -1.03715050e+00 5.19536197e-01 7.33040795e-02 -5.16639873e-02 -7.52433658e-01 4.98848110e-01 -8.97280797e-02 1.19179571e-02 4.21307683e-01 -2.71645844e-01 2.16662943e-01 -4.32860777e-02 3.68177444e-01 9.67107236e-01 -2.75439084e-01 -5.27677014e-02 -1.17143102e-01 9.50380981e-01 -1.61709413e-01 8.58831823e-01 2.95203328e-01 -1.98119119e-01 6.15821898e-01 7.39897490e-01 -4.71439749e-01 -9.77005899e-01 -6.35619283e-01 4.71692979e-02 2.40682408e-01 7.43530393e-01 -2.23096356e-01 -8.45113695e-01 -6.87255323e-01 -1.74144417e-01 4.06376511e-01 -2.78680682e-01 -3.42791706e-01 -7.16753125e-01 -6.66477382e-01 7.18641460e-01 3.04626435e-01 1.47219205e+00 -1.13041019e+00 -4.52099651e-01 1.52494401e-01 -3.04238677e-01 -1.21947765e+00 -5.30400574e-01 -3.49207312e-01 -5.14062166e-01 -1.27464890e+00 -6.39677405e-01 -6.50134206e-01 4.15223897e-01 5.83847761e-01 5.04086673e-01 2.41935328e-01 -3.22506487e-01 -1.49917170e-01 -3.30156803e-01 -2.31050476e-01 -3.08657587e-01 -2.82311410e-01 -4.35059339e-01 4.79097396e-01 3.67821097e-01 -3.47299755e-01 -9.12167490e-01 1.35152310e-01 -1.29230642e+00 1.27844587e-01 6.01586401e-01 5.81649423e-01 1.94385275e-01 6.28505588e-01 3.76064301e-01 -8.56171370e-01 6.91671431e-01 -4.48892921e-01 -5.51627457e-01 -7.58066773e-02 -4.64161515e-01 -1.12611063e-01 8.54580998e-01 -2.24726185e-01 -1.43298578e+00 -2.34406114e-01 -1.30472645e-01 -3.42122257e-01 5.68366051e-02 2.53480494e-01 -6.46906853e-01 -1.05224743e-01 2.24874035e-01 3.21569383e-01 9.46812034e-02 -5.83661973e-01 9.40917358e-02 1.07607615e+00 6.02822125e-01 -9.69033688e-02 5.91524184e-01 7.88777947e-01 -2.23707873e-02 -9.65421736e-01 -3.11556697e-01 -2.67135233e-01 -1.32499561e-02 -4.85316575e-01 8.72995019e-01 -8.96328449e-01 -7.85143256e-01 1.03236282e+00 -1.30086017e+00 8.21100548e-02 7.76424631e-02 3.08991551e-01 1.10103309e-01 9.14263487e-01 -8.40485454e-01 -6.98632777e-01 -4.10625815e-01 -1.51836991e+00 7.61655867e-01 6.15311086e-01 4.40031648e-01 -5.88600695e-01 -4.11928207e-01 4.03545380e-01 3.17786723e-01 1.21983290e-01 6.45960271e-01 -2.32178926e-01 -9.11668777e-01 -2.81086445e-01 -8.11514139e-01 7.29522288e-01 2.56497152e-02 -1.74738899e-01 -1.07714653e+00 -1.82401072e-02 3.04095149e-01 -1.74405742e-02 1.08654118e+00 1.81357145e-01 1.32109296e+00 -3.07129055e-01 -1.04083620e-01 8.48465443e-01 1.71467662e+00 5.99280179e-01 1.33921766e+00 5.01694620e-01 7.17572331e-01 4.88191813e-01 3.53037804e-01 4.02321279e-01 1.78761840e-01 4.02902693e-01 3.22649360e-01 -1.56695768e-01 -4.89926279e-01 -2.15575323e-01 5.01610756e-01 6.90858603e-01 3.84479076e-01 -3.75057757e-01 -6.51926517e-01 4.79590744e-01 -1.64405358e+00 -1.12271130e+00 -5.09413481e-01 1.98415291e+00 5.59670389e-01 2.24610999e-01 -2.96117276e-01 3.78760755e-01 1.02034724e+00 3.49465638e-01 -3.91637355e-01 -3.70536298e-01 -2.45088130e-01 1.39566839e-01 8.06747496e-01 2.26325527e-01 -9.45488274e-01 7.39564300e-01 4.88201523e+00 1.47963583e+00 -1.36655986e+00 9.16494578e-02 9.42402422e-01 2.60837615e-01 -7.64425024e-02 1.94680523e-02 -6.89604461e-01 1.04735553e+00 5.45089006e-01 -6.83870772e-03 1.85997650e-01 7.76290357e-01 3.94107461e-01 -5.74244440e-01 -3.43061030e-01 1.41010058e+00 1.69011578e-01 -1.48168123e+00 3.33604589e-02 1.64350286e-01 4.18274641e-01 -3.67939085e-01 -1.03242947e-02 -1.48488075e-01 -1.33940056e-01 -7.58489430e-01 7.15801418e-01 4.78000700e-01 8.18920135e-01 -9.48792994e-01 1.01496220e+00 2.73955941e-01 -1.19229567e+00 -1.35423079e-01 -4.29281622e-01 -4.28971834e-02 3.47436458e-01 8.64979804e-01 -2.88468808e-01 5.14813900e-01 7.49230802e-01 6.86230361e-01 -5.74445844e-01 1.31682956e+00 -3.22137624e-01 6.50736690e-01 -1.33134067e-01 2.19252586e-01 3.15433681e-01 -2.27217495e-01 3.90223086e-01 1.14193344e+00 3.40189993e-01 1.94474354e-01 -9.65490863e-02 6.13097966e-01 -1.61684602e-01 1.17920354e-01 -4.50915158e-01 8.91104937e-02 4.28954691e-01 1.42799330e+00 -1.06162024e+00 -4.32944715e-01 -5.17383695e-01 1.09158266e+00 -2.62887895e-01 1.65817156e-01 -7.74649799e-01 -9.29256797e-01 4.28940356e-01 2.44850561e-01 3.40824246e-01 6.82515278e-02 -3.98700356e-01 -1.06744790e+00 5.97992539e-02 -8.27266455e-01 2.02158093e-01 -7.75644600e-01 -9.70800161e-01 6.63759768e-01 -3.77184510e-01 -1.34638202e+00 2.84704328e-01 -4.99901444e-01 -8.97215843e-01 9.32599604e-01 -1.68253803e+00 -7.14194775e-01 -7.03972399e-01 6.06117785e-01 5.01956165e-01 -2.33291358e-01 4.81710248e-02 7.15299249e-01 -7.21753538e-01 4.75940406e-01 1.06071115e-01 6.08183980e-01 7.34622359e-01 -6.17834568e-01 2.85170674e-01 1.26896179e+00 -3.65787297e-01 1.98645279e-01 3.27108949e-01 -8.34475756e-01 -1.17607152e+00 -9.49718773e-01 3.02238673e-01 3.20202142e-01 2.64828622e-01 -1.49266526e-01 -8.47788572e-01 9.00301561e-02 3.58113319e-01 8.70515313e-03 3.09806347e-01 -8.69269371e-01 -2.05760777e-01 -4.35120702e-01 -1.28029799e+00 2.69806117e-01 5.91039896e-01 -4.55408096e-01 -8.59620348e-02 -2.06277311e-01 6.11332238e-01 -1.28634617e-01 -3.56595695e-01 6.80458993e-02 5.30627251e-01 -1.32508659e+00 7.75308669e-01 -1.90604609e-02 6.68988526e-01 -5.52618802e-01 3.62919457e-02 -9.72065091e-01 -7.95057714e-02 -5.63294649e-01 3.13980371e-01 1.47914112e+00 -5.09558283e-02 -4.00097698e-01 1.06747496e+00 4.93096381e-01 -5.45966066e-02 -4.69932318e-01 -7.59985745e-01 -4.44148302e-01 -2.58907348e-01 -6.48373425e-01 5.95457792e-01 7.53258526e-01 -1.39202401e-01 2.03796148e-01 -5.80435395e-01 -3.50576006e-02 7.69790113e-01 -6.65026233e-02 7.29355335e-01 -7.50332117e-01 -1.26742139e-01 -4.56244141e-01 -6.19700670e-01 -1.15073121e+00 -1.29137814e-01 -5.59936643e-01 -2.03303814e-01 -1.44470525e+00 2.09050581e-01 -2.57920712e-01 1.21468104e-01 1.21717773e-01 -2.44965941e-01 4.94449586e-01 4.91774052e-01 1.51421487e-01 -3.43428850e-01 5.06458640e-01 1.40009093e+00 -1.90695912e-01 -5.03963344e-02 -2.25575730e-01 -9.12042320e-01 8.20699215e-01 7.05888689e-01 -4.07556266e-01 -3.20526719e-01 -4.34250265e-01 9.06301439e-02 1.36157170e-01 3.22928697e-01 -1.04080260e+00 4.80152726e-01 9.71371606e-02 6.01043999e-01 -6.51838303e-01 2.22946689e-01 -8.75295937e-01 4.39905077e-02 4.81617749e-01 7.25866631e-02 -2.55949259e-01 2.64410406e-01 5.50320864e-01 -3.90880466e-01 -3.12414259e-01 9.38743651e-01 -2.33123779e-01 -8.65990520e-01 3.13780606e-01 -5.61408699e-01 -5.97008355e-02 9.18273509e-01 -5.08756161e-01 -6.99556291e-01 -5.12356520e-01 -9.82867181e-02 1.72102392e-01 3.09942216e-01 2.85262525e-01 9.77301478e-01 -1.36923385e+00 -6.05228841e-01 3.79864335e-01 -2.95510501e-01 -1.06462687e-01 6.99866652e-01 6.82209671e-01 -9.69136298e-01 1.37034819e-01 -4.24823105e-01 -2.03126758e-01 -1.19022346e+00 3.25914383e-01 1.23712465e-01 -1.09336019e-01 -7.11483538e-01 7.01744258e-01 3.04824978e-01 1.85628235e-01 1.20875627e-01 -2.34849546e-02 -3.85394633e-01 2.19404861e-01 1.18310034e+00 7.60012090e-01 1.57752503e-02 -8.18245947e-01 -3.03597271e-01 3.61749977e-01 -8.01285133e-02 7.41502410e-03 1.35390449e+00 -2.62168974e-01 -3.07940990e-01 -1.82135001e-01 1.56729734e+00 1.74822032e-01 -1.17799342e+00 -1.50877133e-01 -3.80420655e-01 -9.68802512e-01 4.78534073e-01 -5.87286532e-01 -1.76903164e+00 1.25012493e+00 6.63242459e-01 1.58558637e-01 1.42173088e+00 -4.59146440e-01 1.32632339e+00 -2.02347279e-01 1.84329405e-01 -1.00630319e+00 6.43034503e-02 -8.33430365e-02 5.01596272e-01 -1.01911604e+00 1.01574890e-01 -5.63692331e-01 -6.43485725e-01 1.41756117e+00 5.82355917e-01 -3.50592345e-01 6.94972694e-01 5.26839852e-01 7.56281763e-02 -6.59274012e-02 -2.04430059e-01 -4.13464271e-02 -4.37549092e-02 5.96384406e-01 -9.50967669e-02 -4.29153711e-01 -3.99330378e-01 7.94200540e-01 1.17191568e-01 9.62007195e-02 1.00237143e+00 6.98089302e-01 -6.27871454e-01 -9.89330471e-01 -6.17233098e-01 5.08256555e-01 -8.52379560e-01 -1.83601648e-01 -1.93048805e-01 5.61520875e-01 4.25758243e-01 1.07837427e+00 1.58796683e-01 -6.40602112e-01 1.15273021e-01 -3.15915585e-01 1.03616469e-01 -9.78307948e-02 -5.09331048e-01 1.37438387e-01 8.29277933e-02 -8.41491938e-01 -3.08298707e-01 -4.70575660e-01 -1.15792131e+00 -6.18438721e-01 -4.27723050e-01 1.19076274e-01 6.91706896e-01 6.89716458e-01 2.72745043e-01 5.55038154e-01 7.09774137e-01 -6.10067844e-01 5.47306016e-02 -7.34455466e-01 -7.28759050e-01 3.06252509e-01 2.22531989e-01 -3.12989116e-01 -4.30440724e-01 7.81146213e-02]
[12.31151294708252, 0.829243540763855]
ea09db61-b42e-47cb-b298-6effa9100cbc
learning-grounded-vision-language
2303.06378
null
https://arxiv.org/abs/2303.06378v2
https://arxiv.org/pdf/2303.06378v2.pdf
Learning Grounded Vision-Language Representation for Versatile Understanding in Untrimmed Videos
Joint video-language learning has received increasing attention in recent years. However, existing works mainly focus on single or multiple trimmed video clips (events), which makes human-annotated event boundaries necessary during inference. To break away from the ties, we propose a grounded vision-language learning framework for untrimmed videos, which automatically detects informative events and effectively excavates the alignments between multi-sentence descriptions and corresponding event segments. Instead of coarse-level video-language alignments, we present two dual pretext tasks to encourage fine-grained segment-level alignments, i.e., text-to-event grounding (TEG) and event-to-text generation (ETG). TEG learns to adaptively ground the possible event proposals given a set of sentences by estimating the cross-modal distance in a joint semantic space. Meanwhile, ETG aims to reconstruct (generate) the matched texts given event proposals, encouraging the event representation to retain meaningful semantic information. To encourage accurate label assignment between the event set and the text set, we propose a novel semantic-aware cost to mitigate the sub-optimal matching results caused by ambiguous boundary annotations. Our framework is easily extensible to tasks covering visually-grounded language understanding and generation. We achieve state-of-the-art dense video captioning performance on ActivityNet Captions, YouCook2 and YouMakeup, and competitive performance on several other language generation and understanding tasks. Our method also achieved 1st place in both the MTVG and MDVC tasks of the PIC 4th Challenge. Our code is publicly available at https://github.com/zjr2000/GVL.
['Ping Luo', 'Ran Cheng', 'Wenhao Jiang', 'Feng Zheng', 'Jinrui Zhang', 'Teng Wang']
2023-03-11
null
null
null
null
['dense-video-captioning', 'natural-language-moment-retrieval']
['computer-vision', 'computer-vision']
[ 2.75193542e-01 2.47211590e-01 -3.30714762e-01 -4.96771783e-01 -1.41137409e+00 -4.56639141e-01 7.38586903e-01 -1.86156452e-01 -2.92908877e-01 7.65940309e-01 7.89921880e-01 -2.02997066e-02 2.69786239e-01 -5.34263492e-01 -1.08000302e+00 -5.05002797e-01 5.66227585e-02 5.44363678e-01 2.67094165e-01 3.54886353e-02 -5.12066623e-03 -1.16045035e-01 -1.44201422e+00 8.23948979e-01 7.68970728e-01 8.14465106e-01 5.77419460e-01 6.68279946e-01 -2.18861103e-01 1.01337945e+00 -4.73853350e-01 -4.54908013e-01 4.91597764e-02 -8.42091203e-01 -8.82906020e-01 3.58919322e-01 5.52326322e-01 -4.29670215e-01 -6.67195320e-01 9.05512989e-01 4.20922011e-01 2.90997356e-01 4.36313391e-01 -1.61750031e+00 -5.93192697e-01 9.17940736e-01 -7.38432050e-01 2.24527508e-01 6.49362326e-01 2.39454955e-01 1.18101168e+00 -1.11097145e+00 9.14687455e-01 1.20525932e+00 3.78738761e-01 5.43973923e-01 -8.34147632e-01 -6.98783517e-01 4.79957432e-01 5.28969109e-01 -1.41891026e+00 -6.37395322e-01 7.37875879e-01 -4.10433888e-01 8.65943849e-01 8.16343650e-02 6.52902305e-01 1.62630236e+00 -1.73849091e-01 1.15156567e+00 5.39028704e-01 -3.65254909e-01 1.00183703e-01 -3.14389825e-01 -2.84606934e-01 8.29722106e-01 -4.70355265e-02 -1.79118708e-01 -1.05310273e+00 3.05539258e-02 7.96650887e-01 -1.58915713e-01 -5.44128060e-01 -2.61472672e-01 -1.68309498e+00 7.97853172e-01 1.23810172e-01 2.06454530e-01 -5.60054898e-01 4.53334600e-01 4.71502453e-01 -3.18992376e-01 5.11989295e-01 1.90037951e-01 -2.31824115e-01 -3.04937512e-01 -1.19143343e+00 2.76367545e-01 5.37505627e-01 1.18592417e+00 6.91082537e-01 -3.89106721e-02 -7.17193127e-01 7.27704048e-01 4.31990683e-01 2.55054265e-01 3.30080032e-01 -1.14686644e+00 9.88730788e-01 2.30025247e-01 1.84925810e-01 -8.81142139e-01 -3.21872830e-02 -9.67186689e-02 -6.21142328e-01 -3.95868301e-01 3.21137339e-01 -1.50126621e-01 -9.68042314e-01 1.90327358e+00 3.10319245e-01 9.29645777e-01 -1.79880988e-02 1.16828310e+00 9.09729719e-01 1.12196362e+00 3.30927312e-01 -2.24433586e-01 1.47619367e+00 -1.27278233e+00 -7.39500761e-01 -4.96636480e-01 4.28432494e-01 -5.88713169e-01 1.08339095e+00 1.66658819e-01 -1.23456979e+00 -5.41016996e-01 -7.69284964e-01 -2.56006628e-01 1.19815491e-01 1.69521362e-01 3.90121400e-01 -1.39448881e-01 -1.02392042e+00 2.15983823e-01 -9.58293855e-01 -3.08186769e-01 5.42379081e-01 -1.94738537e-01 -3.03269744e-01 -1.43271416e-01 -1.37849903e+00 5.34726620e-01 6.66892231e-01 1.12990834e-01 -1.22004759e+00 -6.13983393e-01 -1.13990545e+00 -5.80124035e-02 6.74202979e-01 -8.79142463e-01 1.33471251e+00 -1.15886962e+00 -1.09322619e+00 1.00193846e+00 -3.99104595e-01 -6.92114532e-01 5.55764377e-01 -3.61105531e-01 -3.31543595e-01 6.14240766e-01 3.97122502e-01 1.25538385e+00 7.71024287e-01 -1.20333171e+00 -7.73550928e-01 1.05950147e-01 8.13282877e-02 4.54721212e-01 -7.23798126e-02 9.53142047e-02 -9.69457924e-01 -1.03775477e+00 -8.00143555e-02 -7.67798185e-01 -1.11113645e-01 9.51569676e-02 -5.03407359e-01 -3.25579524e-01 5.70368648e-01 -1.05794549e+00 1.14968896e+00 -2.00941348e+00 3.70888263e-01 -3.51475328e-01 1.69321045e-01 -6.53783008e-02 -4.14365470e-01 3.48755866e-01 -7.28114136e-03 2.04546424e-03 -2.75775760e-01 -6.69115782e-01 7.95828924e-02 2.60117978e-01 -6.09405339e-01 3.01831931e-01 2.19538987e-01 1.14252794e+00 -1.22088301e+00 -9.23053443e-01 2.54671544e-01 5.36457658e-01 -5.40891767e-01 3.28086615e-01 -5.58999479e-01 4.35918242e-01 -5.35754144e-01 5.16054511e-01 1.95697352e-01 -4.77836013e-01 1.82448085e-02 -4.20554966e-01 5.40789366e-02 3.32448572e-01 -9.20220792e-01 2.33272481e+00 -3.68442416e-01 7.93177128e-01 -2.13925824e-01 -1.02406788e+00 5.15772104e-01 5.94613433e-01 6.04752600e-01 -5.79240084e-01 -1.40320957e-01 -1.47431850e-01 -4.89168614e-01 -6.40318990e-01 6.62893593e-01 4.94145043e-02 -2.83140182e-01 2.31801867e-01 2.98281878e-01 1.18790984e-01 2.63895333e-01 6.09269083e-01 7.85708666e-01 7.87718296e-01 3.44702423e-01 1.34632975e-01 2.80264318e-01 5.79481795e-02 6.41603947e-01 6.10497952e-01 -1.23902701e-01 1.02742541e+00 5.24633527e-01 -1.53776392e-01 -9.92682993e-01 -1.07779455e+00 3.73850793e-01 1.16336596e+00 4.09556419e-01 -6.38744712e-01 -7.49942541e-01 -7.20865250e-01 -4.62766886e-01 1.04148805e+00 -4.22836483e-01 -4.89656068e-02 -7.55420864e-01 -4.18386817e-01 6.65056825e-01 7.46508181e-01 4.83044773e-01 -1.29058492e+00 -3.12894642e-01 2.83424288e-01 -1.18234289e+00 -1.50445414e+00 -1.00073302e+00 -3.22787136e-01 -5.10791481e-01 -9.47048903e-01 -6.54490173e-01 -9.17090952e-01 5.69608867e-01 2.31567293e-01 1.35957277e+00 -1.96830794e-01 -1.12406485e-01 4.41437751e-01 -5.98260581e-01 -1.06360428e-01 -3.55437815e-01 -1.70339644e-01 -2.41545498e-01 2.48357698e-01 2.04755366e-01 -2.38175437e-01 -6.90201998e-01 3.05214196e-01 -7.58290231e-01 7.47741342e-01 3.54528308e-01 7.32106864e-01 9.09382761e-01 -3.60604018e-01 6.38344765e-01 -3.52368951e-01 8.25267509e-02 -6.36981785e-01 -3.57462138e-01 3.32173377e-01 5.24009462e-04 -2.48204358e-02 4.62784439e-01 -4.84923333e-01 -1.17869604e+00 7.81802982e-02 -2.11188104e-02 -7.84749210e-01 -1.78054690e-01 3.90621841e-01 -3.16262782e-01 5.57732940e-01 3.67240399e-01 3.57538223e-01 -4.56622958e-01 -1.75398514e-01 6.09597504e-01 5.29791117e-01 8.51818919e-01 -6.87832594e-01 5.20903707e-01 4.54118639e-01 -4.68875051e-01 -5.71746707e-01 -1.21028101e+00 -4.47940707e-01 -3.78282398e-01 -3.53889465e-01 1.30813360e+00 -1.40952933e+00 -2.63544679e-01 2.16519400e-01 -1.48194563e+00 -5.23995042e-01 -3.14130664e-01 4.91542846e-01 -8.50497842e-01 5.38012862e-01 -6.29429102e-01 -5.47156990e-01 -3.54111522e-01 -1.10179734e+00 1.56443584e+00 8.24384615e-02 -3.37342352e-01 -8.40004981e-01 1.36170276e-02 6.75433874e-01 -2.98255414e-01 3.65607917e-01 4.46345598e-01 -4.22698975e-01 -8.61419976e-01 1.52134523e-01 -1.71335980e-01 5.74700274e-02 -1.33602977e-01 -1.44896939e-01 -8.38057458e-01 -1.69307172e-01 -3.43508929e-01 -4.48713779e-01 9.76558506e-01 5.43651938e-01 1.32721472e+00 -4.68670994e-01 -4.29426849e-01 6.31945193e-01 1.05746520e+00 1.44841224e-01 8.61368954e-01 1.18326895e-01 9.82654810e-01 4.64771599e-01 9.30876255e-01 5.24897218e-01 6.61374092e-01 1.00558376e+00 4.35473561e-01 -1.16835490e-01 -4.11422104e-01 -8.27470243e-01 6.38167977e-01 5.13936400e-01 3.22593637e-02 -7.49709070e-01 -6.98125780e-01 8.36997032e-01 -2.23627424e+00 -1.32610846e+00 7.84688741e-02 1.97164893e+00 9.43633735e-01 -1.02148205e-02 1.14763521e-01 -3.83434594e-01 1.05399096e+00 4.89544481e-01 -3.56021643e-01 2.28055790e-01 -1.58981994e-01 -2.12017316e-02 6.73847795e-02 5.43430686e-01 -1.19236815e+00 1.09048676e+00 4.94096613e+00 1.00996077e+00 -7.48863399e-01 3.90395939e-01 7.12197125e-01 -3.59839171e-01 -3.55081588e-01 3.18637751e-02 -8.45216572e-01 6.84498072e-01 7.84565568e-01 -1.56450078e-01 3.23051184e-01 6.19806767e-01 6.68251514e-01 3.12134791e-02 -1.27225614e+00 1.22411013e+00 3.36803675e-01 -1.70161676e+00 3.83585602e-01 -2.68655241e-01 7.38148451e-01 2.46099923e-02 -3.33827138e-01 3.30703020e-01 1.38472676e-01 -8.38380873e-01 1.23916614e+00 5.10466695e-01 8.85095477e-01 -4.78781253e-01 3.42298687e-01 1.39654160e-01 -1.52628016e+00 1.99617669e-01 -2.22277492e-02 2.91655004e-01 7.78274834e-01 3.68635863e-01 -7.24545717e-01 6.65159941e-01 5.07538080e-01 1.03140998e+00 -2.36138672e-01 9.07197952e-01 -4.28821892e-01 6.51961625e-01 -9.99482870e-02 2.94465661e-01 4.24704254e-01 -3.68640162e-02 7.41773725e-01 1.22769296e+00 5.23382306e-01 1.55979231e-01 5.09593189e-01 9.68362689e-01 -2.90755302e-01 -1.80646852e-01 -3.70117038e-01 -1.27895012e-01 6.07527316e-01 1.23445284e+00 -8.15972507e-01 -5.79303026e-01 -4.80989814e-01 1.34374475e+00 1.38025582e-01 4.72093433e-01 -1.40551245e+00 -1.21354900e-01 6.07376873e-01 1.69809967e-01 3.48724931e-01 -4.54499200e-02 6.47717565e-02 -1.48851943e+00 9.91942286e-02 -7.44357705e-01 6.41930997e-01 -1.26671934e+00 -9.94591177e-01 4.31287408e-01 2.38045737e-01 -1.35044909e+00 -5.11794329e-01 -1.03723377e-01 -5.34233570e-01 5.72111666e-01 -1.21478510e+00 -1.33155239e+00 -5.06371260e-01 6.11631334e-01 1.22610331e+00 7.43385106e-02 3.54835063e-01 4.20559525e-01 -4.88283604e-01 4.37890947e-01 -5.28454840e-01 3.19088399e-01 8.05490971e-01 -1.00859606e+00 5.76645732e-01 1.22298980e+00 5.83308101e-01 7.44155347e-02 6.78505957e-01 -7.91749835e-01 -1.07650399e+00 -1.52242053e+00 1.01613855e+00 -4.87950534e-01 6.16566420e-01 -5.25880992e-01 -8.17428946e-01 1.04420018e+00 2.64766842e-01 -3.77805121e-02 2.53946602e-01 -3.57520550e-01 -2.79801279e-01 1.53161183e-01 -6.46232128e-01 8.29877973e-01 1.49456763e+00 -5.77800035e-01 -6.74875140e-01 5.91550708e-01 1.02775335e+00 -6.89360023e-01 -4.18708861e-01 1.15661979e-01 2.28895947e-01 -6.91706300e-01 1.05380130e+00 -5.73787928e-01 7.78578699e-01 -5.00546336e-01 -1.60492420e-01 -1.03574169e+00 -1.37126356e-01 -8.34130943e-01 -2.89500535e-01 1.34473515e+00 2.70367533e-01 1.08989486e-02 6.95671737e-01 2.36388981e-01 -5.07892132e-01 -6.08937263e-01 -9.15060639e-01 -4.82578516e-01 -5.14705420e-01 -6.55418336e-01 3.60282600e-01 9.98170316e-01 -5.50443493e-02 4.27053869e-01 -7.09248126e-01 2.97777832e-01 6.19444847e-01 1.44159898e-01 5.34986496e-01 -6.42499566e-01 -4.52195287e-01 -2.85059065e-01 -1.64143354e-01 -1.36278725e+00 4.62383568e-01 -9.90042269e-01 4.11214739e-01 -1.84521914e+00 4.32431459e-01 1.03961220e-02 -8.14173296e-02 7.79856324e-01 -3.42352957e-01 4.41539794e-01 2.69686729e-01 1.28402174e-01 -1.20532322e+00 6.64698243e-01 1.22172952e+00 -7.57682025e-02 -1.46492347e-01 -3.95172983e-01 -6.52563035e-01 6.96289718e-01 4.73853886e-01 -3.99664432e-01 -5.88593900e-01 -6.02864623e-01 9.37348008e-02 4.87141550e-01 8.16262245e-01 -8.40003908e-01 1.82764366e-01 -2.59280443e-01 2.20983252e-01 -7.57531166e-01 4.88333344e-01 -3.81873757e-01 3.85302812e-01 5.05666211e-02 -5.60779035e-01 -8.11218619e-02 4.21333909e-02 7.69762099e-01 -3.84543926e-01 -7.50517994e-02 5.29511154e-01 -1.37163103e-01 -1.12777770e+00 5.61336339e-01 -1.76616788e-01 4.98462200e-01 1.10454965e+00 -2.09739238e-01 -3.57916802e-01 -6.34864509e-01 -6.75933719e-01 5.10314703e-01 3.33611131e-01 7.56306410e-01 8.10648739e-01 -1.38395226e+00 -9.24321830e-01 -2.76846796e-01 2.56916106e-01 2.46403620e-01 5.31120598e-01 7.72115409e-01 -4.09480810e-01 3.96723092e-01 5.80627322e-02 -6.65702939e-01 -1.22083282e+00 3.63241702e-01 2.93866187e-01 -1.92349136e-01 -8.99225473e-01 9.35846031e-01 6.78091049e-01 1.17835015e-01 3.20814967e-01 -2.86558986e-01 -4.15784586e-03 1.33017883e-01 6.21129811e-01 2.01415941e-01 -3.07948947e-01 -7.61926949e-01 -3.17365199e-01 3.70441020e-01 -2.71300673e-02 -2.77732760e-01 1.05238008e+00 -3.40006769e-01 1.70725957e-01 2.81264693e-01 1.00424254e+00 -2.58285761e-01 -1.76133883e+00 -1.85116798e-01 -1.15856960e-01 -2.71894366e-01 -4.45426106e-02 -6.67449892e-01 -1.02513349e+00 7.11510301e-01 6.56848624e-02 -3.31911832e-01 1.04986703e+00 3.92097384e-01 1.13423514e+00 6.12357371e-02 3.59393805e-01 -9.36555564e-01 4.74798083e-01 2.52317369e-01 9.94793415e-01 -1.05659318e+00 -2.08085507e-01 -4.78004694e-01 -1.04396021e+00 8.19221437e-01 7.51057744e-01 1.78379357e-01 -2.13281270e-02 9.25672427e-02 -1.98472828e-01 -2.75245216e-02 -8.47543657e-01 -2.00256616e-01 4.38741207e-01 5.11684954e-01 3.35011959e-01 -3.43874432e-02 -6.43009543e-02 7.33233452e-01 9.96922180e-02 1.62081346e-01 4.78124499e-01 6.24798179e-01 -4.29131418e-01 -7.21981704e-01 -2.79447734e-01 2.39009604e-01 -3.73054773e-01 -2.88749605e-01 -1.85460821e-01 6.47951901e-01 1.65565372e-01 8.68707120e-01 3.22032183e-01 -1.16877496e-01 1.65547520e-01 1.38811871e-01 4.29142803e-01 -6.00268066e-01 -1.44066393e-01 3.07906896e-01 3.76733750e-01 -8.31521690e-01 -6.71802461e-01 -6.91148758e-01 -1.58030379e+00 7.90940002e-02 -7.46871904e-02 1.91794515e-01 8.93405974e-02 1.04082716e+00 5.34458339e-01 5.84942400e-01 1.88104823e-01 -1.06856132e+00 -6.93508238e-02 -7.47989416e-01 -2.30233327e-01 5.73787093e-01 1.77078873e-01 -5.69743514e-01 -1.63056552e-01 6.82948709e-01]
[10.41744613647461, 0.6920532584190369]
0e842a33-4475-4d7b-97a3-7421c514359a
real-image-denoising-with-feature-attention
1904.07396
null
https://arxiv.org/abs/1904.07396v2
https://arxiv.org/pdf/1904.07396v2.pdf
Real Image Denoising with Feature Attention
Deep convolutional neural networks perform better on images containing spatially invariant noise (synthetic noise); however, their performance is limited on real-noisy photographs and requires multiple stage network modeling. To advance the practicability of denoising algorithms, this paper proposes a novel single-stage blind real image denoising network (RIDNet) by employing a modular architecture. We use a residual on the residual structure to ease the flow of low-frequency information and apply feature attention to exploit the channel dependencies. Furthermore, the evaluation in terms of quantitative metrics and visual quality on three synthetic and four real noisy datasets against 19 state-of-the-art algorithms demonstrate the superiority of our RIDNet.
['Nick Barnes', 'Saeed Anwar']
2019-04-16
real-image-denoising-with-feature-attention-1
http://openaccess.thecvf.com/content_ICCV_2019/html/Anwar_Real_Image_Denoising_With_Feature_Attention_ICCV_2019_paper.html
http://openaccess.thecvf.com/content_ICCV_2019/papers/Anwar_Real_Image_Denoising_With_Feature_Attention_ICCV_2019_paper.pdf
iccv-2019-10
['color-image-denoising']
['computer-vision']
[ 1.16305657e-01 -4.63817656e-01 4.40665394e-01 -3.51622641e-01 -6.47773683e-01 -1.67024285e-01 4.98552740e-01 -6.00121975e-01 -7.46671081e-01 6.46625876e-01 3.83032173e-01 -9.14335772e-02 -6.51478991e-02 -5.89549780e-01 -5.00499129e-01 -9.90283787e-01 -1.36468515e-01 -7.41506100e-01 2.14976504e-01 -4.24754620e-01 7.21858814e-02 4.05058086e-01 -1.29308605e+00 2.99104244e-01 9.24555659e-01 1.25193954e+00 1.66008264e-01 5.97181976e-01 1.81943372e-01 7.13460386e-01 -9.34425712e-01 -4.61899608e-01 6.97079539e-01 -4.67207164e-01 -1.74365595e-01 -1.55256027e-02 4.25910145e-01 -6.17115617e-01 -1.09062123e+00 1.53116453e+00 1.03904688e+00 1.25034243e-01 2.98422158e-01 -8.13518226e-01 -1.02616930e+00 3.45303297e-01 -4.08695191e-01 4.26366061e-01 -4.52994257e-02 5.59740782e-01 3.68013561e-01 -7.53779769e-01 4.64282066e-01 1.26450419e+00 1.05033135e+00 5.55778623e-01 -1.26531017e+00 -7.85973907e-01 1.17789306e-01 3.64324272e-01 -1.33494675e+00 -6.18734121e-01 8.41411114e-01 -6.05147053e-03 6.42308772e-01 2.07837652e-02 3.27798814e-01 1.42089140e+00 3.19682658e-01 3.24034244e-01 1.30679977e+00 -2.36017648e-02 1.81199446e-01 -4.48087156e-01 -6.68916479e-02 3.50958079e-01 3.33515644e-01 4.75386828e-01 -5.49846828e-01 1.94279790e-01 9.06388998e-01 -1.62238926e-02 -7.21252620e-01 1.11040525e-01 -9.16626990e-01 5.49508572e-01 9.57507908e-01 3.96871537e-01 -3.65652680e-01 4.31479812e-01 4.60434586e-01 6.26712263e-01 5.76120436e-01 2.24885702e-01 -3.14614713e-01 1.14604473e-01 -1.04089594e+00 -1.85147673e-01 4.72302020e-01 6.62318707e-01 6.16344035e-01 5.41855216e-01 -3.19089979e-01 9.29104865e-01 2.72670269e-01 4.38889921e-01 4.08598781e-01 -1.13649094e+00 2.42389053e-01 2.08412617e-01 -4.35722694e-02 -9.97784674e-01 -4.43609953e-01 -7.56839871e-01 -1.68875360e+00 5.51624775e-01 1.16423517e-01 -2.18841895e-01 -1.38904870e+00 1.49091792e+00 -2.69644946e-01 1.60468787e-01 1.18023060e-01 1.51182795e+00 1.18892550e+00 6.18686914e-01 6.71630055e-02 -2.56780148e-01 8.73396158e-01 -8.64111960e-01 -9.25652087e-01 -2.77105093e-01 2.81292759e-02 -7.97578812e-01 7.28030086e-01 7.64575541e-01 -9.02254105e-01 -9.44974601e-01 -1.17023289e+00 -1.50312290e-01 -2.18766794e-01 4.56257164e-01 5.36269367e-01 8.88820648e-01 -1.24890184e+00 8.10350895e-01 -6.14635706e-01 -1.28632098e-01 6.79549277e-01 3.42451900e-01 -5.48131824e-01 -4.50086117e-01 -1.31418610e+00 5.97296119e-01 1.51120812e-01 8.89245093e-01 -1.20158601e+00 -4.02844578e-01 -6.95415318e-01 -3.71969528e-02 -3.84547259e-03 -6.84699059e-01 7.44808614e-01 -1.24458778e+00 -1.60566461e+00 5.11997163e-01 1.09153681e-01 -4.59857225e-01 6.25525236e-01 -2.95899570e-01 -7.56736577e-01 3.51062179e-01 -7.25259855e-02 4.96568590e-01 1.23877382e+00 -1.38657844e+00 -1.62015855e-01 -6.40797988e-02 8.08357298e-02 -2.17561334e-01 -3.38734448e-01 2.24914141e-02 -5.79610407e-01 -1.23587263e+00 3.66208225e-01 -4.70782965e-01 -3.74483019e-01 3.63527954e-01 -3.55160236e-01 3.81938815e-01 8.50002229e-01 -1.01015365e+00 1.09142816e+00 -2.60250854e+00 -4.22125868e-02 1.51018441e-01 3.01995397e-01 6.73425615e-01 -5.01297951e-01 2.03922674e-01 -2.45010510e-01 1.60847172e-01 -3.18237245e-01 -2.19292104e-01 -2.87015438e-01 2.93581873e-01 -9.08817723e-02 7.98202217e-01 1.67367488e-01 6.40856802e-01 -6.46414101e-01 -1.09429397e-01 2.15440601e-01 7.18156576e-01 -2.92709291e-01 2.69432694e-01 3.14414173e-01 5.31212509e-01 -1.48065433e-01 6.05541110e-01 1.21209812e+00 1.15816198e-01 -2.05064282e-01 -8.56032014e-01 2.46332809e-02 -7.69468695e-02 -1.23475885e+00 1.85698974e+00 -3.90508950e-01 9.00748670e-01 6.22182667e-01 -7.47447848e-01 8.53095293e-01 2.23944873e-01 8.74833167e-02 -1.11322463e+00 2.69639790e-01 1.33698463e-01 1.94281025e-03 -7.37886131e-01 2.60129333e-01 -8.12541991e-02 3.86763781e-01 -6.06623702e-02 3.53990108e-01 2.78860688e-01 9.92272943e-02 1.46377847e-01 1.24420142e+00 -5.20914793e-02 -1.66917413e-01 -2.01577172e-01 5.39544761e-01 -5.41010559e-01 6.77881062e-01 9.75293100e-01 -3.85157317e-01 1.19085944e+00 4.09224123e-01 -4.58888531e-01 -7.51086831e-01 -1.04285526e+00 -1.43984094e-01 5.97219765e-01 4.40672994e-01 -2.14222759e-01 -7.33654916e-01 -6.48477256e-01 -2.48055071e-01 1.72743335e-01 -6.33105218e-01 -3.24494690e-01 -4.35923934e-01 -9.95721757e-01 7.49119163e-01 4.88314658e-01 1.23046732e+00 -8.02964747e-01 -1.32926658e-01 5.37009314e-02 -2.46647298e-01 -1.13579130e+00 -3.53268474e-01 2.19643116e-01 -6.21393979e-01 -1.07931614e+00 -8.71306896e-01 -7.48378754e-01 6.86324596e-01 5.05309403e-01 9.87357914e-01 3.49064589e-01 -4.09619719e-01 9.56872702e-02 -3.94162208e-01 3.49218398e-03 -1.24648497e-01 -3.07891130e-01 -7.58367330e-02 9.62856784e-02 -1.13959283e-01 -7.14029014e-01 -1.11346304e+00 4.07463521e-01 -1.27491069e+00 -3.59345198e-01 9.30192590e-01 1.21155989e+00 3.40262294e-01 4.08261061e-01 4.27473396e-01 -3.91395837e-01 8.00302923e-01 -1.09031767e-01 -6.48112833e-01 -7.06730224e-03 -3.91863972e-01 -6.46810308e-02 7.01611698e-01 -4.49220002e-01 -1.28616714e+00 -1.83702782e-01 -3.55929404e-01 -3.97056460e-01 -5.29503562e-02 3.41086715e-01 -3.18152100e-01 -5.48841119e-01 8.26588809e-01 2.14284435e-01 5.69397286e-02 -6.76531971e-01 2.30417073e-01 6.59525692e-01 8.86604846e-01 -1.45364091e-01 1.02470744e+00 5.48743069e-01 2.95779333e-02 -8.59572589e-01 -6.30842447e-01 -2.33956739e-01 -3.16565841e-01 -2.38901660e-01 7.98111975e-01 -1.33115387e+00 -7.25492477e-01 1.04100382e+00 -1.21388507e+00 -2.91828066e-01 -6.74934387e-02 3.95980567e-01 1.96201112e-02 6.44213915e-01 -8.98672163e-01 -5.30979693e-01 -4.24429893e-01 -1.20952106e+00 8.32864821e-01 2.79933840e-01 4.82834131e-01 -6.61654472e-01 -3.46871704e-01 1.11931749e-01 9.16051686e-01 -3.62040810e-02 5.46115577e-01 -5.44008277e-02 -7.08918393e-01 -2.63584137e-01 -6.83864832e-01 1.10534441e+00 -5.05252145e-02 -2.07194835e-01 -1.33538437e+00 -4.37879473e-01 2.47152761e-01 -1.49034321e-01 1.45268548e+00 5.32603145e-01 1.47032809e+00 -1.69813588e-01 1.53062806e-01 1.15184343e+00 1.71686745e+00 -1.36176676e-01 1.35013139e+00 3.52448314e-01 5.98011434e-01 2.46784687e-01 1.90917794e-02 1.19744249e-01 -1.31952703e-01 2.09428206e-01 7.86287665e-01 -5.67755342e-01 -6.91551268e-01 3.99407335e-02 3.09571892e-01 5.88656604e-01 -1.81597725e-01 -4.26839024e-01 -4.18186396e-01 5.67439914e-01 -1.43563545e+00 -7.50947535e-01 -2.57906020e-01 1.71399665e+00 5.85859716e-01 1.44378021e-01 -4.82731670e-01 1.63585380e-01 5.96750140e-01 3.45731109e-01 -3.98959905e-01 7.05508441e-02 -7.35582948e-01 3.06938469e-01 9.71159279e-01 1.45090252e-01 -1.36766446e+00 5.55610061e-01 7.07668304e+00 1.05369413e+00 -1.12367332e+00 1.51065707e-01 8.08558464e-01 1.27239928e-01 9.44188461e-02 -2.36366734e-01 -1.86760202e-01 5.09303331e-01 6.70781016e-01 3.81682009e-01 5.62197149e-01 3.77021849e-01 4.32466596e-01 -2.40507483e-01 -4.92279500e-01 1.24394548e+00 -4.90963981e-02 -1.23129404e+00 3.43033336e-02 -1.97081834e-01 8.29249740e-01 1.82559684e-01 2.01901495e-01 3.82326096e-02 1.43844888e-01 -1.19578910e+00 4.57865328e-01 8.06277394e-01 8.51528585e-01 -5.97265422e-01 1.19486284e+00 -2.82464214e-02 -8.06233168e-01 -1.74084172e-01 -5.97031116e-01 -2.23691072e-02 2.32313927e-02 8.55888188e-01 1.37978002e-01 6.68366849e-01 1.22334397e+00 7.59123743e-01 -8.45635176e-01 1.29321086e+00 -5.79164207e-01 7.87798524e-01 -9.41880494e-02 5.26505768e-01 2.34275699e-01 -2.37383023e-01 4.70848262e-01 1.30391383e+00 2.37636819e-01 9.97362658e-02 -2.97223181e-01 6.82977200e-01 -3.80893230e-01 -4.58997309e-01 -3.45624924e-01 2.91366011e-01 -9.35383439e-02 1.37181807e+00 -4.85177606e-01 2.53091613e-03 -5.25132716e-01 1.24697781e+00 -3.43071312e-01 8.95232379e-01 -5.78384638e-01 -6.25539541e-01 8.77379477e-01 -1.86768115e-01 4.48419392e-01 -3.24587971e-01 -4.81254071e-01 -1.15723467e+00 1.50283679e-01 -1.00441515e+00 -4.38897274e-02 -1.06847370e+00 -1.58979678e+00 8.13169658e-01 -5.01443803e-01 -1.33637071e+00 3.92846435e-01 -7.76015162e-01 -6.77885771e-01 1.08437097e+00 -1.74218416e+00 -1.02234960e+00 -7.04758584e-01 7.87486017e-01 2.39455730e-01 -2.98149794e-01 6.67011738e-01 7.02424109e-01 -6.51241422e-01 6.35005116e-01 3.74522597e-01 4.33047801e-01 8.92183304e-01 -8.76594722e-01 4.61898655e-01 1.44508266e+00 -2.72873014e-01 5.07738113e-01 6.78635538e-01 -4.37141091e-01 -1.40061331e+00 -1.13218141e+00 9.70390290e-02 3.38004708e-01 5.00169337e-01 -3.07663351e-01 -1.06475329e+00 1.43037796e-01 4.89855736e-01 4.58013862e-01 1.50416955e-01 -3.53163719e-01 -4.95200932e-01 -5.68373561e-01 -1.36118507e+00 5.10571599e-01 1.14927101e+00 -5.26671588e-01 -5.59537299e-02 9.48045924e-02 6.16847157e-01 -2.25613862e-01 -7.29396760e-01 4.01988894e-01 2.00330362e-01 -1.25124919e+00 1.04130411e+00 -9.62001830e-02 3.49600852e-01 -5.26201487e-01 -1.93228737e-01 -1.53248703e+00 -3.75454098e-01 -9.61586535e-01 1.34559706e-01 1.17335284e+00 1.48692094e-02 -5.75085878e-01 5.31436443e-01 1.06446519e-01 -2.38682672e-01 -3.10566187e-01 -1.02703655e+00 -6.70978367e-01 -2.77577639e-01 -4.47661787e-01 2.57822335e-01 5.35391092e-01 -8.20191026e-01 -5.36989421e-02 -6.69274330e-01 3.09055746e-01 8.63279760e-01 -5.64023733e-01 4.41666037e-01 -9.64418769e-01 2.64543220e-02 -2.70855218e-01 -8.42630684e-01 -1.13012767e+00 -1.50439618e-02 -3.31905872e-01 1.97912931e-01 -1.57104361e+00 -1.59203306e-01 3.64655219e-02 -4.12657499e-01 2.94292033e-01 -1.43796161e-01 8.53510320e-01 -5.33701852e-02 8.10866058e-02 -5.79097688e-01 8.76334190e-01 1.21889389e+00 -4.02608067e-01 2.75766104e-01 -2.31892034e-01 -8.14105093e-01 6.96315527e-01 6.08482420e-01 -3.71545076e-01 -1.37243479e-01 -8.79972875e-01 7.05862418e-02 -2.58731157e-01 7.95735121e-01 -1.29550874e+00 4.80252743e-01 3.25153559e-01 7.83350348e-01 -2.44877964e-01 3.00817519e-01 -9.36448097e-01 -3.08480728e-02 4.77229208e-01 -1.22424640e-01 -1.82235539e-01 3.31445009e-01 8.15644622e-01 -6.90571964e-01 2.68692914e-02 1.05398667e+00 -1.18258879e-01 -1.00646424e+00 1.67702764e-01 -3.86528194e-01 -8.17210078e-02 5.35793006e-01 -2.66027957e-01 -5.80953896e-01 -5.23760676e-01 -6.20136499e-01 -2.52140537e-02 2.34600201e-01 2.66572177e-01 9.28812683e-01 -1.18501687e+00 -7.53923655e-01 4.73640770e-01 -1.23276673e-01 -2.66313016e-01 6.29589260e-01 7.86639273e-01 -7.99823463e-01 -1.37226075e-01 -2.93229610e-01 -4.08808291e-01 -9.95751023e-01 3.43841583e-01 8.16936374e-01 9.05702412e-02 -7.25303590e-01 1.01241994e+00 1.20393798e-01 -9.29472223e-02 5.73444188e-01 -1.38923243e-01 -8.03972706e-02 -2.78700322e-01 6.17742479e-01 4.53342736e-01 3.22486609e-01 -4.64737922e-01 -1.66130602e-01 3.96298528e-01 1.94939882e-01 9.20339972e-02 1.56317937e+00 -3.84329170e-01 -3.31939638e-01 -1.43485397e-01 1.36013591e+00 -2.98837453e-01 -1.58397937e+00 -2.53520131e-01 -4.17276293e-01 -6.68751240e-01 6.30234420e-01 -1.01743007e+00 -1.76244891e+00 7.33474791e-01 1.32727551e+00 9.83594358e-02 1.86983502e+00 -6.42715156e-01 7.83770740e-01 3.67066592e-01 -3.80414771e-03 -9.70095575e-01 7.92565271e-02 3.60631824e-01 1.04618037e+00 -1.38547075e+00 5.31613678e-02 -4.65857267e-01 -2.57269412e-01 1.22539222e+00 4.89164174e-01 -3.31315368e-01 9.53653038e-01 3.48015964e-01 5.34803331e-01 -9.44456607e-02 -2.73362458e-01 -2.58231878e-01 2.60364443e-01 8.75257015e-01 2.63910621e-01 -4.60071802e-01 -3.31626713e-01 6.07263267e-01 2.15111315e-01 2.22667828e-02 4.69570935e-01 5.94717681e-01 -1.93905950e-01 -7.40542591e-01 -5.21283984e-01 3.06476831e-01 -8.77202034e-01 -2.72596806e-01 -2.19675571e-01 6.35164618e-01 4.16101187e-01 1.27756739e+00 -2.94870615e-01 -4.70252186e-01 5.27375758e-01 -5.86403430e-01 1.91171139e-01 1.54099641e-02 -8.73842776e-01 2.68108487e-01 -5.63537963e-02 -7.49903679e-01 -6.38443410e-01 5.21983132e-02 -6.04502678e-01 -3.67749035e-01 -2.29627594e-01 -2.45781079e-01 7.83993840e-01 7.18365133e-01 3.71274471e-01 8.98414731e-01 6.06674016e-01 -1.14988840e+00 -5.33706725e-01 -1.16773081e+00 -7.73304582e-01 4.06703115e-01 7.33598471e-01 -3.57865185e-01 -6.63052917e-01 -1.69089306e-02]
[11.44696044921875, -2.3405251502990723]
fcdb9f8e-8ec1-48ae-8692-7a18274425e0
conditioning-autoencoder-latent-spaces-for
2001.11296
null
https://arxiv.org/abs/2001.11296v1
https://arxiv.org/pdf/2001.11296v1.pdf
Conditioning Autoencoder Latent Spaces for Real-Time Timbre Interpolation and Synthesis
We compare standard autoencoder topologies' performances for timbre generation. We demonstrate how different activation functions used in the autoencoder's bottleneck distributes a training corpus's embedding. We show that the choice of sigmoid activation in the bottleneck produces a more bounded and uniformly distributed embedding than a leaky rectified linear unit activation. We propose a one-hot encoded chroma feature vector for use in both input augmentation and latent space conditioning. We measure the performance of these networks, and characterize the latent embeddings that arise from the use of this chroma conditioning vector. An open source, real-time timbre synthesis algorithm in Python is outlined and shared.
['Joseph T Colonel', 'Sam Keene']
2020-01-30
null
null
null
null
['timbre-interpolation']
['audio']
[ 1.42280513e-03 1.61677480e-01 -3.05824075e-02 -3.11810613e-01 -5.88160634e-01 -5.89167118e-01 6.34018958e-01 -1.48621142e-01 -5.40711939e-01 6.84012890e-01 8.84639561e-01 -1.94212839e-01 2.12562054e-01 -8.17781389e-01 -6.53071046e-01 -9.72715259e-01 7.51670897e-02 6.15632348e-02 -4.31273460e-01 -5.14568985e-02 -2.08187163e-01 3.58393550e-01 -1.33869231e+00 5.57387650e-01 1.81309357e-01 9.20315921e-01 -1.40102401e-01 1.02075779e+00 1.81294397e-01 7.98012555e-01 -1.07488966e+00 -2.42473148e-02 2.28511579e-02 -5.25244832e-01 -8.76439095e-01 -4.71719384e-01 -6.78326283e-03 -4.26912338e-01 -5.35238743e-01 4.90175724e-01 8.52243125e-01 6.11987948e-01 8.26667786e-01 -1.05993021e+00 -7.48448133e-01 8.11163545e-01 2.53787696e-01 2.70197719e-01 1.25567645e-01 3.10489058e-01 9.70754802e-01 -5.43215990e-01 4.72491622e-01 9.23145592e-01 9.55683231e-01 8.10934842e-01 -1.83531165e+00 -5.31804144e-01 -4.09506887e-01 -5.54903597e-02 -1.38566053e+00 -5.82379043e-01 9.83893991e-01 -4.16440338e-01 1.42612946e+00 3.33317250e-01 7.57317424e-01 1.54145885e+00 -6.35135695e-02 6.60860479e-01 9.50799346e-01 -5.76769888e-01 4.24557418e-01 3.71399760e-01 -4.45913263e-02 1.74141660e-01 -5.52703083e-01 3.26258779e-01 -6.59961164e-01 -4.39844310e-01 7.85974860e-01 -3.90915692e-01 -3.92241120e-01 1.48805857e-01 -9.62264419e-01 1.09380686e+00 4.90874559e-01 4.22328413e-01 -5.95570147e-01 7.92765379e-01 4.34984744e-01 3.56073231e-01 5.58125973e-01 8.56243312e-01 -4.50181484e-01 -5.28899372e-01 -8.70986462e-01 4.25018370e-02 7.85101235e-01 5.07676363e-01 5.75296104e-01 6.18318439e-01 -1.67816758e-01 9.09619987e-01 -1.48393095e-01 1.89325735e-01 8.99322867e-01 -1.06211388e+00 -2.93171760e-02 -8.53620917e-02 -1.64094925e-01 -4.31723416e-01 -1.81274638e-01 -3.12497348e-01 -7.37834632e-01 1.09393649e-01 1.85439825e-01 -6.28232718e-01 -6.94972932e-01 1.76323366e+00 -1.32568732e-01 3.51752549e-01 3.06398720e-01 7.62357473e-01 4.91625786e-01 1.09138083e+00 7.97590166e-02 -3.85176241e-02 1.07988477e+00 -7.97523081e-01 -9.74878132e-01 2.79215705e-02 5.62778354e-01 -7.28553176e-01 1.11271667e+00 1.37038276e-01 -1.17173886e+00 -8.57804656e-01 -1.09061623e+00 -3.12552750e-01 -6.25858247e-01 1.26450047e-01 7.09378719e-01 7.48053372e-01 -1.10567284e+00 1.02023399e+00 -6.74662471e-01 1.88600302e-01 2.26651698e-01 3.17852676e-01 -3.22214097e-01 5.61883271e-01 -1.42189610e+00 9.45053339e-01 6.00757718e-01 -1.34598151e-01 -8.57688129e-01 -9.05678689e-01 -7.88284898e-01 2.12253943e-01 -6.32371485e-01 -7.44122088e-01 1.29797328e+00 -1.08438563e+00 -2.02414775e+00 5.52610993e-01 5.06601743e-02 -8.11573505e-01 -8.42292830e-02 -2.99500585e-01 -4.45302725e-01 1.37334451e-01 -5.93256712e-01 1.08656669e+00 1.13475013e+00 -8.92407000e-01 -1.09448209e-01 2.56740481e-01 -2.51227379e-01 1.79617614e-01 -7.10194588e-01 -1.41458601e-01 3.84956151e-01 -7.91284680e-01 -5.85987806e-01 -7.26813138e-01 6.11547157e-02 -4.08079922e-01 -2.27273047e-01 -2.58743048e-01 6.99276030e-01 -5.95745921e-01 1.28535974e+00 -2.44803309e+00 1.45954013e-01 8.59786198e-02 -3.79489027e-02 2.13215668e-02 -5.15467674e-02 5.38290918e-01 -3.56313914e-01 1.10773578e-01 -1.36509195e-01 -5.27880788e-01 3.19481075e-01 1.78591937e-01 -9.12805676e-01 3.85710478e-01 3.10831130e-01 7.76698589e-01 -6.37328804e-01 7.94152394e-02 2.73929626e-01 1.10894907e+00 -9.04759228e-01 4.44343001e-01 -1.05008051e-01 1.85429417e-02 2.38077670e-01 -1.44321755e-01 3.12386721e-01 1.54524967e-01 -6.52790442e-02 -3.40251356e-01 -1.51389450e-01 9.17698562e-01 -7.05930889e-01 1.79608357e+00 -7.19347417e-01 1.24473619e+00 -2.00471789e-01 -5.06548464e-01 1.00334871e+00 8.99062753e-01 8.12477246e-02 -1.46363422e-01 3.90320361e-01 4.18178812e-02 -7.57978782e-02 -1.81040108e-01 5.43904006e-01 -6.84004545e-01 1.12447925e-01 6.49042785e-01 6.03398383e-01 -3.21500123e-01 -4.23463762e-01 -9.79067907e-02 9.24305677e-01 1.52349383e-01 -1.26132756e-01 -1.10401370e-01 -7.31849372e-02 -1.98152766e-01 1.35240123e-01 6.18809938e-01 7.21051842e-02 7.73116112e-01 3.86462182e-01 -4.59352165e-01 -1.52863145e+00 -9.35934603e-01 -2.89107442e-01 1.41858768e+00 -6.87985539e-01 -3.73424143e-01 -8.53218973e-01 -9.29307789e-02 -6.26987517e-02 1.19063759e+00 -8.12129140e-01 -6.40673757e-01 -6.22353911e-01 -6.37748122e-01 1.22747481e+00 7.87020981e-01 1.14204183e-01 -1.51160717e+00 -7.51210272e-01 3.17601830e-01 -1.87428951e-01 -6.06504560e-01 -4.09765452e-01 1.02052534e+00 -9.29904222e-01 -2.02865854e-01 -7.05464065e-01 -6.35206640e-01 2.46351644e-01 -4.99505401e-01 9.49191988e-01 -5.58629036e-01 -2.42910668e-01 3.01225662e-01 -1.41980723e-01 -3.23720247e-01 -5.51184177e-01 3.51866692e-01 1.44782737e-01 -1.50559917e-01 6.62853897e-01 -9.89519238e-01 -6.88107133e-01 -3.65663499e-01 -9.77033496e-01 -4.63256892e-03 8.99967253e-02 9.17560816e-01 3.45084250e-01 6.14464246e-02 4.95252341e-01 -4.94819880e-01 1.26585746e+00 -5.17500758e-01 -5.73945828e-02 -4.17200357e-01 -2.70785302e-01 4.17616397e-01 7.62700200e-01 -6.93088412e-01 -9.38377857e-01 -8.15123767e-02 -5.96581221e-01 -6.20700777e-01 -3.33415598e-01 1.93719998e-01 3.37348938e-01 4.63422179e-01 1.09223485e+00 1.73323974e-01 -1.18262216e-01 -5.86774945e-01 8.92239630e-01 8.68617356e-01 7.17829883e-01 -5.32341063e-01 5.44545531e-01 3.72307777e-01 -6.42016351e-01 -8.80021393e-01 -4.16206181e-01 -2.76826054e-01 -4.61182833e-01 1.97166920e-01 1.02880335e+00 -8.32178950e-01 -7.23147631e-01 2.49274179e-01 -1.34389853e+00 -7.02089727e-01 -1.10699773e+00 6.47472322e-01 -9.24390793e-01 -3.60530406e-01 -9.25113976e-01 -8.56415451e-01 -5.67856491e-01 -6.84504211e-01 6.85063064e-01 1.72839671e-01 -6.75339043e-01 -1.14917994e+00 6.68251038e-01 -2.45786518e-01 6.63994610e-01 1.65118854e-02 8.83058131e-01 -6.90257609e-01 6.96421936e-02 -1.91842213e-01 3.47310483e-01 9.37896490e-01 -1.73943430e-01 -7.78958350e-02 -1.92071259e+00 -2.22378775e-01 2.14949161e-01 -6.72843337e-01 9.84358847e-01 4.42222089e-01 1.09734082e+00 -6.00321054e-01 1.24349512e-01 1.05029047e+00 1.17374706e+00 3.73454392e-02 7.77804494e-01 2.82426357e-01 3.52465510e-01 3.49118531e-01 -3.96802634e-01 3.46065134e-01 -2.42047444e-01 3.14754665e-01 -3.58039029e-02 -1.35769367e-01 -1.04842573e-01 -4.91742373e-01 5.23191214e-01 1.36132634e+00 -8.86819735e-02 -7.03379214e-02 -5.33350945e-01 6.27523482e-01 -1.26127279e+00 -1.13427103e+00 5.55653453e-01 1.85241628e+00 1.43967223e+00 9.00393799e-02 1.40344888e-01 4.79548782e-01 3.27645153e-01 1.99043915e-01 -2.10600585e-01 -1.12680006e+00 -2.86697030e-01 8.91047657e-01 3.42754364e-01 7.14942992e-01 -1.04745936e+00 7.86792815e-01 7.82526112e+00 8.07553649e-01 -1.20159447e+00 1.95599899e-01 3.99859786e-01 -3.82533729e-01 -4.76705968e-01 -3.33716154e-01 -4.85914081e-01 3.64074141e-01 1.62500703e+00 4.04928401e-02 8.16016376e-01 9.60767031e-01 4.66506518e-02 3.95878315e-01 -1.20291615e+00 8.25340927e-01 -2.37881750e-01 -1.44560874e+00 -9.89922509e-02 -4.14142609e-02 6.76630139e-01 7.77749121e-02 4.62039202e-01 4.03929591e-01 4.16480333e-01 -1.31221724e+00 6.08721137e-01 4.40103978e-01 1.04321015e+00 -1.04259682e+00 4.88355666e-01 1.35906404e-02 -7.99988806e-01 -5.64487725e-02 -4.88901973e-01 -2.53801674e-01 -6.35962039e-02 2.07174018e-01 -1.11613190e+00 -2.79130965e-01 3.46329242e-01 1.35662943e-01 -6.16813637e-02 5.81451476e-01 -4.15188879e-01 1.26002884e+00 -4.70155001e-01 -7.80555382e-02 2.67017424e-01 1.09992765e-01 3.59110296e-01 1.83599901e+00 -4.05119285e-02 -4.12014201e-02 -4.94174331e-01 1.20893407e+00 -3.25087696e-01 -6.03488609e-02 -5.88517487e-01 -2.91081607e-01 6.44900739e-01 9.82442856e-01 -9.89396945e-02 -4.06577200e-01 1.59796879e-01 1.33543336e+00 2.51928717e-01 7.85809040e-01 -8.17767680e-01 -8.90506089e-01 9.47889805e-01 -9.47550163e-02 3.38445723e-01 -1.24142885e-01 -2.36103937e-01 -8.84323835e-01 -3.89737129e-01 -6.58652782e-01 1.65102094e-01 -9.91750121e-01 -1.13777852e+00 7.53204346e-01 -8.81765783e-02 -6.12916052e-01 -6.81623876e-01 -5.62496901e-01 -7.49241054e-01 1.30286300e+00 -1.09155309e+00 -7.64564753e-01 1.26611844e-01 6.03014171e-01 2.81926334e-01 -2.06330597e-01 1.64917433e+00 1.26774922e-01 -3.52652043e-01 6.51708603e-01 3.33049417e-01 3.17890495e-01 3.00260186e-01 -1.51655352e+00 5.49395025e-01 1.79460466e-01 3.69094253e-01 8.38348150e-01 7.93330491e-01 -7.15263635e-02 -1.00360560e+00 -7.99676895e-01 7.65746295e-01 -2.93317884e-01 6.80135131e-01 -5.50259590e-01 -9.81682599e-01 9.38111305e-01 6.96280420e-01 -1.33823678e-01 1.02867520e+00 2.25442290e-01 -5.83404303e-01 6.29407912e-02 -8.89831722e-01 4.73142147e-01 3.58146161e-01 -1.28191888e+00 -9.46040452e-01 2.01356132e-02 1.01168799e+00 -6.33415133e-02 -1.00230312e+00 -4.81090397e-02 6.11613572e-01 -6.18641317e-01 1.06448877e+00 -8.67661655e-01 4.91423190e-01 -3.13728303e-02 -1.55605420e-01 -1.64574993e+00 -4.95364070e-01 -1.00944543e+00 -2.37627223e-01 9.66301739e-01 3.31732839e-01 -4.77343708e-01 5.22643387e-01 2.62571067e-01 -5.39152436e-02 -5.88162124e-01 -9.61192012e-01 -3.38870257e-01 4.76392359e-01 -5.18977165e-01 4.04754132e-01 9.31111097e-01 2.84027010e-01 5.93525052e-01 -4.11352932e-01 -2.40354910e-01 4.27386425e-02 -3.18316758e-01 4.68019873e-01 -6.56064332e-01 -5.06355941e-01 -5.20198882e-01 -3.65112364e-01 -1.09893453e+00 -2.35318299e-02 -9.84603524e-01 4.49328683e-02 -1.07214153e+00 -3.06410670e-01 -3.02200943e-01 -6.27795935e-01 5.54120898e-01 1.69615269e-01 4.90865797e-01 5.72180040e-02 -8.00003335e-02 1.52899057e-01 7.72811234e-01 5.65388441e-01 4.01066132e-02 -4.38783228e-01 -3.11027735e-01 -2.79590130e-01 6.25352442e-01 1.15139973e+00 -6.06677175e-01 -3.93795788e-01 -4.45852607e-01 2.50531673e-01 -2.67564505e-01 4.52623039e-01 -1.02337646e+00 7.42712840e-02 2.22100094e-01 6.16689980e-01 -2.53802955e-01 7.91069984e-01 -3.98178637e-01 1.12236947e-01 1.62912309e-01 -7.99959004e-01 -2.55300045e-01 5.55465102e-01 5.40128238e-02 -1.37408420e-01 -3.49014640e-01 7.87996233e-01 5.79071455e-02 5.94211034e-02 -2.19777152e-01 -6.94228649e-01 5.15351724e-03 3.85710716e-01 -6.38160035e-02 -2.41119951e-01 -4.90498960e-01 -9.31849360e-01 -5.37655950e-01 1.27292484e-01 2.12413788e-01 5.03086030e-01 -1.66272044e+00 -6.99421108e-01 4.14897978e-01 -3.64976615e-01 -4.39563125e-01 -3.98686193e-02 4.59586591e-01 -3.94774288e-01 3.64759177e-01 -3.73840153e-01 -2.47106269e-01 -7.69430816e-01 4.16549981e-01 6.31640732e-01 1.49755493e-01 -7.29021072e-01 1.08907509e+00 -1.04164243e-01 -2.74634331e-01 4.09296870e-01 -3.60486001e-01 3.95112894e-02 1.73347756e-01 5.77950418e-01 2.93816239e-01 -1.68820128e-01 -2.26473495e-01 3.09147015e-02 -2.80075278e-02 2.83025622e-01 -6.92284882e-01 1.47773480e+00 2.86670268e-01 1.88394099e-01 1.14816737e+00 1.54230714e+00 8.00416321e-02 -1.26552618e+00 6.04692996e-02 -4.73814994e-01 7.62680406e-03 3.11793983e-01 -6.85060561e-01 -6.26940906e-01 1.33024263e+00 6.93796456e-01 4.20791388e-01 9.53627348e-01 -2.95412332e-01 8.02601576e-01 3.30912441e-01 -3.91313702e-01 -1.36101258e+00 6.77221194e-02 3.60290080e-01 8.49543273e-01 -4.41450477e-01 -2.10921913e-01 4.65717822e-01 -5.66350698e-01 1.31218755e+00 9.29444879e-02 -4.81595337e-01 6.23135090e-01 7.03726590e-01 2.84745455e-01 3.93770933e-02 -9.85707760e-01 5.93653917e-02 2.36293942e-01 6.69698119e-01 7.68110931e-01 1.87152699e-01 2.30062738e-01 6.21213913e-01 -8.37062776e-01 -1.97857618e-01 2.97399402e-01 6.82434678e-01 -3.87703061e-01 -7.51064837e-01 -3.71537626e-01 1.33631915e-01 -4.67011750e-01 -4.08519149e-01 -2.41394043e-01 5.34989774e-01 2.81999946e-01 6.58694863e-01 5.11588514e-01 -4.72525299e-01 -9.39933234e-04 8.72956753e-01 4.72547680e-01 -4.22236860e-01 -1.04393137e+00 2.91233480e-01 4.16538939e-02 -1.53419957e-01 -6.36989698e-02 -1.95194349e-01 -1.22751248e+00 -3.97016823e-01 -3.90493810e-01 2.69279957e-01 9.45047796e-01 5.19277990e-01 1.73380956e-01 8.89145792e-01 5.94603360e-01 -7.51525283e-01 -6.51642621e-01 -1.41627574e+00 -3.22865367e-01 4.97990042e-01 5.44933379e-01 -7.89219886e-02 -6.49457455e-01 4.79934245e-01]
[15.45969009399414, 5.720325946807861]
d5434072-6034-45c8-8c7a-8a9c436796b0
answering-questions-about-data-visualizations
1908.01801
null
https://arxiv.org/abs/1908.01801v2
https://arxiv.org/pdf/1908.01801v2.pdf
Answering Questions about Data Visualizations using Efficient Bimodal Fusion
Chart question answering (CQA) is a newly proposed visual question answering (VQA) task where an algorithm must answer questions about data visualizations, e.g. bar charts, pie charts, and line graphs. CQA requires capabilities that natural-image VQA algorithms lack: fine-grained measurements, optical character recognition, and handling out-of-vocabulary words in both questions and answers. Without modifications, state-of-the-art VQA algorithms perform poorly on this task. Here, we propose a novel CQA algorithm called parallel recurrent fusion of image and language (PReFIL). PReFIL first learns bimodal embeddings by fusing question and image features and then intelligently aggregates these learned embeddings to answer the given question. Despite its simplicity, PReFIL greatly surpasses state-of-the art systems and human baselines on both the FigureQA and DVQA datasets. Additionally, we demonstrate that PReFIL can be used to reconstruct tables by asking a series of questions about a chart.
['Brian Price', 'Kushal Kafle', 'Scott Cohen', 'Robik Shrestha', 'Christopher Kanan']
2019-08-05
null
null
null
null
['chart-question-answering', 'chart-question-answering']
['computer-code', 'computer-vision']
[-6.79462329e-02 5.40975332e-02 2.17770725e-01 -3.33698750e-01 -1.21240103e+00 -1.12062955e+00 7.37364531e-01 5.30882478e-01 4.60169539e-02 1.53846189e-01 4.15904373e-01 -9.16880786e-01 1.44268245e-01 -7.68947423e-01 -8.51590931e-01 -2.25069210e-01 2.13030517e-01 6.15132570e-01 2.15636060e-01 -2.61321902e-01 3.91205877e-01 6.32918417e-01 -1.50557983e+00 6.75093412e-01 7.56508708e-01 1.11225212e+00 -8.24932680e-02 1.15267158e+00 -8.58099580e-01 1.33141303e+00 -7.32286632e-01 -7.93106019e-01 -1.13685623e-01 -4.47783083e-01 -9.30514753e-01 1.80809163e-02 1.14118135e+00 -7.09103465e-01 -4.47566062e-01 7.63788164e-01 2.28562504e-01 -8.88740346e-02 8.33835006e-01 -1.60091579e+00 -1.60156560e+00 -1.43262684e-01 -5.69710791e-01 3.28108937e-01 9.74826634e-01 4.44345921e-01 1.44336474e+00 -1.32062864e+00 7.16438651e-01 1.61315262e+00 1.40222713e-01 4.73605245e-01 -1.49526930e+00 -2.86868274e-01 2.83118725e-01 2.57974356e-01 -8.86202157e-01 -1.94053248e-01 8.59091580e-01 -5.67832410e-01 9.78887498e-01 5.24449527e-01 6.31517231e-01 8.79497707e-01 1.49805576e-01 1.17528009e+00 9.50179100e-01 -2.58579105e-01 4.12585586e-01 1.16429456e-01 2.56788760e-01 1.19323599e+00 9.61826444e-02 -3.02934557e-01 -8.16417873e-01 -9.07669067e-02 8.99693489e-01 -1.35605633e-01 -1.49740845e-01 -8.33950043e-01 -1.49479711e+00 8.31253588e-01 6.31838381e-01 -1.49545372e-01 -1.40247449e-01 2.65700370e-01 2.44047627e-01 4.53827500e-01 4.96449601e-03 6.80081904e-01 -5.01940660e-02 7.15444833e-02 -7.00282991e-01 5.21489561e-01 7.46675372e-01 1.07870853e+00 8.04354191e-01 -4.17206921e-02 -6.96853220e-01 2.29755953e-01 1.88815340e-01 8.67219508e-01 -2.92032287e-02 -1.13126886e+00 6.58469737e-01 1.02667224e+00 3.72457296e-01 -1.12589872e+00 -3.23010474e-01 1.27799660e-01 -5.96162319e-01 4.96794760e-01 8.73048961e-01 3.04277658e-01 -1.30828249e+00 1.24983978e+00 3.22157055e-01 -6.48930609e-01 9.49914940e-03 1.06773293e+00 1.36063826e+00 9.36722517e-01 2.54098885e-02 2.62725770e-01 1.85471058e+00 -1.06211782e+00 -8.89995575e-01 -1.84762269e-01 3.50955248e-01 -5.97990572e-01 1.88731897e+00 3.09247047e-01 -1.10856831e+00 -6.91262424e-01 -1.04240382e+00 -1.09167075e+00 -6.83712602e-01 4.81808931e-02 3.86134863e-01 4.57124650e-01 -1.31298816e+00 -1.61777183e-01 -4.44326550e-01 -2.45454967e-01 6.30332470e-01 -2.12614313e-01 -4.45992887e-01 -3.83305460e-01 -6.50887549e-01 7.85374641e-01 -3.82861137e-01 -1.04618631e-01 -1.02445829e+00 -1.06222212e+00 -1.17093205e+00 2.42823943e-01 4.07380849e-01 -9.39892530e-01 1.47043908e+00 -3.03616226e-01 -1.13516963e+00 1.02093375e+00 -5.23535371e-01 -3.08104694e-01 4.76134300e-01 -4.58507925e-01 -3.64259750e-01 7.67745018e-01 1.37303531e-01 9.80721831e-01 1.09362984e+00 -1.45583498e+00 -1.29328892e-01 -4.94098037e-01 4.22088474e-01 -3.10668536e-02 6.38822541e-02 -4.27420080e-01 -6.56290948e-01 -4.04082954e-01 2.15227772e-02 -2.44636685e-01 1.72985956e-01 1.02146530e+00 -4.58016217e-01 -4.23499584e-01 1.17941523e+00 -9.43138301e-01 1.09875059e+00 -2.05660391e+00 8.08394477e-02 1.00274041e-01 6.78228974e-01 1.03971018e-02 -4.04013515e-01 5.69368601e-01 7.29480013e-02 5.97008467e-02 -1.53936788e-01 -2.44514614e-01 3.73815984e-01 2.77556896e-01 -8.22160840e-01 1.57970667e-01 5.83627105e-01 1.55955791e+00 -9.20904517e-01 -8.01542461e-01 3.94044220e-01 4.96857911e-01 -4.74678010e-01 5.80187261e-01 -7.82905340e-01 2.41171300e-01 -2.70193100e-01 1.00408494e+00 5.57199836e-01 -7.55305946e-01 -1.27267942e-01 -4.45530415e-01 -2.15981361e-02 1.95483603e-02 -7.58761823e-01 1.88803947e+00 -2.66244829e-01 1.06644154e+00 -1.30852222e-01 -6.70223355e-01 1.12703443e+00 -1.03723146e-01 -1.27223283e-01 -1.31118727e+00 -7.49790892e-02 -1.76826298e-01 -6.18817508e-01 -7.59199023e-01 6.24535978e-01 4.03872341e-01 1.72450338e-02 4.67421561e-01 1.15283318e-01 -6.59762323e-01 2.94236004e-01 8.28252256e-01 9.69915569e-01 2.48023227e-01 2.35891059e-01 -5.85831776e-02 5.06171644e-01 3.25071841e-01 -3.37711602e-01 8.37355316e-01 -3.49863440e-01 8.63565743e-01 9.72154140e-01 -6.18633270e-01 -1.23282897e+00 -1.74807572e+00 3.30078125e-01 1.18249130e+00 1.58937111e-01 -4.92079198e-01 -4.53870296e-01 -7.58003056e-01 4.45513546e-01 9.60662127e-01 -1.01615632e+00 3.55194449e-01 -3.07529449e-01 3.13109785e-01 2.04462111e-01 7.27256596e-01 3.26901734e-01 -1.08471048e+00 -8.12157333e-01 -2.63182431e-01 1.46579742e-03 -9.53525305e-01 -6.12979293e-01 -2.43043885e-01 -6.65591180e-01 -1.30840147e+00 -9.06681240e-01 -8.67422998e-01 6.97451532e-01 4.04997855e-01 1.63220274e+00 4.34614569e-02 -3.54343385e-01 1.10349989e+00 -1.38193876e-01 -3.76578838e-01 -1.90703765e-01 -2.94441015e-01 -6.90848053e-01 -3.48021723e-02 3.36438626e-01 -2.84823537e-01 -8.11364174e-01 -3.85816544e-02 -1.06967390e+00 1.42071351e-01 4.76319730e-01 6.37445092e-01 7.73980021e-01 -9.63178933e-01 4.08990413e-01 -7.75160134e-01 9.73961651e-01 -6.54990226e-02 -7.19245493e-01 8.89003277e-01 -4.71417606e-01 6.42373443e-01 4.52639252e-01 -1.57263204e-01 -9.37983155e-01 -1.15125403e-01 2.55454555e-02 -6.04158998e-01 -7.90358558e-02 1.45044297e-01 -1.84786633e-01 2.30052963e-01 7.51285136e-01 1.34352073e-01 1.00004189e-01 -2.59153008e-01 1.20189035e+00 1.28663540e-01 1.09254551e+00 -4.20270085e-01 1.00207090e+00 7.47918427e-01 -1.75057035e-02 -6.62867248e-01 -7.27274656e-01 -4.38587368e-01 -5.32740355e-01 -4.08592165e-01 1.32822919e+00 -7.12492347e-01 -1.35690916e+00 -1.32542968e-01 -1.41104329e+00 -1.05624788e-01 -4.88166541e-01 -3.01976085e-01 -6.42976761e-01 3.71968895e-01 -2.73017377e-01 -9.48247373e-01 -3.45633894e-01 -9.76151764e-01 1.44107044e+00 3.13708335e-01 -2.21989438e-01 -8.78778219e-01 -6.53044879e-02 5.26686609e-01 3.04543853e-01 5.22383630e-01 1.40811622e+00 -2.13644445e-01 -1.00627196e+00 1.57621354e-01 -8.87231290e-01 -7.23786131e-02 -9.88737214e-03 4.82431464e-02 -1.01617384e+00 -7.88493529e-02 -5.59088528e-01 -9.30463374e-01 1.01586115e+00 -4.02507838e-03 1.28149736e+00 -1.11530676e-01 -2.22174488e-02 3.92888784e-01 1.17135334e+00 1.56955436e-01 6.86509788e-01 4.86126319e-02 8.82198453e-01 5.75387657e-01 3.15860808e-01 2.36809373e-01 9.53620255e-01 1.23750336e-01 6.70386434e-01 -3.58341187e-01 -3.28369141e-01 -7.36100674e-01 3.88131291e-02 4.76333052e-01 6.13441885e-01 -1.90296531e-01 -1.12249994e+00 8.46608818e-01 -1.68360972e+00 -7.44194984e-01 -2.18520224e-01 1.70888376e+00 4.84450787e-01 -5.17245829e-02 -4.71136793e-02 1.71600848e-01 8.94360393e-02 5.48712373e-01 -1.00416553e+00 -7.46257186e-01 -2.11935684e-01 2.03569949e-01 -9.59352963e-03 3.95449817e-01 -9.81084108e-01 6.47414923e-01 6.47706270e+00 2.69418150e-01 -6.56052768e-01 -2.60956049e-01 6.44597054e-01 1.94584697e-01 -1.01518750e+00 -1.45545885e-01 -1.14784718e-01 -3.10928553e-01 6.01449490e-01 -4.04378809e-02 4.16332781e-01 6.65941834e-01 -3.22805613e-01 -1.58161938e-01 -1.31883097e+00 1.44354951e+00 5.98000884e-01 -1.76290095e+00 7.51474977e-01 -5.68459868e-01 3.72545928e-01 -5.16406953e-01 5.62128901e-01 3.96011531e-01 4.93477881e-01 -1.33979297e+00 8.22252750e-01 8.11024129e-01 1.12486851e+00 -4.11910743e-01 1.98520511e-01 -2.13417828e-01 -1.14278626e+00 -7.86250308e-02 -4.83239517e-02 8.46041143e-02 2.67995477e-01 2.00091779e-01 -6.04126036e-01 5.99726260e-01 8.76619935e-01 3.60232830e-01 -1.20992827e+00 7.24945009e-01 -4.06772137e-01 4.05793697e-01 2.34728158e-02 -2.42459774e-01 2.28506193e-01 -3.00275888e-02 1.53763577e-01 8.14208746e-01 4.73165475e-02 1.97525680e-01 -1.39568210e-01 1.14070773e+00 -2.58823127e-01 7.37653375e-02 -5.83635271e-01 -4.15170759e-01 8.95655304e-02 9.62334335e-01 -3.79733503e-01 -4.60858464e-01 -7.08847940e-01 1.13527524e+00 6.29690766e-01 8.45970213e-01 -5.15438080e-01 -4.88904744e-01 5.50128996e-01 4.05957475e-02 5.11343598e-01 -4.89166617e-01 -4.92409676e-01 -1.21213353e+00 1.63892701e-01 -9.51244950e-01 6.91789448e-01 -1.65829396e+00 -1.25627983e+00 5.97791731e-01 -1.92852825e-01 -1.10537159e+00 -2.74741799e-01 -9.67230916e-01 -3.92559677e-01 7.19176650e-01 -1.59438860e+00 -1.33917856e+00 -7.31316686e-01 8.23495984e-01 4.70825672e-01 9.99564007e-02 8.94663751e-01 -1.66954070e-01 3.06155365e-02 3.18961203e-01 -1.96549907e-01 3.37652624e-01 6.95323229e-01 -1.69407082e+00 9.19350445e-01 6.05203867e-01 6.51012540e-01 3.48999232e-01 7.55794406e-01 -2.89926589e-01 -1.89992166e+00 -7.49807537e-01 7.51527071e-01 -9.57297802e-01 7.16727972e-01 -8.90734255e-01 -1.19132221e+00 6.48724079e-01 5.58927715e-01 2.94600099e-01 4.70584154e-01 -4.34442535e-02 -1.26455581e+00 -1.56925976e-01 -8.80905092e-01 7.43384182e-01 6.37909770e-01 -1.17807019e+00 -8.82396638e-01 3.56811315e-01 1.05013800e+00 -5.68837225e-01 -4.83044595e-01 -1.83644861e-01 7.17341602e-01 -9.69031334e-01 1.09272408e+00 -8.89094234e-01 4.88633126e-01 -5.40857315e-01 -2.95877308e-01 -1.02827716e+00 -1.05698168e-01 -5.18932045e-01 -4.62178886e-01 8.02627504e-01 4.55874473e-01 -2.24910110e-01 6.51222765e-01 6.74046636e-01 2.83476293e-01 -4.20154780e-01 -8.06837797e-01 -3.39368612e-01 8.64802003e-02 -4.01581854e-01 7.33225524e-01 6.96210861e-01 -5.64790592e-02 4.14806843e-01 -2.36078680e-01 3.04806024e-01 6.44022048e-01 5.53694546e-01 1.16194725e+00 -1.05292845e+00 1.09815329e-01 -2.78613806e-01 -3.37263703e-01 -1.36969435e+00 -2.45028570e-01 -6.69112325e-01 -2.74015099e-01 -2.30307889e+00 -1.61446735e-01 4.72345263e-01 7.03386664e-02 2.16027409e-01 -1.72369942e-01 1.05704516e-01 6.00116313e-01 -6.04285449e-02 -8.27871561e-01 6.57702923e-01 1.69790006e+00 -6.12133682e-01 -2.63992827e-02 -7.95435369e-01 -6.52024329e-01 1.60989568e-01 4.45098281e-01 9.09806117e-02 -5.87722123e-01 -7.58212566e-01 5.21954000e-01 5.09019077e-01 7.60383666e-01 -7.47037113e-01 4.28380042e-01 8.75725001e-02 7.87756324e-01 -1.15258706e+00 3.50131303e-01 -5.64749360e-01 -5.51452875e-01 4.04370762e-02 -5.45199275e-01 7.29207575e-01 3.66607457e-01 7.61102378e-01 -4.04449254e-01 5.05106568e-01 2.90606350e-01 1.19627997e-01 -7.64597952e-01 8.13080743e-02 -1.50856197e-01 3.44853669e-01 7.23492146e-01 -4.34720777e-02 -9.53077137e-01 -9.07578647e-01 -6.61104381e-01 8.05989802e-01 2.30897263e-01 5.81688046e-01 1.17345476e+00 -1.39164686e+00 -5.48801661e-01 1.62817314e-01 6.87467694e-01 3.59844603e-02 5.50188243e-01 3.77339840e-01 -9.76938188e-01 5.23759305e-01 -1.74942046e-01 -6.97124958e-01 -1.07647753e+00 1.05797327e+00 6.14901744e-02 -1.76157847e-01 -6.34296238e-01 7.99085379e-01 4.71323341e-01 -4.75969255e-01 3.97753537e-01 -8.38111103e-01 -2.32721210e-01 5.86987846e-02 8.78196895e-01 -1.71846524e-02 -1.85309172e-01 -2.99763590e-01 -3.63739222e-01 5.42042255e-01 -1.75422966e-01 -2.28253812e-01 7.71792948e-01 -4.51469608e-02 2.54806310e-01 5.99335253e-01 1.21307552e+00 -7.35792667e-02 -1.43065798e+00 -1.54229209e-01 -1.49681255e-01 -4.46769744e-01 -2.45138407e-01 -1.03160262e+00 -6.68738961e-01 1.50469840e+00 4.71792668e-01 2.93021083e-01 1.03281498e+00 2.59343952e-01 5.59263527e-01 6.59165680e-01 -1.03679657e-01 -6.19773149e-01 7.61926115e-01 1.39283091e-01 1.54645598e+00 -1.38192499e+00 1.35492701e-02 -1.35612115e-01 -9.57265258e-01 1.25211358e+00 7.03840435e-01 -9.57119241e-02 3.97908777e-01 -1.32850096e-01 5.57873964e-01 -7.02785313e-01 -9.93376434e-01 -3.10609192e-01 6.92776740e-01 7.77177453e-01 1.52129143e-01 -7.10956231e-02 3.99921298e-01 1.21483371e-01 -2.64948636e-01 -2.27060780e-01 4.34514254e-01 8.30177724e-01 -2.92603612e-01 -4.10194725e-01 -5.91158628e-01 5.06542742e-01 1.57366291e-01 -3.44777927e-02 -5.98601937e-01 9.90857899e-01 -6.01302147e-01 1.08539104e+00 4.23671603e-01 -1.32600963e-01 6.29333794e-01 1.88206628e-01 4.14240867e-01 -5.35520352e-02 -3.53221983e-01 -4.90055889e-01 -9.39248055e-02 -9.31456625e-01 -1.44392356e-01 -4.61522013e-01 -1.28298604e+00 -1.21207356e-01 4.50454116e-01 -2.16576643e-03 4.95173901e-01 5.70085049e-01 3.87588203e-01 5.04948974e-01 -6.92532063e-02 -1.68288007e-01 -3.28898430e-01 -5.62649548e-01 -3.40279996e-01 7.12305188e-01 1.05120122e+00 -4.60130930e-01 -2.46140435e-01 1.19094066e-01]
[11.112095832824707, 2.0152506828308105]
12fb1034-6102-4c5e-9c03-5d2dc3b05140
an-application-of-deep-reinforcement-learning
2004.06627
null
https://arxiv.org/abs/2004.06627v3
https://arxiv.org/pdf/2004.06627v3.pdf
An Application of Deep Reinforcement Learning to Algorithmic Trading
This scientific research paper presents an innovative approach based on deep reinforcement learning (DRL) to solve the algorithmic trading problem of determining the optimal trading position at any point in time during a trading activity in stock markets. It proposes a novel DRL trading strategy so as to maximise the resulting Sharpe ratio performance indicator on a broad range of stock markets. Denominated the Trading Deep Q-Network algorithm (TDQN), this new trading strategy is inspired from the popular DQN algorithm and significantly adapted to the specific algorithmic trading problem at hand. The training of the resulting reinforcement learning (RL) agent is entirely based on the generation of artificial trajectories from a limited set of stock market historical data. In order to objectively assess the performance of trading strategies, the research paper also proposes a novel, more rigorous performance assessment methodology. Following this new performance assessment approach, promising results are reported for the TDQN strategy.
['Thibaut Théate', 'Damien Ernst']
2020-04-07
null
null
null
null
['algorithmic-trading']
['time-series']
[-4.89914387e-01 -1.79670483e-01 -6.67853281e-03 1.68414816e-01 -4.18196112e-01 -6.49917960e-01 7.98121035e-01 1.71014126e-02 -5.76441526e-01 1.04788768e+00 -1.95249513e-01 -4.93213326e-01 -6.41062200e-01 -1.26106358e+00 -4.45721477e-01 -6.32174253e-01 -1.79428309e-01 6.69551611e-01 3.78117301e-02 -4.62893903e-01 6.91621125e-01 7.06844032e-01 -1.23727596e+00 -3.08547646e-01 8.10306311e-01 1.41816831e+00 -1.82655379e-01 4.66637820e-01 -2.83631712e-01 9.24598873e-01 -8.86891127e-01 -2.59520978e-01 8.38680387e-01 -7.68176079e-01 -4.26989287e-01 -2.68939119e-02 -3.29406470e-01 -5.11965215e-01 6.15402274e-02 8.97696555e-01 3.89380485e-01 2.46368706e-01 4.82216299e-01 -1.16566479e+00 -2.19118908e-01 6.50130451e-01 -1.98138475e-01 5.46139538e-01 -1.57066301e-01 1.66818172e-01 1.09080279e+00 -5.54559112e-01 2.15843961e-01 6.46312535e-01 4.37452883e-01 6.70203343e-02 -1.00917494e+00 -7.21097529e-01 -3.09733659e-01 -1.46896141e-02 -7.01491952e-01 2.36141711e-01 8.42626810e-01 -3.19482714e-01 8.60963702e-01 -5.56166209e-02 1.26717615e+00 6.29680514e-01 6.15954995e-01 5.42966247e-01 1.38060451e+00 -2.29900107e-01 7.66261160e-01 -1.56589508e-01 -2.93739200e-01 7.13699013e-02 4.58084345e-01 1.00214326e+00 -1.31263420e-01 -1.91386163e-01 1.02887452e+00 -9.64545980e-02 3.47479612e-01 -6.15863562e-01 -8.33132148e-01 1.31088102e+00 2.92309672e-01 6.33671284e-01 -8.57531726e-01 1.16478115e-01 4.87761289e-01 1.00483835e+00 5.39653420e-01 8.92250955e-01 -6.60092771e-01 -3.40180784e-01 -1.16631413e+00 8.53453338e-01 1.07764113e+00 9.80465263e-02 4.00133997e-01 7.94006824e-01 -1.48003653e-01 3.67059737e-01 3.80660266e-01 5.06895602e-01 7.14938641e-01 -1.18544042e+00 7.46659786e-02 4.24698561e-01 4.94972944e-01 -6.53078914e-01 -4.77306992e-01 -7.52189040e-01 -4.23120767e-01 9.36389983e-01 5.67157507e-01 -5.75411797e-01 -2.69410133e-01 1.25050616e+00 3.05835694e-01 4.76059355e-02 3.86261880e-01 4.86915261e-01 -2.20864322e-02 8.37333739e-01 -2.46707410e-01 -5.17296731e-01 7.98301697e-01 -7.14989126e-01 -4.94107187e-01 5.39747536e-01 1.58010706e-01 -4.76569325e-01 4.88132864e-01 6.71388566e-01 -1.21123123e+00 -5.49321830e-01 -1.16121435e+00 9.24136579e-01 -5.82163274e-01 -5.10669827e-01 2.25296631e-01 6.37047172e-01 -1.22335231e+00 1.05637681e+00 -3.74810487e-01 2.65412331e-01 4.45192680e-02 5.30953646e-01 4.24249262e-01 8.52618217e-01 -1.56798601e+00 1.08462203e+00 6.55574083e-01 1.30445749e-01 -8.13050807e-01 -6.80936098e-01 -3.80156130e-01 5.40676862e-02 5.13049841e-01 -2.24102288e-01 1.62091780e+00 -1.33574295e+00 -2.10292816e+00 5.06764114e-01 9.16097462e-01 -1.30158150e+00 1.17264259e+00 -5.37050106e-02 -3.15680116e-01 1.51787713e-01 -6.90615326e-02 2.96835691e-01 1.11792886e+00 -8.09994221e-01 -8.43460739e-01 -5.70250601e-02 5.00432700e-02 1.53063461e-01 -3.58722243e-03 -2.37615228e-01 6.17603838e-01 -1.27259219e+00 -6.20511234e-01 -6.35801077e-01 -1.82255208e-01 -5.76864362e-01 3.43955815e-01 -4.32154089e-01 6.92837954e-01 -4.27275270e-01 9.47216153e-01 -1.79134965e+00 -2.44665980e-01 5.33434749e-01 -3.04830056e-02 5.00791669e-01 -9.28749666e-02 7.07980633e-01 -4.97734956e-02 -3.37637812e-01 -3.61492664e-01 5.29811621e-01 3.10055017e-01 1.63048938e-01 -5.56779087e-01 2.31442228e-01 1.30630240e-01 1.10794413e+00 -9.31941271e-01 -8.70545115e-03 2.57038832e-01 -1.35614038e-01 -2.59432793e-01 3.73568356e-01 -7.19136000e-01 2.76053011e-01 -5.03828466e-01 3.14740419e-01 5.85855007e-01 1.43325537e-01 -8.50454867e-02 4.36604261e-01 -5.97216606e-01 -9.78762656e-02 -1.15757704e+00 8.70652080e-01 -3.84314805e-01 3.24685246e-01 -1.91028759e-01 -1.18427193e+00 1.40321732e+00 3.70558798e-01 9.20054197e-01 -1.28236425e+00 1.95228487e-01 6.95870578e-01 1.84409887e-01 3.16818506e-02 3.65651906e-01 -7.06157804e-01 2.83180699e-02 9.62605715e-01 -6.86262920e-02 -1.81406364e-01 5.01407981e-01 -5.80823183e-01 8.85167480e-01 3.10303003e-01 3.49765748e-01 -4.84619975e-01 6.48036182e-01 5.55690005e-02 2.99355984e-01 7.05813885e-01 -4.31325972e-01 -2.97800809e-01 6.78189397e-01 -8.10656190e-01 -1.16584575e+00 -1.18303382e+00 -8.41453448e-02 7.66366243e-01 -1.54492557e-01 5.46968043e-01 -7.31901884e-01 -5.26167393e-01 4.06558186e-01 8.27605188e-01 -6.80436969e-01 -2.01451294e-02 -5.25057018e-01 -7.28091300e-01 2.66446829e-01 2.59365082e-01 8.61516654e-01 -1.91322649e+00 -1.33656466e+00 8.27500403e-01 7.44747460e-01 -4.25945491e-01 -9.95060652e-02 2.81920075e-01 -1.03485096e+00 -1.18584991e+00 -1.02672851e+00 -5.02413988e-01 -1.07040636e-01 -4.09919232e-01 1.10641289e+00 -1.85855150e-01 2.71256417e-01 3.80371511e-01 -4.74680215e-01 -6.25333548e-01 -7.89191365e-01 5.11925779e-02 -1.19696520e-01 9.52911973e-02 3.39735150e-01 -4.87031668e-01 -5.48050344e-01 2.15324745e-01 -1.13839996e+00 -6.42881274e-01 7.23075628e-01 8.88300776e-01 3.02788645e-01 5.09519339e-01 1.26304221e+00 -5.03343761e-01 1.14901400e+00 -4.24986035e-01 -1.57069921e+00 8.06265026e-02 -1.17318773e+00 3.75713229e-01 8.90320301e-01 -1.94283918e-01 -9.13404047e-01 -5.20042062e-01 -1.19839638e-01 -2.77884007e-01 2.42496505e-01 3.90936792e-01 3.23576927e-01 -8.00261721e-02 5.10652512e-02 5.71097255e-01 4.46266592e-01 -3.24719071e-01 8.18918794e-02 1.97195396e-01 2.84698308e-01 -2.79592156e-01 8.16639245e-01 9.96761918e-02 6.01497069e-02 -3.57281029e-01 -4.30961102e-01 -3.12973075e-02 -6.08812630e-01 -3.74393344e-01 5.67116201e-01 -2.32159629e-01 -1.01884842e+00 9.61842775e-01 -6.69178486e-01 -6.38634920e-01 -7.94637144e-01 4.25570101e-01 -1.10868943e+00 -6.08984977e-02 -5.60175896e-01 -9.47336078e-01 -5.17954886e-01 -7.64528394e-01 3.26208353e-01 2.44913161e-01 -4.47122306e-02 -1.31746769e+00 7.84758687e-01 -1.82685688e-01 5.88039339e-01 6.67542398e-01 7.47096896e-01 -1.12064433e+00 -4.52892929e-01 -1.17219046e-01 1.04988277e-01 7.50694931e-01 1.14670575e-01 -1.31497066e-02 -4.61912811e-01 -3.39836717e-01 5.81018567e-01 -1.90459475e-01 5.01119614e-01 5.39107263e-01 3.75378877e-01 -2.45378017e-01 4.96927112e-01 2.09150225e-01 1.58486605e+00 1.03042519e+00 3.76770526e-01 1.14040399e+00 -2.85617225e-02 4.96453583e-01 9.66569781e-01 7.59611011e-01 -1.92619994e-01 2.39984259e-01 5.44468701e-01 2.29064524e-01 6.83247685e-01 -6.33248463e-02 4.99061912e-01 4.61528152e-01 1.46550551e-01 1.20065540e-01 -6.73269331e-01 2.63143480e-01 -1.58431876e+00 -1.23986232e+00 4.65415627e-01 2.08507872e+00 6.72820568e-01 7.82146752e-01 8.13412070e-01 3.10360402e-01 5.19850791e-01 3.16232949e-01 -9.01452184e-01 -9.31746185e-01 -6.63410723e-02 6.69594228e-01 6.55743659e-01 3.69374454e-01 -9.45153058e-01 5.47268391e-01 6.39676619e+00 7.80895770e-01 -1.22378421e+00 -3.82191956e-01 5.95782995e-01 1.11131772e-01 -2.62681514e-01 -3.28532606e-01 -4.95624036e-01 5.70172250e-01 1.41955805e+00 -5.89184165e-01 5.06810367e-01 5.73139310e-01 5.00394285e-01 -7.09784701e-02 -5.95432103e-01 5.72330952e-01 -5.86281419e-01 -1.65164137e+00 1.21221252e-01 2.97907859e-01 6.52653098e-01 -8.84127915e-02 3.15446228e-01 4.97572273e-01 3.58000755e-01 -7.81920791e-01 9.42951441e-01 7.46505320e-01 6.58246949e-02 -1.40949833e+00 9.23914313e-01 2.58912355e-01 -9.18937266e-01 -5.29283106e-01 -6.07158393e-02 -1.11062691e-01 -1.74161404e-01 2.63630271e-01 -5.89801073e-01 4.88513112e-01 3.54648054e-01 6.32093251e-01 -3.02388638e-01 1.03330743e+00 -8.70318897e-03 5.90565979e-01 -1.05022259e-01 -5.57211936e-01 9.69181418e-01 -7.82130480e-01 4.84064698e-01 7.08981454e-01 4.76295888e-01 -1.73071951e-01 -1.00448504e-01 1.19414985e+00 5.60226440e-02 1.21712714e-01 -6.15370035e-01 -1.40536889e-01 2.55393952e-01 9.58671510e-01 -1.06394684e+00 -1.66663393e-01 -2.04665214e-01 6.43121302e-01 -1.80600494e-01 -1.57192349e-02 -6.00946188e-01 -3.73735100e-01 3.78260255e-01 6.68049231e-03 7.40580440e-01 -3.38272639e-02 -7.11567029e-02 -5.60626328e-01 -5.91612130e-04 -8.91198575e-01 4.69675511e-01 -2.23498836e-01 -1.31849885e+00 3.86027485e-01 -5.16116954e-02 -1.51258969e+00 -9.08011615e-01 -6.54380739e-01 -8.48196566e-01 6.69258118e-01 -1.76390326e+00 -2.25458086e-01 4.77001816e-01 4.36992109e-01 4.22464550e-01 -8.04331839e-01 4.78608549e-01 -1.48138091e-01 -1.99723631e-01 7.08352253e-02 8.78953636e-01 1.68252781e-01 5.69842989e-03 -1.71090543e+00 5.14259815e-01 5.02728522e-01 1.41336665e-01 -1.30777434e-01 6.69396639e-01 -6.34190261e-01 -9.52157974e-01 -7.36711621e-01 3.80139202e-01 -1.86194658e-01 1.25607252e+00 2.67467409e-01 -7.88915336e-01 8.65770578e-02 5.68671882e-01 -3.87649029e-01 3.69811684e-01 -6.96551383e-01 1.65073350e-01 -5.03242314e-01 -1.18328118e+00 2.65222579e-01 6.62425011e-02 -9.03245285e-02 -9.94333684e-01 -2.82146156e-01 5.57222903e-01 -1.10821210e-01 -1.00629008e+00 3.05308670e-01 4.65639055e-01 -1.35061073e+00 5.80722928e-01 -2.89119214e-01 2.03564167e-01 -1.32533722e-02 3.20345342e-01 -1.44165909e+00 6.91977367e-02 -1.01169968e+00 -3.78792226e-01 6.84271693e-01 2.17310563e-01 -1.01778543e+00 9.09967959e-01 7.37103969e-02 2.47531027e-01 -8.75649571e-01 -1.21322227e+00 -1.19169533e+00 5.87313950e-01 2.22814996e-02 8.80839586e-01 6.73158765e-01 -4.14444059e-01 -1.85583889e-01 -1.25786483e-01 -4.49181437e-01 8.32790017e-01 5.32871246e-01 3.10860842e-01 -1.24210525e+00 -3.63696843e-01 -1.03954220e+00 -1.26262829e-01 -3.03457528e-01 1.76549509e-01 -4.38434780e-01 -2.06585303e-01 -9.05810952e-01 -6.15210414e-01 -9.15836319e-02 -9.61541057e-01 -1.99463680e-01 3.36710006e-01 -1.00681081e-01 3.97647440e-01 7.49601126e-02 -2.60575354e-01 6.81880713e-01 1.25498605e+00 -7.67654777e-02 -3.17941368e-01 7.00540900e-01 -3.16621661e-01 4.11019832e-01 1.09290588e+00 -2.45845571e-01 -4.75208282e-01 3.95960510e-01 3.50863248e-01 5.57693958e-01 3.59497160e-01 -9.66305733e-01 -9.42297280e-02 -2.13136822e-01 3.27237517e-01 -8.96143675e-01 -3.22917402e-01 -7.97711194e-01 6.84561878e-02 1.28199780e+00 -3.96079838e-01 6.78826571e-01 1.60906583e-01 4.12923843e-01 -6.35015547e-01 -4.45259094e-01 9.73652601e-01 -2.43225187e-01 -8.45550537e-01 2.21115202e-01 -5.69673479e-01 2.99806535e-01 1.44995189e+00 -4.21329468e-01 3.29568624e-01 -3.70911062e-01 -7.02026606e-01 3.48539799e-01 2.04133153e-01 2.33892635e-01 5.64020097e-01 -1.31730628e+00 -7.82954633e-01 1.05018698e-01 -3.57349426e-01 -5.57004929e-01 -2.47831911e-01 3.31883192e-01 -9.52252388e-01 6.05495989e-01 -7.35494733e-01 7.51348659e-02 -3.64367127e-01 5.47381341e-01 9.98845756e-01 -8.28590214e-01 -5.54723680e-01 3.32089722e-01 -4.98581856e-01 -2.09509552e-01 -7.12450072e-02 -3.64883929e-01 -3.33949357e-01 5.84527254e-01 3.62266213e-01 7.19966114e-01 1.56406432e-01 -2.66821146e-01 1.55530378e-01 3.52066547e-01 3.00589561e-01 -5.81594169e-01 1.63365376e+00 3.72439921e-01 -2.51458678e-02 5.22574723e-01 1.01806819e+00 -2.52918869e-01 -1.80377829e+00 2.17703916e-02 7.18479633e-01 -7.70249777e-03 -1.72588453e-01 -8.12228858e-01 -1.12823045e+00 3.57752860e-01 6.61801696e-01 9.50009108e-01 9.80730355e-01 -7.13172138e-01 9.91861105e-01 3.02984476e-01 4.28440064e-01 -1.62045205e+00 1.16617642e-01 5.93610466e-01 8.19729567e-01 -9.27745342e-01 -3.20621163e-01 7.59156764e-01 -3.88091654e-01 1.49441826e+00 2.11930558e-01 -8.74504983e-01 1.00618267e+00 3.23013484e-01 2.52408177e-01 -8.89820457e-02 -5.57186782e-01 -2.93903612e-02 -6.05143309e-02 3.16342890e-01 -5.64093962e-02 -5.90124652e-02 -5.36767960e-01 1.57948822e-01 -3.14470142e-01 2.57247180e-01 4.76944894e-01 9.48748648e-01 -5.54714918e-01 -1.19501770e+00 -4.10567403e-01 4.58946526e-01 -4.15424258e-01 1.81639701e-01 -3.95244986e-01 1.21104157e+00 -1.20469682e-01 4.95305330e-01 2.72003800e-01 1.20573267e-01 4.49705780e-01 -3.98618877e-02 3.04592282e-01 -9.93592218e-02 -1.13524091e+00 1.50713161e-01 -4.40600216e-01 -3.69790971e-01 -5.75421631e-01 -8.66473198e-01 -1.35719252e+00 -2.94520229e-01 1.77801400e-01 6.50772989e-01 2.77468503e-01 1.04912460e+00 -1.76374137e-01 3.50246370e-01 1.24277985e+00 -6.56554997e-01 -1.39007151e+00 -6.22567475e-01 -1.28772604e+00 1.50490791e-01 5.72992265e-01 -5.67205906e-01 -2.28771880e-01 -6.73722565e-01]
[4.451768398284912, 3.921389102935791]
f99e7c13-259d-4bf1-88f6-f7a6bbabf91e
a-weighting-scheme-for-open-information
null
null
https://aclanthology.org/N12-2011
https://aclanthology.org/N12-2011.pdf
A Weighting Scheme for Open Information Extraction
null
['Yuval Merhav']
2012-06-01
null
null
null
naacl-2012-6
['text-clustering']
['natural-language-processing']
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
[-7.379974365234375, 3.756701946258545]
8f0deb00-62e6-452d-8b01-db24f2486e24
190503678
1905.03678
null
https://arxiv.org/abs/1905.03678v1
https://arxiv.org/pdf/1905.03678v1.pdf
What Do Single-view 3D Reconstruction Networks Learn?
Convolutional networks for single-view object reconstruction have shown impressive performance and have become a popular subject of research. All existing techniques are united by the idea of having an encoder-decoder network that performs non-trivial reasoning about the 3D structure of the output space. In this work, we set up two alternative approaches that perform image classification and retrieval respectively. These simple baselines yield better results than state-of-the-art methods, both qualitatively and quantitatively. We show that encoder-decoder methods are statistically indistinguishable from these baselines, thus indicating that the current state of the art in single-view object reconstruction does not actually perform reconstruction but image classification. We identify aspects of popular experimental procedures that elicit this behavior and discuss ways to improve the current state of research.
['René Ranftl', 'Zhuwen Li', 'Vladlen Koltun', 'Thomas Brox', 'Maxim Tatarchenko', 'Stephan R. Richter']
2019-05-09
what-do-single-view-3d-reconstruction
http://openaccess.thecvf.com/content_CVPR_2019/html/Tatarchenko_What_Do_Single-View_3D_Reconstruction_Networks_Learn_CVPR_2019_paper.html
http://openaccess.thecvf.com/content_CVPR_2019/papers/Tatarchenko_What_Do_Single-View_3D_Reconstruction_Networks_Learn_CVPR_2019_paper.pdf
cvpr-2019-6
['single-view-3d-reconstruction']
['computer-vision']
[-8.78283195e-03 -6.90271556e-02 -1.63154036e-01 -6.03817046e-01 -8.44851792e-01 -6.18676424e-01 9.34599042e-01 -3.06164950e-01 -2.18525514e-01 3.44825178e-01 4.82956529e-01 -2.72645652e-01 1.23587780e-01 -7.45334804e-01 -1.16897440e+00 -5.21048486e-01 1.10900573e-01 5.85969388e-01 1.35327712e-01 -7.19635189e-02 4.25691843e-01 6.17359042e-01 -1.65546381e+00 8.19765747e-01 -9.07430053e-02 1.07560480e+00 -9.00933519e-02 8.60278904e-01 2.22147897e-01 9.87767518e-01 -3.88666630e-01 -7.62069345e-01 4.31037903e-01 -2.74383456e-01 -1.03616750e+00 1.68689162e-01 9.47067261e-01 -9.26428795e-01 -5.86606264e-01 9.62700009e-01 5.93369901e-01 -2.08428949e-01 7.64922678e-01 -9.49745178e-01 -1.00018942e+00 2.71094054e-01 -3.43529224e-01 2.79307097e-01 4.54775304e-01 -6.40534982e-02 1.11274242e+00 -1.22122109e+00 8.77403259e-01 1.37101221e+00 6.83633864e-01 3.92008454e-01 -1.26143944e+00 -4.12234485e-01 -3.56792798e-03 2.45571330e-01 -1.34303677e+00 -6.80952191e-01 6.30823791e-01 -4.81304795e-01 1.28773379e+00 1.58704951e-01 5.14730930e-01 1.12957895e+00 3.29498440e-01 1.07167292e+00 1.05507147e+00 -3.32307577e-01 -1.06249482e-03 1.25072718e-01 -5.61001897e-02 8.92362177e-01 2.92622000e-01 3.04763258e-01 -6.20463729e-01 6.61320910e-02 1.02773917e+00 6.17407216e-03 -2.36496195e-01 -9.42680478e-01 -1.30724120e+00 7.40840375e-01 5.00187039e-01 1.67024240e-01 -1.77370593e-01 4.10091549e-01 6.37797892e-01 3.43898863e-01 5.73795974e-01 3.47185880e-01 -2.81104296e-01 6.53079376e-02 -1.06034768e+00 3.07000399e-01 7.30485439e-01 1.02129591e+00 5.23358703e-01 3.66780348e-02 4.56500575e-02 6.41129136e-01 3.73137921e-01 1.12081870e-01 2.81604588e-01 -1.27476358e+00 3.75402123e-01 3.08028847e-01 1.82760879e-01 -8.54259014e-01 -1.37981549e-01 -4.98511374e-01 -5.76061070e-01 4.07828122e-01 2.65200853e-01 3.74452651e-01 -9.87958848e-01 1.53909063e+00 -1.56475902e-01 -1.51445016e-01 2.50009000e-01 9.68412519e-01 8.34334910e-01 4.69039679e-01 -3.01700264e-01 2.72874206e-01 1.27044976e+00 -1.04795563e+00 -4.87575084e-01 -1.03435600e-02 3.80299926e-01 -8.13091159e-01 8.65473509e-01 5.16265690e-01 -1.58441210e+00 -6.55146778e-01 -1.23502052e+00 -4.18234020e-01 -2.35934973e-01 2.19945937e-01 7.41080463e-01 4.50128704e-01 -1.33369195e+00 7.65277743e-01 -8.98120761e-01 -3.56618136e-01 6.39815331e-01 2.84815520e-01 -7.29602218e-01 -2.53700376e-01 -6.63354814e-01 1.29816175e+00 2.18520269e-01 3.97767052e-02 -1.28600776e+00 -4.96635437e-01 -7.55726457e-01 4.97654416e-02 1.02640934e-01 -1.07981193e+00 1.66868782e+00 -9.65408504e-01 -1.25654244e+00 1.43088245e+00 -1.49884820e-01 -5.16979337e-01 4.92804170e-01 -3.86903018e-01 -9.91658643e-02 3.99297476e-01 -3.26637132e-03 8.92428756e-01 8.04927409e-01 -1.75407255e+00 -2.61763602e-01 -4.68366861e-01 3.31192553e-01 2.64665663e-01 -9.89518091e-02 -6.21251389e-02 -5.20624220e-01 -5.16613603e-01 3.59601974e-01 -7.50922620e-01 1.01140581e-01 4.79654759e-01 -4.87940729e-01 -2.76332516e-02 5.26746392e-01 -4.70314115e-01 7.40417480e-01 -2.09499478e+00 2.86402494e-01 -1.89218640e-01 3.45643908e-01 6.19381890e-02 -1.23498805e-01 7.10001349e-01 -3.83007526e-01 2.74984568e-01 1.59414425e-01 -6.17933452e-01 3.61244865e-02 1.50863409e-01 -5.09625435e-01 7.89458096e-01 -2.48013772e-02 9.70480800e-01 -5.77951968e-01 -3.40557039e-01 2.48823896e-01 4.18707639e-01 -7.44270504e-01 3.56659174e-01 -2.37260476e-01 1.91657797e-01 -1.45682976e-01 7.70600855e-01 6.17538214e-01 -5.05514622e-01 1.16595462e-01 -5.22385299e-01 2.49604471e-02 6.69656098e-01 -1.01785624e+00 1.94695926e+00 -2.88412035e-01 9.43423033e-01 6.84603751e-02 -1.21728313e+00 5.21502554e-01 3.82171750e-01 4.54709381e-01 -5.38189530e-01 8.23538825e-02 1.35699794e-01 -2.66804826e-02 -6.04155838e-01 5.17110944e-01 -3.91852409e-01 3.68929833e-01 5.71986258e-01 2.69870311e-01 -1.48286104e-01 -2.14957044e-01 1.77095875e-01 1.03456116e+00 4.16029990e-01 3.17334235e-01 -2.40044743e-01 2.45124981e-01 -5.76953888e-02 -6.58995658e-02 8.22320223e-01 -9.34684798e-02 9.96961176e-01 3.32499951e-01 -6.60344541e-01 -1.36514878e+00 -1.36960566e+00 -1.64118618e-01 8.03393900e-01 2.23432317e-01 -4.45669025e-01 -6.73940599e-01 -6.36251092e-01 -7.74805024e-02 4.77699757e-01 -6.32923007e-01 -5.67155853e-02 -4.91638124e-01 -2.89472371e-01 7.27515876e-01 8.78237605e-01 6.55015588e-01 -8.32218647e-01 -7.45646536e-01 -5.25713339e-02 -2.08139345e-01 -1.16433716e+00 -8.16102885e-03 2.45641753e-01 -1.09965944e+00 -1.01849496e+00 -7.64290631e-01 -7.10712910e-01 6.81562781e-01 6.56719923e-01 1.67189455e+00 1.54263437e-01 -1.16659015e-01 6.83010757e-01 -2.05402941e-01 -3.02964538e-01 -4.76480424e-01 -1.17254406e-01 -1.40550453e-02 -3.67122352e-01 2.59114444e-01 -6.40343189e-01 -8.36369216e-01 3.37129354e-01 -1.03601170e+00 1.80957004e-01 6.15132570e-01 7.50160992e-01 6.04802310e-01 -2.44019747e-01 1.65666088e-01 -8.10031593e-01 2.07784116e-01 -3.05670977e-01 -5.18644750e-01 2.52652436e-01 -5.11610806e-01 3.01232010e-01 5.36348999e-01 -9.32286680e-02 -8.20770681e-01 1.19871005e-01 -3.92612547e-01 -8.25043321e-01 -4.08018202e-01 2.84433633e-01 3.80328856e-02 -1.11531205e-01 6.21236622e-01 3.14843714e-01 9.12407115e-02 -6.24533117e-01 4.24828708e-01 5.03550529e-01 4.20379877e-01 -3.50948244e-01 6.24147356e-01 8.90809357e-01 -1.57132939e-01 -3.64082634e-01 -1.08827353e+00 -4.55850840e-01 -6.30911708e-01 -1.48666218e-01 7.85496414e-01 -1.23048699e+00 -7.25332975e-01 3.57233495e-01 -1.49391484e+00 -1.31285563e-01 -2.67061710e-01 4.20098126e-01 -9.09940302e-01 3.48627716e-01 -7.18459249e-01 -4.96306837e-01 -7.52445757e-02 -1.34444439e+00 1.49877954e+00 -3.21276039e-01 -4.87599559e-02 -8.81594360e-01 5.71251102e-02 3.92460346e-01 4.68772560e-01 -9.94630083e-02 8.64440024e-01 -5.63635767e-01 -9.05759871e-01 -1.88291259e-02 -5.17488360e-01 4.59816068e-01 -2.06211388e-01 -1.57097906e-01 -1.25456119e+00 -4.19249892e-01 1.74863815e-01 -6.23206675e-01 1.13543606e+00 3.02036881e-01 1.22477257e+00 -1.66142762e-01 -3.15502107e-01 6.99554920e-01 1.68266726e+00 -5.62053993e-02 9.95972872e-01 2.68740177e-01 4.76707369e-01 4.25032675e-01 1.76652342e-01 1.58488438e-01 4.53666985e-01 8.22669327e-01 7.37961888e-01 -1.01813681e-01 -3.25866014e-01 -3.85911286e-01 3.91227871e-01 7.15054512e-01 -9.99491066e-02 -5.30886352e-01 -8.08034182e-01 6.12769783e-01 -1.66323698e+00 -1.21364522e+00 8.25165287e-02 1.97674072e+00 4.84396905e-01 1.75967619e-01 -7.77937919e-02 4.09966074e-02 2.53212005e-01 3.58798206e-01 -3.25388163e-01 -3.68035555e-01 5.16800284e-02 1.78652838e-01 4.92480695e-01 3.12535346e-01 -1.12886345e+00 7.17052579e-01 8.37897491e+00 4.57077712e-01 -1.03499591e+00 -3.70653234e-02 5.39843261e-01 -3.73331085e-03 -2.46460155e-01 -1.20900176e-01 -7.31179833e-01 -1.44428194e-01 6.43533587e-01 3.95806849e-01 4.35030997e-01 1.04029381e+00 -2.50695348e-01 -1.65658370e-01 -1.72926140e+00 1.12263715e+00 5.45543909e-01 -1.57467520e+00 3.12520653e-01 4.00396585e-02 6.86063230e-01 3.67072374e-01 2.56674737e-01 7.90569410e-02 3.39840859e-01 -1.17054570e+00 9.28484023e-01 5.39662361e-01 7.92928278e-01 -2.24024400e-01 5.93283653e-01 2.76251167e-01 -8.94430518e-01 8.03771839e-02 -3.96309555e-01 4.36271820e-03 1.46016851e-01 2.75447994e-01 -5.35129428e-01 5.83629727e-01 8.01612377e-01 9.01600778e-01 -6.51453853e-01 8.96986842e-01 7.86676779e-02 4.60299760e-01 -9.69777107e-02 2.78778762e-01 1.59093961e-01 1.93529397e-01 4.87620145e-01 1.06703544e+00 9.39848125e-02 -1.18963532e-01 1.89756546e-02 1.01314020e+00 -3.09661597e-01 -2.73600221e-01 -1.26036251e+00 1.17563419e-01 -1.17272660e-02 6.78096712e-01 -7.26969719e-01 -4.74805295e-01 -7.41888285e-01 1.06119442e+00 5.73859215e-01 2.68148452e-01 -6.70036435e-01 2.11839527e-01 6.99930727e-01 2.12530583e-01 6.69105470e-01 -3.73712152e-01 -4.21806425e-01 -1.43592751e+00 1.37874931e-01 -1.00484598e+00 9.79699567e-02 -1.05617321e+00 -1.40643990e+00 4.89029825e-01 2.52167135e-01 -1.37516785e+00 -3.61041784e-01 -1.02089226e+00 -2.59339660e-01 3.63978714e-01 -1.61404848e+00 -9.35929537e-01 -2.37112358e-01 4.96870786e-01 7.14401126e-01 -1.99218243e-01 1.12208033e+00 3.42298001e-01 -7.76434317e-02 4.40231919e-01 1.72918543e-01 2.12841988e-01 5.94244480e-01 -1.09165037e+00 5.00950277e-01 6.29716754e-01 6.47897482e-01 7.53908753e-01 6.91598654e-01 -3.50104183e-01 -1.72029150e+00 -5.66104412e-01 5.80163598e-01 -6.56762719e-01 3.01605880e-01 -4.69867229e-01 -6.14647508e-01 1.03772604e+00 4.84273583e-01 1.56939402e-01 5.31654716e-01 6.78807497e-02 -7.83290267e-01 -5.53868385e-03 -8.56909811e-01 4.82938468e-01 1.08456826e+00 -9.69177067e-01 -5.57115972e-01 4.32214588e-01 3.52294713e-01 -6.03059888e-01 -6.93132639e-01 5.19547462e-01 9.26975727e-01 -1.56071365e+00 1.34671271e+00 -7.87298381e-01 8.71299744e-01 6.24557259e-03 -7.02211499e-01 -1.08531380e+00 -2.89244354e-01 -1.00859754e-01 -9.60049778e-02 6.80284619e-01 2.50526816e-01 -3.43196779e-01 8.99691463e-01 3.15164328e-01 -2.81543106e-01 -8.83428454e-01 -7.38163173e-01 -7.47193456e-01 2.31235728e-01 -4.36742038e-01 1.21745571e-01 4.98011708e-01 -3.74491960e-01 4.25452858e-01 -2.82697558e-01 -3.73660177e-02 3.96667868e-01 3.21614623e-01 9.30419624e-01 -9.59930301e-01 -3.93079102e-01 -4.23917234e-01 -7.07425117e-01 -1.43376732e+00 1.45952523e-01 -9.81227875e-01 6.68608397e-02 -1.75567627e+00 4.65456754e-01 -3.26127224e-02 -8.38124007e-02 1.59800500e-01 3.70299369e-01 6.29808486e-01 1.31528825e-01 3.15087438e-01 -8.12081873e-01 3.54981571e-01 1.32428110e+00 -8.57435353e-03 4.95706707e-01 -1.95236251e-01 -6.36071801e-01 8.19193304e-01 5.72709024e-01 -3.83778751e-01 -3.72976780e-01 -8.81944537e-01 3.70435089e-01 6.50622249e-02 8.02035630e-01 -9.33392465e-01 8.40798542e-02 3.27666581e-01 6.31927788e-01 -6.89441562e-01 6.87443733e-01 -9.59522128e-01 1.68065608e-01 3.92474294e-01 -5.00790834e-01 3.11251968e-01 1.36591598e-01 5.70413470e-01 -4.03816432e-01 -2.55838335e-01 8.29828143e-01 -6.04603887e-01 -5.06194353e-01 1.19095817e-01 -4.22648698e-01 -4.90324572e-02 7.66672850e-01 -2.84446180e-01 -3.83581370e-01 -7.74700284e-01 -7.19339252e-01 -3.58590603e-01 6.07399046e-01 1.92796066e-01 6.72731936e-01 -1.27494931e+00 -6.61373734e-01 1.88134536e-01 6.97069094e-02 -2.49327004e-01 -1.36401966e-01 6.48136139e-01 -7.43267715e-01 5.47844052e-01 -2.99615383e-01 -7.24128842e-01 -1.08726072e+00 7.28997767e-01 4.90003824e-01 -2.66474098e-01 -8.26028943e-01 7.43090272e-01 3.98711622e-01 -3.51986617e-01 5.26962936e-01 -1.36050820e-01 8.01803023e-02 -2.19091520e-01 2.84113854e-01 9.40712690e-02 1.85883209e-01 -5.71954370e-01 -1.94953293e-01 5.88863611e-01 -1.94629639e-01 -2.65651643e-01 1.42363989e+00 -3.08455038e-03 -5.29378057e-02 4.84554738e-01 1.52206028e+00 -3.00042957e-01 -9.91126120e-01 -9.89519432e-02 -2.36093894e-01 -5.76766133e-01 9.79471579e-02 -7.04217732e-01 -1.01858771e+00 1.20627952e+00 5.88844717e-01 2.43508175e-01 9.46340501e-01 1.77171260e-01 6.87720358e-01 6.31190002e-01 3.06907773e-01 -6.85778022e-01 3.55077803e-01 5.16767859e-01 1.10529268e+00 -1.29438591e+00 3.08699340e-01 -3.77900392e-01 -3.87944490e-01 1.22009921e+00 6.15630150e-01 -5.65206945e-01 8.68325531e-01 4.36992407e-01 -3.28149907e-02 -7.01774597e-01 -7.76708245e-01 -2.36556515e-01 2.87333339e-01 4.37105030e-01 6.44880414e-01 -2.86636949e-01 9.97683555e-02 1.43770084e-01 -1.24101967e-01 -6.11942075e-02 2.38954395e-01 9.24884617e-01 -1.47534996e-01 -9.38426077e-01 -2.37141803e-01 4.48251277e-01 -6.65666640e-01 -5.58183491e-02 -6.40728295e-01 9.59030330e-01 -2.27000996e-01 5.20430624e-01 8.50068107e-02 -4.64456111e-01 3.32478553e-01 1.84732124e-01 8.35499704e-01 -6.15117192e-01 -5.70679665e-01 2.71345582e-02 2.10720211e-01 -8.43024135e-01 -6.79255545e-01 -5.83276570e-01 -8.83498192e-01 -3.58657211e-01 -4.80681092e-01 -4.42393780e-01 7.07552731e-01 8.60689819e-01 3.63558859e-01 2.71012813e-01 3.87849361e-01 -1.20769143e+00 -8.48276556e-01 -6.95528984e-01 -2.07369879e-01 4.02281016e-01 4.66822982e-01 -7.33246326e-01 -5.95616885e-02 1.25506982e-01]
[8.526097297668457, -3.0632779598236084]
4bf4659b-4abf-4f4a-85f1-04e6daa79ecc
semi-supervised-temporal-action-detection
2207.07059
null
https://arxiv.org/abs/2207.07059v1
https://arxiv.org/pdf/2207.07059v1.pdf
Semi-Supervised Temporal Action Detection with Proposal-Free Masking
Existing temporal action detection (TAD) methods rely on a large number of training data with segment-level annotations. Collecting and annotating such a training set is thus highly expensive and unscalable. Semi-supervised TAD (SS-TAD) alleviates this problem by leveraging unlabeled videos freely available at scale. However, SS-TAD is also a much more challenging problem than supervised TAD, and consequently much under-studied. Prior SS-TAD methods directly combine an existing proposal-based TAD method and a SSL method. Due to their sequential localization (e.g, proposal generation) and classification design, they are prone to proposal error propagation. To overcome this limitation, in this work we propose a novel Semi-supervised Temporal action detection model based on PropOsal-free Temporal mask (SPOT) with a parallel localization (mask generation) and classification architecture. Such a novel design effectively eliminates the dependence between localization and classification by cutting off the route for error propagation in-between. We further introduce an interaction mechanism between classification and localization for prediction refinement, and a new pretext task for self-supervised model pre-training. Extensive experiments on two standard benchmarks show that our SPOT outperforms state-of-the-art alternatives, often by a large margin. The PyTorch implementation of SPOT is available at https://github.com/sauradip/SPOT
['Tao Xiang', 'Yi-Zhe Song', 'Xiatian Zhu', 'Sauradip Nag']
2022-07-14
null
null
null
null
['classification']
['methodology']
[ 3.43930691e-01 -1.51821971e-01 -5.61723471e-01 -3.49921167e-01 -9.69308555e-01 -4.88340199e-01 5.83075106e-01 -1.26146600e-02 -3.79532009e-01 5.53391516e-01 8.72656927e-02 -7.69917965e-02 2.89555281e-01 -4.21833724e-01 -5.18818498e-01 -7.09452748e-01 -1.16791189e-01 3.52774441e-01 1.01314390e+00 6.89860731e-02 1.50956456e-02 1.61540404e-01 -1.47546446e+00 4.88653094e-01 7.35555053e-01 1.12173843e+00 2.65485317e-01 3.76502991e-01 8.99144933e-02 8.71418178e-01 -4.37272072e-01 2.03870758e-02 4.51639622e-01 -6.42597198e-01 -8.69225204e-01 3.25238794e-01 4.02167499e-01 -3.74160498e-01 -1.10935792e-01 6.27219021e-01 3.71659875e-01 3.04052532e-01 4.01791483e-01 -1.25958312e+00 -5.15073016e-02 5.53118289e-01 -7.32178271e-01 2.49934986e-01 3.26676279e-01 3.50415945e-01 9.57164407e-01 -1.05819941e+00 6.51892900e-01 9.57522511e-01 9.15123284e-01 7.16044903e-01 -1.53515065e+00 -4.67469335e-01 3.69591415e-01 2.71726966e-01 -1.53818262e+00 -5.69166362e-01 7.85342455e-01 -4.99463886e-01 1.01950347e+00 1.85927033e-01 7.50341535e-01 1.41758144e+00 -9.13186967e-02 1.22047913e+00 1.20617807e+00 -3.87674183e-01 5.30895948e-01 -6.68850988e-02 6.80592954e-02 7.61213303e-01 -1.46776810e-01 6.52761851e-03 -6.47192240e-01 6.84027672e-02 8.57703507e-01 9.14138183e-03 -1.72694728e-01 -6.75520599e-01 -1.27034199e+00 4.83925283e-01 2.48215958e-01 3.41840208e-01 -2.43993282e-01 -1.44099584e-02 6.25493348e-01 1.93861887e-01 4.73614484e-01 1.33934200e-01 -6.42835796e-01 -3.26216191e-01 -1.24545515e+00 8.23734924e-02 4.49223727e-01 8.58013451e-01 6.42000318e-01 -2.46222347e-01 -5.15203714e-01 7.12596834e-01 1.03842534e-01 8.72353464e-02 6.74376011e-01 -1.02158666e+00 3.83133352e-01 7.36561835e-01 1.18049219e-01 -5.93623817e-01 -4.86497253e-01 -3.85635525e-01 -7.59075046e-01 4.30390716e-01 5.78525603e-01 2.79860020e-01 -1.06470013e+00 1.77938461e+00 6.50288343e-01 4.27499205e-01 -2.96019047e-01 9.95232165e-01 4.11187917e-01 3.80120039e-01 1.50421739e-01 -3.18698198e-01 1.00928128e+00 -1.57833540e+00 -6.11179054e-01 -2.23728970e-01 1.00787485e+00 -5.22842884e-01 1.15758753e+00 4.04616565e-01 -8.43310833e-01 -6.79576516e-01 -8.17842722e-01 -4.57948297e-02 -1.80958435e-01 4.90730882e-01 6.08911276e-01 3.33903134e-01 -8.43721807e-01 7.11517334e-01 -1.32526004e+00 -6.09401107e-01 7.14541674e-01 2.18648031e-01 -6.07685804e-01 8.01386163e-02 -8.67131352e-01 6.97254896e-01 4.34090585e-01 1.31536171e-01 -8.73844624e-01 -4.51494038e-01 -8.43643069e-01 -3.50506485e-01 9.19982910e-01 -4.46692467e-01 1.60693300e+00 -1.11393118e+00 -1.68492448e+00 7.20042288e-01 -1.86488360e-01 -5.80879033e-01 8.33635747e-01 -2.96539873e-01 -2.36395612e-01 1.81841165e-01 3.69119316e-01 7.46861219e-01 1.05260181e+00 -9.27664101e-01 -6.86462402e-01 -8.85447338e-02 -4.82808193e-03 1.59642860e-01 -1.73662454e-01 -1.79004371e-02 -7.29201674e-01 -9.02310312e-01 9.19107944e-02 -1.02467370e+00 -3.30659360e-01 3.09134126e-01 -2.85103500e-01 -3.82837147e-01 9.38470542e-01 -5.08522332e-01 1.55548739e+00 -2.17849827e+00 1.46497622e-01 -8.10303017e-02 2.08099976e-01 5.08932352e-01 -1.63984850e-01 2.47062087e-01 8.40528961e-03 -8.32237974e-02 -3.83542001e-01 -7.85715103e-01 -7.44324923e-02 2.92869180e-01 2.81339791e-02 3.81423116e-01 2.65919387e-01 7.13896453e-01 -1.15210760e+00 -8.37520123e-01 3.70434076e-01 2.33617708e-01 -5.31052947e-01 1.27371445e-01 -3.55400026e-01 6.69354200e-01 -4.79740947e-01 8.55118871e-01 3.45095068e-01 -5.01812577e-01 2.15121403e-01 -2.29549333e-01 -3.15395087e-01 5.75043559e-01 -1.31881261e+00 2.10802770e+00 -1.91688195e-01 3.62205595e-01 -2.39476889e-01 -1.06616378e+00 5.82040131e-01 3.46101463e-01 7.75204480e-01 -5.70556343e-01 7.30952695e-02 2.51249522e-01 -2.77488172e-01 -3.11750442e-01 2.92388231e-01 2.11804166e-01 1.12584447e-02 6.62975252e-01 2.08228022e-01 3.34511220e-01 4.48957026e-01 3.24484646e-01 1.51419020e+00 9.38427329e-01 4.00135368e-01 7.97604099e-02 4.08007890e-01 1.60535678e-01 1.11721170e+00 7.61363328e-01 -6.78200662e-01 7.05655515e-01 3.32582116e-01 -4.47144657e-01 -7.39799738e-01 -7.20531881e-01 -3.72047983e-02 1.04234469e+00 1.68917924e-01 -8.40195298e-01 -7.85942495e-01 -1.37072408e+00 -1.82823107e-01 4.39587474e-01 -8.17611217e-01 -2.53509581e-02 -7.42217004e-01 -5.24963796e-01 3.69459838e-01 7.27202058e-01 5.65860152e-01 -1.00728667e+00 -7.22770810e-01 3.30233842e-01 -4.05352771e-01 -1.26176608e+00 -6.14837945e-01 3.72480005e-01 -9.64590371e-01 -9.44768786e-01 -6.55934989e-01 -4.71565664e-01 6.94797933e-01 5.03207862e-01 9.87529039e-01 -1.20211855e-01 -9.08768848e-02 2.33253956e-01 -7.83780575e-01 -1.17005467e-01 -3.74789745e-01 1.45351574e-01 1.29657328e-01 3.30765814e-01 3.78513128e-01 -6.07321203e-01 -7.35375226e-01 8.95730793e-01 -6.92003429e-01 5.10390520e-01 7.11874962e-01 6.69416428e-01 1.02266896e+00 -1.35511477e-02 2.95428753e-01 -6.42702579e-01 -1.10673249e-01 -3.55725735e-01 -5.40002286e-01 9.35952216e-02 -6.17902815e-01 -4.41629514e-02 3.39926511e-01 -7.14000046e-01 -1.11285245e+00 5.92561424e-01 8.70990530e-02 -7.96106875e-01 -4.30909842e-01 3.49546939e-01 -3.89666297e-02 -2.77079660e-02 7.00088620e-01 1.49964884e-01 8.70266855e-02 -8.22664499e-01 1.08801700e-01 2.64395118e-01 3.37480247e-01 -2.60287285e-01 7.89928198e-01 6.88825607e-01 -2.06744432e-01 -4.80577230e-01 -1.17618680e+00 -6.56038761e-01 -1.07098246e+00 -5.13487041e-01 8.36122453e-01 -9.51445699e-01 -1.60622030e-01 6.43360615e-01 -9.15145099e-01 -9.11422312e-01 -5.72792292e-01 5.28770745e-01 -5.89141905e-01 4.40376967e-01 -3.97152871e-01 -6.27823770e-01 -1.99152101e-02 -9.17788982e-01 1.29980361e+00 -1.11798950e-01 -4.84141499e-01 -6.55231297e-01 2.16131493e-01 5.94926119e-01 2.01012149e-01 2.77254909e-01 1.23171017e-01 -5.93174338e-01 -6.67885721e-01 -2.09602028e-01 -6.00480884e-02 3.83757532e-01 1.61860779e-01 -2.22361341e-01 -8.20018947e-01 -1.85597047e-01 -1.75472185e-01 -5.43879926e-01 9.96320486e-01 3.69230419e-01 1.08021259e+00 -8.10272023e-02 -4.29412425e-01 2.87515849e-01 1.05417049e+00 -1.63717881e-01 4.68336731e-01 3.86905849e-01 7.00621903e-01 4.83931750e-01 1.21273243e+00 5.28707922e-01 3.25745493e-01 1.32130885e+00 2.16226682e-01 -1.01033524e-01 -4.67554122e-01 -2.54010081e-01 7.22008407e-01 6.56018436e-01 -3.34335804e-01 -2.16429323e-01 -8.25881839e-01 4.70081061e-01 -2.36615920e+00 -1.09584665e+00 -2.76603460e-01 2.12286997e+00 1.09191680e+00 3.11788529e-01 5.63701034e-01 5.36627650e-01 5.03186643e-01 1.50434330e-01 -3.63772362e-01 2.44748995e-01 -8.77544080e-05 3.95478942e-02 3.64124298e-01 2.41052389e-01 -1.59651887e+00 1.11678958e+00 5.01426315e+00 1.15192473e+00 -8.90309215e-01 4.36028868e-01 2.94851899e-01 -1.39419153e-01 4.22155976e-01 1.70223653e-01 -8.59846354e-01 5.27880847e-01 6.92312419e-01 3.25562984e-01 1.01042669e-02 8.45375359e-01 6.24451458e-01 -5.01423180e-01 -1.25354791e+00 9.52202141e-01 -1.60443351e-01 -1.28305137e+00 -1.91734463e-01 -2.04186991e-01 5.80797732e-01 7.88825899e-02 -3.44001949e-01 3.37809324e-01 5.40706255e-02 -4.88600403e-01 1.00502586e+00 3.32153350e-01 7.03384936e-01 -3.37942392e-02 5.75922012e-01 4.37788367e-01 -1.57286215e+00 -1.09693542e-01 1.51747093e-01 -8.20040554e-02 3.84443998e-01 5.96041739e-01 -5.74790418e-01 4.45774168e-01 8.54064405e-01 1.21150887e+00 -6.18991494e-01 1.08502328e+00 -4.39659625e-01 8.56367946e-01 -3.74412537e-01 2.97517926e-01 3.48277450e-01 1.60882160e-01 6.65227413e-01 1.29507494e+00 6.04317635e-02 1.82418395e-02 5.88730574e-01 7.25471914e-01 2.12180018e-01 -3.48637812e-02 -2.21970439e-01 1.13428280e-01 2.39675224e-01 1.18676138e+00 -9.99221027e-01 -3.57855290e-01 -4.89028186e-01 1.23002970e+00 2.46043861e-01 1.91087008e-01 -1.17467225e+00 1.07968077e-01 3.28599840e-01 3.09606045e-01 5.19504011e-01 -3.41735303e-01 -4.67371754e-02 -1.26067042e+00 2.26629257e-01 -7.93607712e-01 5.96000552e-01 -5.92309237e-01 -1.02548063e+00 3.33623439e-01 3.63852791e-02 -1.78175414e+00 5.91141060e-02 -1.85936972e-01 -3.37108254e-01 1.83861241e-01 -1.54705381e+00 -1.26783288e+00 -3.33092719e-01 7.68940330e-01 9.48671341e-01 1.48825422e-01 7.34418511e-01 5.58495045e-01 -9.69635010e-01 5.70713937e-01 -3.87234837e-01 1.19649716e-01 8.99478018e-01 -1.19768322e+00 1.99080870e-01 9.09586728e-01 2.54427850e-01 1.47039577e-01 3.93605083e-01 -6.03814006e-01 -1.02429020e+00 -1.50874352e+00 8.75699461e-01 -6.38420105e-01 6.82905018e-01 -4.67392564e-01 -9.49778795e-01 6.80023730e-01 -3.24425280e-01 3.40824068e-01 5.44001281e-01 -1.33880183e-01 -3.67170244e-01 -1.77386492e-01 -8.00758123e-01 4.65920627e-01 1.49913406e+00 -3.62846315e-01 -4.38515186e-01 5.50488710e-01 6.22388005e-01 -4.08794165e-01 -6.26847208e-01 4.49787557e-01 5.08562922e-01 -1.24445999e+00 6.92294776e-01 -1.70688778e-01 2.34891281e-01 -6.79740489e-01 1.59676149e-01 -7.73594737e-01 -3.78326535e-01 -9.89840031e-01 -5.25366902e-01 1.20021987e+00 3.59321058e-01 -2.69576341e-01 8.24031353e-01 2.55855471e-01 -2.87002444e-01 -8.25475991e-01 -1.03033018e+00 -1.19775760e+00 -5.75280964e-01 -5.66365600e-01 8.02722350e-02 9.33149934e-01 -9.37560201e-02 2.69898951e-01 -4.64518696e-01 9.59523991e-02 6.29202962e-01 1.39961187e-02 9.36425865e-01 -9.98318315e-01 -4.43729997e-01 -3.87600958e-01 -5.48458338e-01 -1.32504332e+00 7.99400359e-03 -6.60846531e-01 2.49777585e-01 -1.32245088e+00 1.55320540e-01 -7.23824203e-01 -4.24849331e-01 1.08700097e+00 6.30743504e-02 5.09944439e-01 -3.41962278e-02 6.01181030e-01 -1.25444412e+00 5.81081688e-01 1.08102787e+00 1.08245388e-01 -5.90646148e-01 1.73981249e-01 -1.01280445e-02 9.49120402e-01 8.69197845e-01 -7.23484099e-01 -4.84490842e-01 -2.93285787e-01 -2.37722531e-01 -1.37554303e-01 6.45494044e-01 -1.21753097e+00 1.81960180e-01 -2.27605745e-01 -5.80180809e-02 -6.59870386e-01 3.34706813e-01 -7.38136053e-01 -4.44030464e-02 3.06139261e-01 -3.61529440e-01 -4.17226523e-01 1.80441160e-02 8.02579403e-01 -2.06934214e-01 8.03766623e-02 8.41232598e-01 -5.23822233e-02 -1.06801987e+00 4.18743581e-01 -4.51906025e-01 -7.19888061e-02 1.22917092e+00 -5.00103176e-01 2.45267060e-02 -8.33180547e-02 -8.22485268e-01 2.27348059e-01 3.57462108e-01 3.93187404e-01 3.58467489e-01 -1.24835026e+00 -4.12454009e-01 4.27973233e-02 2.72393346e-01 5.25120944e-02 2.11818963e-01 1.58924806e+00 -3.79658118e-02 1.27196088e-01 8.67689960e-03 -7.54672706e-01 -1.39866531e+00 3.22279334e-01 2.51286268e-01 -5.52129745e-01 -9.72465932e-01 8.39278460e-01 9.47446898e-02 -2.04953581e-01 5.35901904e-01 -3.98672193e-01 -7.36146234e-03 -7.18045887e-03 4.73040313e-01 3.56937677e-01 -5.50419511e-03 -6.37231410e-01 -4.14021045e-01 4.57396924e-01 1.65892895e-02 -4.67666909e-02 1.31001186e+00 -2.12395743e-01 1.59311920e-01 5.82225561e-01 7.85575628e-01 -2.25335762e-01 -1.59735954e+00 -4.73943263e-01 1.55951425e-01 -4.97926295e-01 5.45451678e-02 -7.81161666e-01 -9.33731496e-01 6.85859263e-01 4.87079769e-01 1.09612145e-01 1.24619436e+00 9.32750702e-02 9.05802548e-01 2.29062170e-01 4.88524914e-01 -1.36322391e+00 5.25000215e-01 4.07229334e-01 7.57170796e-01 -1.39839578e+00 1.30498141e-01 -5.08599937e-01 -6.09310448e-01 8.63472939e-01 8.50160241e-01 1.13527045e-01 5.41572392e-01 1.64683029e-01 -8.88955127e-03 -4.38925289e-02 -8.09814036e-01 -3.79921079e-01 3.01238954e-01 3.76645863e-01 2.05060035e-01 -1.88414320e-01 -3.60741109e-01 6.62238002e-01 5.66149712e-01 3.34131837e-01 2.61044145e-01 1.43643427e+00 -2.72529125e-01 -1.25837421e+00 -1.33574739e-01 2.21995980e-01 -2.22264662e-01 1.47524416e-01 -3.28924537e-01 6.09084845e-01 4.03547525e-01 7.98218012e-01 -1.59473121e-01 -4.84317392e-01 2.13293448e-01 1.25417367e-01 3.37899178e-01 -7.30896533e-01 -3.52783740e-01 3.92232001e-01 2.13704735e-01 -1.14004576e+00 -1.03982270e+00 -9.28508162e-01 -1.23929024e+00 1.25803962e-01 -5.93234539e-01 -8.55530873e-02 2.29193076e-01 9.13821697e-01 5.99009573e-01 4.19351995e-01 5.64812541e-01 -1.10659587e+00 -4.21387553e-01 -8.23782504e-01 -4.70594317e-01 3.32597047e-01 2.01198995e-01 -9.68431473e-01 -3.28075230e-01 3.70222598e-01]
[8.429756164550781, 0.5927621722221375]
ed37d7a9-cae4-46a9-b400-de6e20706899
electrode-clustering-and-bandpass-analysis-of
2302.12710
null
https://arxiv.org/abs/2302.12710v1
https://arxiv.org/pdf/2302.12710v1.pdf
Electrode Clustering and Bandpass Analysis of EEG Data for Gaze Estimation
In this study, we validate the findings of previously published papers, showing the feasibility of an Electroencephalography (EEG) based gaze estimation. Moreover, we extend previous research by demonstrating that with only a slight drop in model performance, we can significantly reduce the number of electrodes, indicating that a high-density, expensive EEG cap is not necessary for the purposes of EEG-based eye tracking. Using data-driven approaches, we establish which electrode clusters impact gaze estimation and how the different types of EEG data preprocessing affect the models' performance. Finally, we also inspect which recorded frequencies are most important for the defined tasks.
['Roger Wattenhofer', 'Nicolas Langer', 'Joël Küchler', 'Martyna Beata Plomecka', 'Ard Kastrati']
2023-02-19
null
null
null
null
['gaze-estimation']
['computer-vision']
[-1.23049663e-02 4.25884388e-02 3.86695087e-01 -1.82190001e-01 -1.50066197e-01 -3.61800909e-01 1.52737126e-01 3.19525987e-01 -7.33791828e-01 8.48958433e-01 -1.16404193e-02 -3.74824464e-01 -4.18304801e-01 -1.68950796e-01 -7.69607663e-01 -4.30613577e-01 1.91582646e-02 -2.61496276e-01 -5.98835014e-03 1.06533863e-01 7.54047513e-01 3.90145689e-01 -1.90352488e+00 5.68878371e-03 8.67260218e-01 9.61560726e-01 1.86935335e-01 2.30087891e-01 5.47287166e-01 2.79456735e-01 -1.00970864e+00 -2.23048255e-01 -2.15709299e-01 -4.69455272e-01 -3.52709532e-01 -3.88647199e-01 2.89949983e-01 -3.10444087e-01 1.90150458e-02 7.83759296e-01 7.33154535e-01 1.26333281e-01 5.85318983e-01 -1.32876980e+00 -6.54726684e-01 4.04612064e-01 -5.80004990e-01 8.01652908e-01 6.53338253e-01 1.87080845e-01 5.16475201e-01 -6.27573609e-01 5.39856553e-01 4.03894037e-01 6.33274019e-01 3.00758839e-01 -1.18801224e+00 -1.12221837e+00 2.37574235e-01 4.03085560e-01 -1.72075617e+00 -7.47712374e-01 7.32915461e-01 -4.64373618e-01 1.11289346e+00 3.89305294e-01 1.05849600e+00 1.23969531e+00 3.47373098e-01 -1.08252920e-01 1.49949098e+00 -3.39999646e-01 3.07474434e-01 5.52077532e-01 4.62046057e-01 -1.25802616e-02 7.00980008e-01 -1.11180887e-01 -9.23246384e-01 -7.30611756e-02 5.48709691e-01 -4.35994923e-01 -7.30266809e-01 3.57856676e-02 -1.02239025e+00 4.44514930e-01 1.25458911e-01 6.06118262e-01 -4.57567692e-01 -4.83540539e-03 9.58309025e-02 2.72293594e-02 4.82978076e-01 9.31061208e-01 -8.64706635e-02 -5.23545504e-01 -1.15267527e+00 -1.59336150e-01 5.37867844e-01 6.53591037e-01 2.43398041e-01 -1.49647817e-01 -2.68016696e-01 4.05042797e-01 3.75225216e-01 3.30324084e-01 2.55985379e-01 -7.49938488e-01 2.09370285e-01 4.06551629e-01 3.06037009e-01 -9.81877565e-01 -7.17028618e-01 -3.29578191e-01 4.45428528e-02 2.07447842e-01 3.79356831e-01 -4.47427690e-01 -5.01973450e-01 1.54001558e+00 -2.58627445e-01 1.48529783e-01 -6.08935893e-01 9.11462784e-01 4.26691860e-01 7.77947977e-02 3.49130839e-01 -5.68320036e-01 1.53225362e+00 -1.11195669e-01 -9.82791245e-01 -2.78782785e-01 4.69012141e-01 -6.30145967e-01 1.29453111e+00 7.18915164e-01 -1.10220838e+00 -4.32087272e-01 -1.11171699e+00 2.46973544e-01 -2.76659936e-01 -1.26774367e-02 6.59192562e-01 1.14538944e+00 -1.27169287e+00 3.05977166e-01 -7.14849830e-01 -3.94984215e-01 5.20232856e-01 6.06198788e-01 -2.32156128e-01 4.72848833e-01 -9.54650998e-01 1.42046368e+00 1.26801774e-01 -8.08987767e-03 -3.32555294e-01 -7.17916489e-01 -4.31082696e-01 2.54911512e-01 -1.61936030e-01 -2.94974238e-01 6.93736553e-01 -9.13970947e-01 -1.18595755e+00 6.72475457e-01 -4.12746906e-01 -1.04299515e-01 2.05567703e-02 2.27561183e-02 -5.42571306e-01 2.73240477e-01 -3.17569077e-01 7.22462833e-01 6.07416570e-01 -1.02648771e+00 -5.96180439e-01 -4.87196952e-01 -1.00418873e-01 4.38834988e-02 -5.96072853e-01 4.15129215e-01 -2.23186553e-01 -2.36885130e-01 -4.09471631e-01 -7.57756114e-01 4.08684850e-01 -3.99233460e-01 -9.61191766e-03 -3.65634084e-01 2.68496782e-01 -7.29287624e-01 1.55023420e+00 -2.37953520e+00 -1.04454808e-01 5.05184829e-01 4.44243968e-01 -1.24908634e-01 2.11650237e-01 1.04331084e-01 -2.49237731e-01 3.85706007e-01 3.00494768e-02 -3.14647168e-01 1.96214169e-01 -5.32883704e-01 6.54511303e-02 6.46982312e-01 4.63283248e-02 8.01830947e-01 -4.69734430e-01 -2.29636058e-01 2.73887157e-01 7.93497980e-01 -4.57243830e-01 -1.44946679e-01 6.04392529e-01 5.85413933e-01 2.79981673e-01 3.33782315e-01 6.17331326e-01 -1.37762204e-01 -1.99061055e-02 -1.35598883e-01 -5.45218050e-01 5.21549582e-01 -7.33957171e-01 1.26001120e+00 -1.52271599e-01 1.44887865e+00 -2.26972327e-01 -3.54841739e-01 5.77518165e-01 2.43972123e-01 3.09269756e-01 -9.67611015e-01 5.08081317e-01 -2.15774905e-02 6.21845961e-01 -4.52418536e-01 3.55867475e-01 3.18346083e-01 5.36544561e-01 7.12765455e-01 -1.29589215e-01 6.97388798e-02 2.20457196e-01 -2.03141272e-01 7.83571959e-01 -1.15017690e-01 -2.88353041e-02 -6.75785661e-01 -2.60402113e-01 -2.73448139e-01 7.18271658e-02 5.37580073e-01 -2.45444030e-01 5.81395328e-01 6.90473557e-01 3.62087935e-01 -5.10736644e-01 -5.87300956e-01 -7.76799560e-01 9.39504862e-01 1.47728965e-01 -5.16833305e-01 -1.28033900e+00 -1.34313792e-01 -2.11383358e-01 1.18267334e+00 -9.54870343e-01 -2.43379727e-01 1.35302097e-01 -9.50768828e-01 2.97711253e-01 4.06988382e-01 -2.14819342e-01 -1.02683687e+00 -1.23137927e+00 -2.28712589e-01 -4.03730897e-03 -8.72327209e-01 -2.82175183e-01 2.19071254e-01 -7.78405368e-01 -1.11271584e+00 -6.43365741e-01 -4.48180556e-01 9.06251073e-01 2.62657583e-01 6.92829847e-01 1.46175787e-01 9.18738991e-02 5.37150741e-01 -3.00222158e-01 -9.77254093e-01 3.01638842e-01 1.32721469e-01 8.74724463e-02 -1.50795564e-01 1.13721633e+00 -5.69740593e-01 -6.81943417e-01 1.10336505e-01 -2.90542573e-01 -1.78670540e-01 3.70412171e-01 1.41262695e-01 1.21014282e-01 -1.57410337e-03 6.21013939e-01 -3.95486474e-01 1.18524516e+00 -5.17892718e-01 -5.35733163e-01 1.08456403e-01 -1.00074422e+00 -2.63377160e-01 -6.53941631e-02 -5.93043923e-01 -7.93316305e-01 -4.77149725e-01 2.17882171e-01 -1.52203262e-01 -4.04574454e-01 3.56533557e-01 3.76430333e-01 -4.70625669e-01 8.04263771e-01 -1.13635167e-01 -2.56050587e-01 -7.09322542e-02 -5.20406842e-01 8.70593011e-01 -5.76817840e-02 1.44757017e-01 2.21547633e-01 2.86446035e-01 -3.76540989e-01 -7.55975187e-01 -2.08331019e-01 -1.20984592e-01 -5.93037903e-01 -4.88145769e-01 8.57472599e-01 -8.88641000e-01 -1.26742744e+00 2.28699282e-01 -9.17326391e-01 -3.72725517e-01 2.48414293e-01 1.07857668e+00 -2.26082236e-01 -1.17994107e-01 -5.92157543e-02 -1.09400249e+00 -2.15573877e-01 -1.19825006e+00 7.89198101e-01 3.10383260e-01 -9.32280719e-01 -7.94829488e-01 -5.43379672e-02 1.75484136e-01 3.15108508e-01 3.14057693e-02 5.42933285e-01 -5.24475694e-01 -1.53841197e-01 -1.35742858e-01 -1.63117737e-01 -1.84576869e-01 1.15480669e-01 1.49668336e-01 -1.35156691e+00 -1.47469267e-01 2.57227391e-01 1.43337816e-01 2.67097741e-01 7.69819140e-01 1.09997833e+00 3.28725904e-01 -4.08919841e-01 6.03744864e-01 1.04695356e+00 3.44043195e-01 8.25996399e-01 3.81229192e-01 3.53962213e-01 7.01509058e-01 3.09177339e-01 3.56205940e-01 4.47869897e-01 4.86605436e-01 2.39779174e-01 -1.73573475e-02 1.33804947e-01 -6.45623431e-02 2.44999930e-01 5.10152996e-01 -4.20957774e-01 -7.45602027e-02 -9.33416128e-01 4.99165416e-01 -1.26294923e+00 -5.05085886e-01 -3.77863348e-01 2.39858150e+00 5.38621724e-01 2.41175562e-01 2.64570206e-01 2.08690584e-01 5.03865540e-01 -3.86822462e-01 -4.51090276e-01 -3.56134027e-01 1.12423718e-01 5.43508768e-01 4.75967377e-01 1.63740497e-02 -3.69711488e-01 1.86419159e-01 8.18323231e+00 1.37021825e-01 -1.17648613e+00 6.27814978e-02 2.70061910e-01 -7.64603019e-01 -2.12026253e-01 -3.00046772e-01 -4.92659956e-01 1.03204226e+00 1.41358566e+00 -2.25525454e-01 6.49795413e-01 3.52151245e-01 6.45168900e-01 -8.65482748e-01 -1.16514468e+00 1.14864564e+00 3.64438504e-01 -9.10096705e-01 -6.37785137e-01 2.69658804e-01 9.94099900e-02 -9.23283100e-02 1.53985202e-01 -6.17151074e-02 -4.90284979e-01 -1.11623919e+00 8.97271752e-01 6.81006074e-01 8.33027124e-01 -5.90754747e-01 6.26054943e-01 6.29541406e-04 -5.85084796e-01 -1.60547346e-01 -1.78732097e-01 -4.09056038e-01 -6.79231882e-02 1.54066563e-01 -4.49131042e-01 -1.93727836e-02 1.17400277e+00 3.75190169e-01 -9.63984847e-01 1.28989947e+00 -2.32382819e-01 8.94318461e-01 -3.42193842e-01 -3.05213630e-01 -3.25794071e-01 2.84543298e-02 2.21106231e-01 8.83177042e-01 5.38643658e-01 2.03868315e-01 -9.21130419e-01 1.17751181e+00 2.32345790e-01 -2.53588869e-03 -4.33924139e-01 9.28409249e-02 8.02248776e-01 1.02392650e+00 -9.35997069e-01 2.35164791e-01 -5.95306575e-01 7.12516665e-01 2.42205113e-01 6.40159547e-01 -8.95378828e-01 -6.11601710e-01 3.97893578e-01 2.87768543e-01 -4.96614259e-03 -1.13182999e-01 -7.94333875e-01 -8.18954825e-01 7.42756948e-02 -6.13421738e-01 -1.78981140e-01 -1.37791800e+00 -6.96323931e-01 4.13698673e-01 3.67676988e-02 -8.46321642e-01 5.30274911e-03 -4.72962350e-01 -4.70395714e-01 1.23937917e+00 -1.29502499e+00 -4.87316221e-01 -4.94839609e-01 5.79839706e-01 -4.07388108e-03 3.15069616e-01 7.28932858e-01 2.38797247e-01 -7.96682715e-01 6.42512798e-01 -1.50992095e-01 -2.66357929e-01 9.05246854e-01 -8.90463889e-01 1.87210172e-01 6.70280159e-01 5.38906157e-02 1.02250898e+00 7.48746693e-01 -4.89165992e-01 -1.21890545e+00 -2.46421218e-01 7.74479151e-01 -7.91115165e-01 3.55898321e-01 -7.31151342e-01 -7.46262312e-01 6.88809693e-01 6.33367896e-01 -6.58478141e-01 9.24202979e-01 5.12577713e-01 2.74301827e-01 8.97104219e-02 -1.02106464e+00 5.93864620e-01 1.01590812e+00 -5.78238487e-01 -7.67982185e-01 -3.64367738e-02 1.52134866e-01 -2.52495021e-01 -9.10878539e-01 6.12735860e-02 7.39832878e-01 -1.09340346e+00 5.57674825e-01 -2.02946693e-01 -7.75921270e-02 1.47566617e-01 4.90864366e-01 -1.41152692e+00 -6.31573975e-01 -4.17554438e-01 2.45649717e-03 1.05556726e+00 5.97663343e-01 -8.55136096e-01 3.88094604e-01 1.19477320e+00 9.94370133e-02 -3.71415943e-01 -7.05746233e-01 -4.52530265e-01 -1.63894415e-01 -4.64119136e-01 5.52235842e-01 7.91759074e-01 6.55860603e-01 9.79463905e-02 2.77135894e-02 2.20236331e-01 1.26999274e-01 -5.08910418e-01 2.50239164e-01 -1.44477558e+00 2.11150050e-01 -5.40002465e-01 -4.93465811e-01 -3.27301830e-01 4.83864881e-02 -3.45388085e-01 -2.02234924e-01 -1.30748188e+00 1.65850207e-01 -1.01776615e-01 -6.91067398e-01 3.36644053e-01 -3.74464840e-01 4.47685361e-01 2.16834769e-01 -1.65575594e-02 -3.07542980e-01 1.53686740e-02 5.59949934e-01 3.90691817e-01 -2.29085937e-01 -1.55408651e-01 -1.18150365e+00 3.93464625e-01 8.63623857e-01 -6.20366514e-01 -6.05838239e-01 -4.87599105e-01 4.74692345e-01 -4.95563686e-01 2.97032773e-01 -1.17319846e+00 4.87037301e-01 1.94891602e-01 7.15649545e-01 -2.95999438e-01 3.36112738e-01 -8.13983679e-01 1.98556542e-01 2.17631742e-01 -1.76520497e-01 1.84810266e-01 7.98041880e-01 1.27416387e-01 2.41691381e-01 -4.45151716e-01 4.75112408e-01 4.54369277e-01 -2.49457300e-01 -2.65098393e-01 -6.75140023e-01 -1.97407752e-01 1.33724487e+00 -8.00122201e-01 -2.29277387e-01 -2.70565212e-01 -5.11274338e-01 -1.06871963e-01 5.95875621e-01 2.37255722e-01 2.37357616e-01 -8.46873641e-01 -3.59901875e-01 4.37272072e-01 1.69115476e-02 -6.55758321e-01 1.66539565e-01 1.58833373e+00 -1.22248523e-01 5.99248409e-01 -5.45960665e-01 -5.43906629e-01 -1.29090917e+00 4.49092299e-01 2.12987646e-01 6.53667808e-01 -4.17947739e-01 7.97719359e-01 -5.99913672e-03 4.18776780e-01 3.21367711e-01 -4.58122581e-01 -5.85464239e-01 5.51379085e-01 6.73376858e-01 4.96415973e-01 3.96735966e-01 -3.68347883e-01 -6.77532613e-01 3.36352199e-01 2.61546791e-01 -1.93592846e-01 1.38135219e+00 -3.52088749e-01 -5.50161349e-03 7.95817971e-01 6.50516927e-01 1.69586942e-01 -1.34743237e+00 7.71483004e-01 -2.10670993e-01 -4.60875601e-01 4.06135559e-01 -8.31611931e-01 -7.19694912e-01 8.97464693e-01 9.03101683e-01 3.80812109e-01 1.32840824e+00 -2.01441988e-01 2.41586547e-02 -5.02920300e-02 3.79920959e-01 -1.23671091e+00 -6.31535888e-01 3.09465062e-02 7.31096327e-01 -8.18032324e-01 1.01654753e-01 -1.71053037e-01 -7.70389020e-01 7.00856924e-01 5.48694074e-01 9.62471217e-02 7.27476656e-01 4.21696126e-01 -3.37618351e-01 -5.26455104e-01 -7.25855470e-01 -1.80678114e-01 4.91738051e-01 9.23931003e-01 5.50388217e-01 -1.80392325e-01 -5.54184258e-01 8.89243722e-01 -4.57994491e-01 3.94651115e-01 6.45774901e-01 7.78800488e-01 -2.51656502e-01 -4.35884327e-01 -5.89063048e-01 9.17045355e-01 -4.91349548e-01 -3.28337938e-01 -6.29578590e-01 1.12924170e+00 -1.34613201e-01 1.21288967e+00 5.88177085e-01 -5.06006598e-01 5.61826468e-01 3.78085196e-01 7.14901626e-01 -6.45748794e-01 -7.92257071e-01 -1.60028964e-01 3.64888390e-03 -4.79702979e-01 -4.65641916e-01 -8.84731591e-01 -9.55069959e-01 -3.68420482e-01 -7.93284118e-01 9.59734395e-02 8.45064700e-01 9.39164817e-01 8.31275046e-01 7.76328564e-01 1.82955414e-01 -6.49613440e-01 1.68227136e-01 -1.38990510e+00 -6.55543804e-01 1.18465893e-01 8.12525824e-02 -1.00269771e+00 -6.48496926e-01 7.80440792e-02]
[13.231764793395996, 3.255540132522583]
72e165aa-a164-410a-990d-57fc5fb17270
deep-curiosity-loops-in-social-environments
1806.03645
null
http://arxiv.org/abs/1806.03645v1
http://arxiv.org/pdf/1806.03645v1.pdf
Deep Curiosity Loops in Social Environments
Inspired by infants' intrinsic motivation to learn, which values informative sensory channels contingent on their immediate social environment, we developed a deep curiosity loop (DCL) architecture. The DCL is composed of a learner, which attempts to learn a forward model of the agent's state-action transition, and a novel reinforcement-learning (RL) component, namely, an Action-Convolution Deep Q-Network, which uses the learner's prediction error as reward. The environment for our agent is composed of visual social scenes, composed of sitcom video streams, thereby both the learner and the RL are constructed as deep convolutional neural networks. The agent's learner learns to predict the zero-th order of the dynamics of visual scenes, resulting in intrinsic rewards proportional to changes within its social environment. The sources of these socially informative changes within the sitcom are predominantly motions of faces and hands, leading to the unsupervised curiosity-based learning of social interaction features. The face and hand detection is represented by the value function and the social interaction optical-flow is represented by the policy. Our results suggest that face and hand detection are emergent properties of curiosity-based learning embedded in social environments.
['Jonatan Barkan', 'Goren Gordon']
2018-06-10
null
null
null
null
['hand-detection']
['computer-vision']
[-1.60769835e-01 5.84425032e-01 2.31434837e-01 -3.57754439e-01 3.53597581e-01 -1.29155010e-01 6.02729201e-01 -5.04491106e-02 -5.45996249e-01 6.31700337e-01 2.16085374e-01 3.84665400e-01 -9.45289806e-03 -8.29440892e-01 -1.19543731e+00 -9.74316001e-01 -5.20583808e-01 -8.08289126e-02 2.29014009e-01 -2.11489096e-01 2.00777784e-01 3.62477601e-02 -1.96825910e+00 3.08417261e-01 5.82842469e-01 1.00026011e+00 5.33382058e-01 1.31234825e+00 7.82893375e-02 1.81081915e+00 -3.07806656e-02 1.96745666e-03 -3.44326124e-02 -7.63274133e-01 -6.47067487e-01 -1.08875304e-01 -2.11964741e-01 -7.83414483e-01 -3.73563051e-01 9.14063573e-01 3.58227313e-01 4.86092716e-01 6.25455737e-01 -1.36395490e+00 -6.62133098e-01 5.87461770e-01 -1.41423522e-03 2.03338221e-01 5.75288236e-01 5.05271792e-01 1.02781844e+00 -4.80214357e-01 6.51353061e-01 1.53970659e+00 5.42656071e-02 1.21854436e+00 -9.76400137e-01 -3.03994656e-01 3.80290836e-01 3.98300588e-01 -6.03024960e-01 -3.95007372e-01 9.02274430e-01 -7.17230499e-01 8.39332104e-01 -1.44606739e-01 1.37943947e+00 1.48806071e+00 8.88730884e-02 1.10660732e+00 9.53717887e-01 -2.40253091e-01 6.81746364e-01 2.81013906e-01 -3.94810408e-01 7.59119749e-01 -5.16954124e-01 8.14417839e-01 -8.63132536e-01 1.70430854e-01 9.91246462e-01 7.52176940e-02 -1.58975163e-05 -6.46250784e-01 -9.03365195e-01 7.91344404e-01 5.97104967e-01 1.22294158e-01 -5.51959157e-01 5.84949374e-01 3.95735241e-02 7.40306914e-01 3.09084922e-01 3.91779214e-01 -4.52859521e-01 -4.19250250e-01 -1.46414833e-02 2.30757177e-01 6.89965606e-01 6.41779780e-01 9.13358033e-01 -1.17933257e-02 -1.07986808e-01 5.87936759e-01 6.80739582e-01 5.37620008e-01 5.03621936e-01 -1.39558744e+00 -1.12659790e-01 6.51970148e-01 6.04119934e-02 -6.97676182e-01 -5.37091255e-01 -3.03453282e-02 -1.71734467e-01 4.34808999e-01 1.87086031e-01 -8.12513173e-01 -2.51770765e-01 2.35803151e+00 6.04636490e-01 3.08943838e-01 2.97990739e-01 8.61122012e-01 9.32345212e-01 5.41931033e-01 3.66135091e-02 -4.27922696e-01 7.44587839e-01 -7.77512670e-01 -5.42103350e-01 3.31381559e-02 4.32719558e-01 -1.28865153e-01 7.33909070e-01 7.25532621e-02 -1.48579991e+00 -4.99273539e-01 -5.17357707e-01 4.23933953e-01 -1.72827736e-01 -3.90432507e-01 4.14986938e-01 -3.25208642e-02 -1.46836972e+00 8.59444261e-01 -8.19762170e-01 -6.91806495e-01 5.32349527e-01 3.45670462e-01 -1.57268167e-01 5.82954645e-01 -9.75552082e-01 7.39896595e-01 1.73304174e-02 -2.01736778e-01 -1.63397014e+00 -3.69308919e-01 -6.61941588e-01 1.70559555e-01 2.15105563e-01 -5.23999691e-01 1.34830809e+00 -1.77906239e+00 -2.12340879e+00 8.32365036e-01 1.88212290e-01 -2.45850593e-01 4.23889607e-01 -1.49194211e-01 -1.04833930e-03 6.51853502e-01 -1.22600935e-01 1.08324897e+00 1.37922359e+00 -1.26912999e+00 -7.99704432e-01 -2.88072944e-01 3.01408440e-01 6.07077539e-01 -4.69951510e-01 5.84641881e-02 1.65703714e-01 -3.72817069e-01 -3.40177745e-01 -9.42760408e-01 -1.89084679e-01 3.48203957e-01 1.83998138e-01 -4.54856783e-01 5.62673807e-01 -9.08424482e-02 7.16489315e-01 -2.35609770e+00 5.20635784e-01 -9.03428346e-02 3.96358103e-01 -1.15718961e-01 -4.46261883e-01 4.65934753e-01 -5.47602251e-02 -4.97864991e-01 2.43667886e-01 -3.09222043e-01 -2.57036090e-01 2.22133413e-01 -5.24348915e-02 4.86463636e-01 3.78762394e-01 9.45400298e-01 -1.46284556e+00 -3.61579090e-01 1.79039985e-01 5.38341343e-01 -1.01239038e+00 1.16505861e+00 -6.54337168e-01 1.02648234e+00 -6.14145935e-01 1.15088478e-03 1.96317539e-01 -1.51992112e-01 1.83565181e-03 7.79293537e-01 -4.31346178e-01 2.14742631e-01 -6.37050807e-01 1.35843122e+00 -1.47880837e-01 5.97907960e-01 1.21519931e-01 -9.43120062e-01 7.93345273e-01 3.57487768e-01 6.65662587e-01 -8.69096339e-01 4.90137815e-01 -1.81020394e-01 2.83766955e-01 -1.23738408e+00 -1.82246163e-01 7.28721172e-02 5.68238139e-01 8.26541722e-01 3.10384780e-01 -9.54837427e-02 -1.25924677e-01 3.32342088e-01 1.19804275e+00 4.41136032e-01 -1.64276555e-01 -2.42439926e-01 4.11840081e-01 -7.16381013e-01 4.28444684e-01 5.93407631e-01 -3.58628809e-01 1.38031524e-02 9.25437510e-01 -6.20867908e-01 -6.65326178e-01 -1.16441798e+00 3.52650613e-01 1.68888593e+00 5.29215261e-02 1.67732626e-01 -1.14445329e+00 -4.36753750e-01 -1.63282573e-01 3.30938637e-01 -1.13957155e+00 -5.74104726e-01 -5.52356064e-01 -5.60562164e-02 -2.85374731e-01 3.55753124e-01 4.10163194e-01 -2.24882340e+00 -1.39232421e+00 2.96997577e-01 2.33661652e-01 -8.41151237e-01 -3.85417223e-01 1.16743617e-01 -6.48973703e-01 -1.15800118e+00 -5.74724376e-01 -9.50485229e-01 7.35438704e-01 -1.68310314e-01 8.42105806e-01 2.51505286e-01 -3.14198822e-01 1.06227374e+00 -5.64544797e-01 -4.05573189e-01 -5.90955317e-01 -6.00839257e-01 3.76625717e-01 5.32190442e-01 7.19874650e-02 -1.03003824e+00 -1.13891697e+00 -7.78990760e-02 -6.50972605e-01 2.53071245e-02 1.83916977e-03 6.73410952e-01 -1.21790417e-01 -5.95359743e-01 5.78064978e-01 -1.57304361e-01 3.89262885e-01 -6.82065606e-01 -6.90848291e-01 5.13631292e-02 -1.05778746e-01 1.60345614e-01 6.02448046e-01 -7.27938771e-01 -1.12963581e+00 1.62517682e-01 1.39967084e-01 -3.15326899e-01 -2.73463994e-01 -1.78002253e-01 3.95196736e-01 1.55340852e-02 4.52152401e-01 1.90742150e-01 3.06105882e-01 -1.51177526e-01 2.99968570e-01 4.77666348e-01 2.10828155e-01 -6.42272592e-01 4.87155139e-01 3.20216984e-01 -1.27443254e-01 -1.01654327e+00 -7.48753667e-01 2.13970169e-02 -5.47605753e-01 -9.93164718e-01 1.46792400e+00 -7.58765638e-01 -1.68862271e+00 7.93420315e-01 -1.19280493e+00 -9.04767931e-01 -7.10530460e-01 7.59654224e-01 -1.07038510e+00 -2.12814808e-01 -5.73326528e-01 -1.26168406e+00 -8.58023986e-02 -1.00214207e+00 5.92159688e-01 6.73083067e-01 3.13975364e-02 -9.22590256e-01 4.80513811e-01 4.74704633e-04 2.51979619e-01 2.17577532e-01 7.67093241e-01 -4.08889771e-01 -7.93937087e-01 5.33850729e-01 4.69168425e-02 4.46356803e-01 -1.65650651e-01 3.53738099e-01 -1.22057426e+00 -3.77881348e-01 -2.13953145e-02 -7.80804455e-01 4.34667975e-01 5.66808879e-01 9.01425600e-01 -3.85601521e-01 1.75424486e-01 5.07576406e-01 8.33443761e-01 6.42421007e-01 2.06663564e-01 -8.01397040e-02 4.05419677e-01 1.17654955e+00 2.97874898e-01 7.72736788e-01 8.48301411e-01 1.93126708e-01 1.10273969e+00 1.14552580e-01 1.05131634e-01 -5.63064575e-01 8.14309299e-01 7.47676730e-01 -4.50169265e-01 3.07789296e-01 -5.07718086e-01 4.58499849e-01 -2.13491011e+00 -1.23902702e+00 3.75322163e-01 2.03049302e+00 7.65394807e-01 -2.02245802e-01 5.38620651e-01 -3.48682880e-01 3.75976503e-01 1.21544711e-01 -1.03269315e+00 -6.77449167e-01 8.14485103e-02 -7.69025385e-02 -3.01578254e-01 5.15864074e-01 -7.86948204e-01 1.11607337e+00 6.16797113e+00 -1.29451334e-01 -1.03712726e+00 -8.88544470e-02 4.31888044e-01 -3.58525515e-01 -2.12836280e-01 -3.67604643e-01 -3.47134680e-01 5.19941330e-01 7.99732864e-01 3.97427231e-01 1.12852085e+00 6.49756134e-01 3.79922360e-01 -2.46469483e-01 -1.38589716e+00 7.24604130e-01 -1.29365802e-01 -1.01018476e+00 -5.34767568e-01 -4.49377932e-02 6.89391613e-01 2.46457890e-01 2.52079666e-01 3.29345018e-01 7.32229888e-01 -7.43835926e-01 8.73663485e-01 8.65164161e-01 5.30223370e-01 -4.56811249e-01 7.87945464e-03 6.99002981e-01 -8.60158801e-01 -7.45150089e-01 -1.00032605e-01 -5.27360916e-01 -5.29442012e-01 -3.43442321e-01 -6.11073911e-01 -7.29941189e-01 1.13458300e+00 9.70646977e-01 -1.60804138e-01 4.70225006e-01 -3.79603416e-01 4.29681808e-01 -2.27521602e-02 -8.15658391e-01 3.06450546e-01 -3.62410814e-01 7.54449427e-01 6.88004196e-01 -7.21265003e-02 5.44263184e-01 -7.08735287e-02 8.46048653e-01 -1.03403218e-02 5.86292781e-02 -7.47925222e-01 8.45975280e-02 6.08308299e-04 9.34010208e-01 -3.18352908e-01 -1.97629482e-01 -3.43052119e-01 7.53706932e-01 8.17938507e-01 4.58639801e-01 -4.88004446e-01 1.61655977e-01 9.66539443e-01 6.22804649e-03 3.75380874e-01 2.90954411e-01 4.10285294e-01 -1.25533307e+00 -3.37183505e-01 -5.62901139e-01 7.26590902e-02 -7.83747435e-01 -9.39335585e-01 5.08330345e-01 -2.51043439e-01 -8.36655080e-01 -3.70931417e-01 -3.13013166e-01 -9.27977502e-01 1.99599311e-01 -1.45436656e+00 -7.60190308e-01 -1.20092764e-01 1.07653558e+00 4.04928267e-01 -5.10836601e-01 6.37843668e-01 -3.17228794e-01 -5.57614028e-01 4.10542369e-01 3.29849273e-02 8.33626166e-02 1.35570541e-01 -1.35113573e+00 6.55278843e-03 3.09146464e-01 -9.29672942e-02 1.16728321e-01 7.19600976e-01 -3.60146850e-01 -1.35533059e+00 -7.94408143e-01 5.05432010e-01 -3.99614006e-01 8.98094952e-01 -4.47103292e-01 -5.89317262e-01 5.73358595e-01 3.00494343e-01 2.92723686e-01 3.17216665e-01 -3.46830547e-01 -1.04195915e-01 -2.13972136e-01 -1.25803685e+00 6.78927183e-01 1.24463785e+00 -5.38915575e-01 -5.25191605e-01 2.96493530e-01 9.81515944e-01 -6.02757260e-02 -5.31856418e-01 -8.07691589e-02 6.39114022e-01 -1.21433032e+00 6.60538793e-01 -8.55345190e-01 8.32696736e-01 2.64585584e-01 1.93677992e-02 -1.46333528e+00 -2.33813465e-01 -9.01264966e-01 -3.36840123e-01 6.30864024e-01 -1.39760645e-02 -6.72585011e-01 6.04209602e-01 4.90351588e-01 2.14285746e-01 -7.16027558e-01 -1.07087982e+00 -1.88559577e-01 2.27318425e-02 -8.76181051e-02 6.10090017e-01 6.55026138e-01 3.16699773e-01 3.56671393e-01 -3.86338860e-01 -7.64281303e-02 6.86498165e-01 -1.37240961e-01 5.51050007e-01 -1.02964473e+00 -5.17273128e-01 -3.25783581e-01 -9.90041643e-02 -7.97163606e-01 2.39676803e-01 -6.48723960e-01 3.69407624e-01 -9.26860988e-01 1.99664429e-01 -6.51286989e-02 -4.93774951e-01 1.85334608e-01 -1.76182613e-02 -3.29892188e-01 3.86015713e-01 -2.10131958e-01 -1.00560486e+00 8.57771158e-01 1.67201662e+00 3.05931568e-01 -5.59065461e-01 1.64941236e-01 -2.08096668e-01 1.08404469e+00 5.73200643e-01 -4.04326648e-01 -4.29678649e-01 -3.52894038e-01 6.09982550e-01 6.53483152e-01 5.26518822e-01 -8.20668936e-01 1.04592249e-01 -3.54510009e-01 2.91448265e-01 -8.73628855e-02 3.48468661e-01 -8.77987921e-01 -4.82543349e-01 9.82334614e-01 -8.50627363e-01 -2.05128878e-01 -4.32667345e-01 5.65264046e-01 3.05967361e-01 6.30486682e-02 1.14912248e+00 -2.49544308e-01 -3.52655113e-01 5.80449760e-01 -6.96834385e-01 1.50296807e-01 1.38115001e+00 -4.42161970e-03 -9.29185972e-02 -6.02181256e-01 -9.85094428e-01 5.08540988e-01 1.83499351e-01 5.62009573e-01 7.82193303e-01 -1.05077732e+00 -6.22663975e-01 5.59948504e-01 -1.27832979e-01 -3.58817615e-02 2.24909678e-01 5.41363120e-01 -2.65648991e-01 -3.09358448e-01 -4.77989465e-01 -6.57444358e-01 -1.15476322e+00 5.78640997e-01 5.70549250e-01 3.15397412e-01 -5.57550490e-01 1.26881313e+00 5.76472223e-01 -3.51282001e-01 5.17356694e-01 -2.58390188e-01 -8.47091377e-01 1.95368484e-01 5.87267041e-01 4.83388007e-01 -8.54061186e-01 -5.97400904e-01 -1.07541122e-01 4.99725014e-01 2.56823421e-01 -9.02883112e-02 1.80705559e+00 -2.08104447e-01 -3.08486879e-01 6.57965124e-01 1.27031279e+00 -7.20329821e-01 -2.01811457e+00 -2.06215844e-01 -2.38453686e-01 -2.01132223e-01 -2.92029351e-01 -6.06592655e-01 -1.06010079e+00 1.08344734e+00 7.90695131e-01 4.36706156e-01 1.09948862e+00 4.02858108e-01 4.79596406e-01 4.51065272e-01 3.76786083e-01 -1.43306196e+00 1.02829897e+00 6.16393924e-01 1.08659327e+00 -1.58004594e+00 -5.34038424e-01 5.28772473e-01 -8.25697899e-01 9.08965886e-01 7.98446298e-01 -5.01759171e-01 9.55524564e-01 1.39607996e-01 -6.43051276e-03 -2.19698042e-01 -1.38286602e+00 -1.78906724e-01 -8.12700670e-03 7.59689271e-01 -6.34758696e-02 2.58165449e-01 2.37429902e-01 3.28675002e-01 -2.24769726e-01 -6.15951717e-02 2.81851679e-01 6.55268788e-01 -8.24191809e-01 -6.14919305e-01 -3.32393534e-02 3.29102993e-01 -1.16497353e-01 1.88477844e-01 -1.96737379e-01 9.32622850e-02 3.73346895e-01 8.74266922e-01 3.57685953e-01 -3.65327507e-01 1.24973930e-01 -3.18618149e-01 5.70596278e-01 -6.43386602e-01 -8.00851941e-01 -1.08035825e-01 -4.39683139e-01 -1.03099382e+00 -5.84431171e-01 -9.63538766e-01 -1.46500659e+00 -1.41292080e-01 2.33915880e-01 -7.84368664e-02 5.06520629e-01 8.57388318e-01 -6.83227479e-02 3.57497841e-01 1.59371912e+00 -8.93163681e-01 -3.32155585e-01 -7.52219021e-01 -5.05841613e-01 4.36071187e-01 1.05226839e+00 -5.14993668e-01 -6.37707591e-01 5.27934320e-02]
[4.211451530456543, 1.4380632638931274]
cbf81895-9bed-404f-a570-7afde56651cb
expert-concept-modeling-ground-truth
null
null
https://aclanthology.org/2020.coling-main.586
https://aclanthology.org/2020.coling-main.586.pdf
Expert Concept-Modeling Ground Truth Construction for Word Embeddings Evaluation in Concept-Focused Domains
We present a novel, domain expert-controlled, replicable procedure for the construction of concept-modeling ground truths with the aim of evaluating the application of word embeddings. In particular, our method is designed to evaluate the application of word and paragraph embeddings in concept-focused textual domains, where a generic ontology does not provide enough information. We illustrate the procedure, and validate it by describing the construction of an expert ground truth, QuiNE-GT. QuiNE-GT is built to answer research questions concerning the concept of naturalized epistemology in QUINE, a 2-million-token, single-author, 20th-century English philosophy corpus of outstanding quality, cleaned up and enriched for the purpose. To the best of our ken, expert concept-modeling ground truths are extremely rare in current literature, nor has the theoretical methodology behind their construction ever been explicitly conceptualised and properly systematised. Expert-controlled concept-modeling ground truths are however essential to allow proper evaluation of word embeddings techniques, and increase their trustworthiness in specialised domains in which the detection of concepts through their expression in texts is important. We highlight challenges, requirements, and prospects for future work.
['Jelke Bloem', 'Andrew Salway', 'Yvette Oortwijn', 'Thijs Ossenkoppele', 'Martin Reynaert', 'Arianna Betti']
2020-12-01
null
null
null
coling-2020-8
['embeddings-evaluation']
['natural-language-processing']
[ 3.28980014e-02 4.81018633e-01 3.74706425e-02 -9.07808095e-02 -5.26712716e-01 -6.80101156e-01 1.02781928e+00 8.20369363e-01 -8.48013878e-01 3.46893370e-01 6.52771533e-01 -6.23761356e-01 -6.16986811e-01 -6.83330178e-01 -4.26774412e-01 -2.69647330e-01 2.24657148e-01 7.43446589e-01 -1.23124093e-01 -6.36417627e-01 7.27211237e-01 4.73939739e-02 -1.66110408e+00 1.30253181e-01 9.69691455e-01 7.27831185e-01 2.92717189e-01 1.90767035e-01 -4.34228271e-01 7.16469109e-01 -5.87393343e-01 -9.16499972e-01 -3.11598748e-01 -1.50638446e-01 -1.08693683e+00 -5.71946278e-02 2.14950025e-01 1.76737219e-01 -1.45560701e-03 1.05981195e+00 4.90934819e-01 -6.63456926e-03 9.80046630e-01 -1.04547250e+00 -1.25509620e+00 9.07433927e-01 3.56838167e-01 2.66364932e-01 4.08194721e-01 -2.41738528e-01 1.40118337e+00 -1.10312009e+00 9.20814157e-01 1.21164024e+00 6.36045516e-01 5.44772148e-01 -1.00655961e+00 -2.76610732e-01 -2.03823641e-01 3.79604280e-01 -1.30332971e+00 -3.53687018e-01 5.83738089e-01 -8.73731196e-01 8.39762747e-01 -1.30317956e-01 6.35777891e-01 1.50202048e+00 7.59197995e-02 3.06991845e-01 9.82966483e-01 -9.09488022e-01 4.94101971e-01 6.84874952e-01 2.42113650e-01 3.71298969e-01 6.98245704e-01 -7.16799945e-02 -5.00576377e-01 -2.43977830e-01 4.76397693e-01 -4.73489255e-01 -4.26255912e-01 -3.02383900e-01 -1.39939463e+00 1.12964845e+00 -8.24333057e-02 1.01720226e+00 -5.89356959e-01 -1.67876244e-01 6.91428244e-01 1.94819957e-01 5.58104455e-01 9.92971301e-01 -5.76975703e-01 -5.77033281e-01 -8.91751111e-01 4.76731002e-01 7.36153841e-01 9.36103821e-01 2.62989908e-01 -1.21109277e-01 1.37583673e-01 5.93044162e-01 2.73715585e-01 2.97554970e-01 8.72798979e-01 -9.18123782e-01 8.66103023e-02 8.74124527e-01 2.91421115e-01 -1.29609072e+00 -1.66014895e-01 -3.87194484e-01 -2.34258786e-01 -7.61513412e-02 1.23345457e-01 -7.73136795e-04 -3.61553043e-01 1.36039555e+00 2.51700997e-01 -1.37049109e-01 4.80847090e-01 7.47898161e-01 1.08926976e+00 3.07958543e-01 2.08342105e-01 -1.01846806e-03 1.87801290e+00 -1.35578960e-01 -8.83072257e-01 -2.78122932e-01 8.53642166e-01 -5.30003905e-01 1.30768323e+00 4.68055367e-01 -7.04921484e-01 -3.55906367e-01 -1.12288678e+00 -1.18185423e-01 -7.79757798e-01 -1.37891039e-01 4.96188819e-01 8.03782821e-01 -6.73597157e-01 5.76335132e-01 -3.61484259e-01 -5.18527150e-01 3.68665427e-01 -4.15661126e-01 -5.31186640e-01 -8.17876607e-02 -1.62015998e+00 1.68278122e+00 8.85301769e-01 -2.11698502e-01 -4.66162354e-01 -8.88239622e-01 -8.84909451e-01 -1.00870922e-01 4.66685504e-01 -4.86471057e-01 9.71682131e-01 -7.64269054e-01 -8.40833008e-01 1.35201037e+00 9.66640189e-02 -5.66551328e-01 6.97195381e-02 -1.09237954e-01 -9.15132821e-01 1.23496704e-01 4.08716708e-01 1.38124406e-01 6.36246860e-01 -1.35221219e+00 -5.87998211e-01 -2.39139974e-01 2.26236254e-01 -1.33161604e-01 -7.08141327e-01 2.69950144e-02 8.52890238e-02 -6.17801368e-01 -6.52438998e-02 -2.79613197e-01 3.54865789e-02 -2.77838588e-01 2.91146100e-01 -5.45650542e-01 1.29799381e-01 -6.41107142e-01 1.47402525e+00 -2.19625926e+00 9.09241661e-02 -8.38491991e-02 5.28006375e-01 1.84624195e-01 1.55241922e-01 9.91103947e-01 -3.37030478e-02 5.44788599e-01 -3.19035590e-01 2.77671106e-02 5.96198380e-01 3.80218774e-01 -4.96117294e-01 4.19311583e-01 2.93064892e-01 8.25193286e-01 -1.31543100e+00 -7.06893206e-01 2.85433143e-01 5.35501838e-01 -4.43641841e-01 -1.83479175e-01 -1.20175496e-01 -3.74060571e-01 -4.07737672e-01 4.35576737e-01 8.68758187e-02 4.58367504e-02 3.84784669e-01 -2.94846892e-01 -2.66422540e-01 5.62990129e-01 -9.56430435e-01 1.58473647e+00 -6.17081821e-01 1.07963324e+00 -3.92897278e-01 -1.09054291e+00 1.00310099e+00 5.07843435e-01 3.23582590e-01 -7.57538199e-01 3.86093140e-01 4.97785509e-01 -4.19690460e-02 -7.05806375e-01 9.41837907e-01 -7.01528192e-01 -1.97115809e-01 2.83645302e-01 3.14461052e-01 -4.11270618e-01 3.00512344e-01 2.70401955e-01 9.84148800e-01 1.07702635e-01 4.93920892e-01 -7.65360594e-01 4.17514592e-01 4.14658487e-01 2.45401278e-01 2.96697468e-01 -8.54092464e-02 4.45759624e-01 4.59263355e-01 -1.76966593e-01 -1.25872147e+00 -7.29437768e-01 -8.48572552e-01 8.51019502e-01 -8.79699886e-02 -8.28447998e-01 -7.26755321e-01 -3.24454427e-01 1.56519823e-02 1.41732049e+00 -9.56335187e-01 -2.01029368e-02 -4.61652167e-02 -4.55016881e-01 5.39405167e-01 2.82083660e-01 -1.31013647e-01 -1.17644167e+00 -1.08682740e+00 6.24305367e-01 -2.33647883e-01 -1.11200678e+00 3.30241889e-01 1.33612251e-04 -4.32672381e-01 -1.44327331e+00 -3.76122862e-01 -6.52243912e-01 2.24159464e-01 -2.28153452e-01 1.63122976e+00 -5.62556647e-02 -9.95866805e-02 6.98771656e-01 -9.16377425e-01 -6.99810863e-01 -6.70851409e-01 -2.70955205e-01 1.68514788e-01 -3.81478846e-01 1.12130702e+00 -4.49623913e-01 -1.42266363e-01 -1.57808036e-01 -1.34195793e+00 -6.27134502e-01 2.79375643e-01 9.42957938e-01 7.50269741e-02 3.24348450e-01 5.87353885e-01 -9.23345804e-01 1.18724954e+00 -6.69432700e-01 -1.33828327e-01 2.70624161e-01 -7.96717107e-01 4.73768935e-02 2.74874479e-01 -2.88734376e-01 -6.48361146e-01 -9.37079489e-01 -3.03479195e-01 -1.13451360e-02 -5.03193177e-02 1.01755226e+00 3.94462124e-02 4.10299331e-01 1.24451303e+00 -1.06137181e-02 -1.19231492e-01 -3.74446183e-01 7.95627534e-01 8.72146964e-01 6.47255182e-01 -9.87285256e-01 6.19979918e-01 2.51079142e-01 -4.75033462e-01 -1.06153095e+00 -7.14408875e-01 -6.03931963e-01 -6.28212094e-01 -4.69988175e-02 7.84874201e-01 -7.95351803e-01 -3.75404388e-01 -3.70210975e-01 -1.25893235e+00 1.15730554e-01 -6.27574563e-01 4.75578874e-01 -4.58360970e-01 3.93405467e-01 1.26291290e-01 -8.58923972e-01 -2.82269150e-01 -5.81044972e-01 7.44272768e-01 -2.29938731e-01 -8.26553226e-01 -1.52550566e+00 1.88434795e-01 2.02036098e-01 1.84249446e-01 5.10059118e-01 1.26498210e+00 -1.03507435e+00 3.51820260e-01 -2.81013280e-01 7.70255476e-02 6.00075901e-01 1.48786739e-01 2.24483758e-01 -1.02171218e+00 -2.08825357e-02 2.91641116e-01 -3.24859262e-01 2.95787632e-01 -1.34054229e-01 4.37194228e-01 -3.18108559e-01 -2.05283333e-02 -2.82294780e-01 1.71178746e+00 5.48707042e-03 6.70117557e-01 7.30182946e-01 2.61883140e-01 9.54113007e-01 6.02721751e-01 5.97002625e-01 4.34016615e-01 3.09441626e-01 2.49707147e-01 4.62394536e-01 2.74896175e-01 -1.29428208e-01 1.92183048e-01 1.15000939e+00 -3.83497961e-02 -2.90213406e-01 -1.39062846e+00 1.37190652e+00 -1.44427407e+00 -1.00016415e+00 -3.96346629e-01 1.94388723e+00 9.46884215e-01 3.65591049e-01 -1.38771757e-01 4.44355577e-01 4.30402249e-01 -1.87100135e-02 2.24257275e-01 -4.53092843e-01 -1.17151022e-01 6.36487603e-01 1.52755260e-01 4.45001602e-01 -7.90861130e-01 8.39339077e-01 6.01408339e+00 7.66734838e-01 -3.81898344e-01 3.76327962e-01 -4.00222689e-01 3.61406654e-01 -9.77402270e-01 2.58493900e-01 -4.14750874e-01 5.67685425e-01 1.11034060e+00 -4.49500352e-01 3.81084229e-03 7.87421405e-01 8.58976878e-03 6.73671514e-02 -1.11597419e+00 7.91357875e-01 2.35996380e-01 -1.49546945e+00 -4.99998126e-03 7.51228025e-03 3.53065044e-01 -3.30003083e-01 -2.23047465e-01 4.40342963e-01 2.95690268e-01 -1.18005037e+00 1.02845240e+00 3.66568238e-01 6.51369452e-01 -7.85361350e-01 1.08148193e+00 1.54860362e-01 -6.36553824e-01 1.20636150e-01 -6.64459169e-01 -9.53602418e-02 1.35234937e-01 7.34804928e-01 -5.36674619e-01 7.97098935e-01 4.88978654e-01 5.73803842e-01 -5.01228690e-01 4.34080422e-01 -3.36711109e-01 6.74084127e-01 -1.69317741e-02 -4.06563669e-01 5.49001813e-01 -5.02356663e-02 4.79924172e-01 1.37898791e+00 1.34622186e-01 3.82040814e-02 -3.26158583e-01 1.22709846e+00 2.15129420e-01 2.84196079e-01 -5.77265322e-01 -8.49857688e-01 7.63134122e-01 1.05194366e+00 -5.51324308e-01 -1.84676155e-01 -5.30326068e-01 6.67889535e-01 1.91043794e-01 9.48560536e-02 -5.13142288e-01 -6.24261439e-01 7.46195257e-01 2.36313000e-01 3.26006323e-01 -1.31863967e-01 -2.89940834e-01 -1.07951725e+00 1.72071144e-01 -1.04253435e+00 3.55801247e-02 -6.20080888e-01 -1.78278100e+00 6.60520196e-01 2.03522429e-01 -9.21127975e-01 -2.98855513e-01 -1.09895813e+00 -3.14669281e-01 7.67038584e-01 -1.08004904e+00 -9.91102219e-01 9.62904934e-03 2.17967004e-01 1.17956415e-01 -2.07003459e-01 1.28132260e+00 1.86198235e-01 -1.33325145e-01 2.62385547e-01 1.22248836e-01 6.16192631e-02 4.43869501e-01 -1.45556557e+00 4.26390976e-01 5.67671835e-01 2.89392143e-01 1.10960078e+00 1.28323925e+00 -5.44726074e-01 -1.15969217e+00 -5.70043564e-01 1.50352657e+00 -1.03342509e+00 1.35157335e+00 -3.77395928e-01 -1.12387133e+00 3.88370305e-01 3.79997313e-01 -5.23540437e-01 1.01072383e+00 4.79746342e-01 -5.04526854e-01 3.04433167e-01 -1.13602591e+00 5.90762436e-01 7.26909876e-01 -9.58466351e-01 -1.86298120e+00 3.51078957e-01 7.75619328e-01 1.20285647e-02 -1.28742111e+00 -2.51615234e-02 3.50266188e-01 -6.06010735e-01 8.56260896e-01 -7.67238617e-01 7.49830604e-01 -6.47913069e-02 -5.07696807e-01 -1.22156477e+00 -1.65607646e-01 -3.35380673e-01 1.88365847e-01 1.41533208e+00 3.93760502e-01 -6.05366647e-01 3.13158929e-01 5.96926272e-01 -3.15067351e-01 -5.03667891e-01 -1.09085488e+00 -7.75105298e-01 7.93318212e-01 -8.93605828e-01 6.08282745e-01 1.39056289e+00 4.91004080e-01 3.38507742e-01 2.62877911e-01 -1.45485371e-01 3.78631860e-01 -3.43058169e-01 4.14111286e-01 -1.56315684e+00 1.75105348e-01 -5.52466214e-01 -7.92619407e-01 -3.16273630e-01 3.76841336e-01 -8.31138074e-01 -8.22818279e-02 -1.87043691e+00 -1.73817292e-01 -1.36183262e-01 -1.65623605e-01 7.36659020e-02 -2.99794339e-02 -1.48615107e-01 2.30716597e-02 -1.35016412e-01 -3.29604000e-02 6.28627241e-01 8.26126218e-01 -3.10708676e-02 1.81145981e-01 -8.17691267e-01 -1.17726827e+00 5.95711529e-01 5.59339941e-01 -6.60880804e-01 -6.12745285e-01 -1.80742994e-01 7.41191566e-01 -6.10465825e-01 6.60592198e-01 -7.06827462e-01 8.04842934e-02 -1.87955916e-01 -2.24108890e-01 -2.19455019e-01 1.47065362e-02 -1.00211418e+00 -2.74420884e-02 1.22064631e-02 -2.24287778e-01 2.95403600e-01 3.13365966e-01 2.87487477e-01 -2.83917844e-01 -1.04144144e+00 3.47895116e-01 -3.59640330e-01 -1.10633206e+00 -2.49315321e-01 -6.00521207e-01 8.24193120e-01 7.66760230e-01 -3.77937585e-01 -9.68620926e-02 5.77653125e-02 -3.00833553e-01 1.29464632e-02 6.64051294e-01 5.48732221e-01 7.67133355e-01 -1.47720134e+00 -9.58932161e-01 -2.09546953e-01 6.31868362e-01 -2.30875343e-01 7.59168565e-02 2.91267037e-01 -6.07258379e-01 4.75821227e-01 -1.59252807e-01 -1.68382358e-02 -4.90657628e-01 8.15450668e-01 5.51717058e-02 2.02816427e-01 -7.74912298e-01 6.99888647e-01 -1.55596942e-01 -4.23411369e-01 8.21358636e-02 -4.82544214e-01 -6.56727850e-01 5.79485714e-01 3.91804636e-01 2.28270054e-01 1.54446736e-01 -7.06885159e-01 -4.76560771e-01 2.47774377e-01 3.13297689e-01 -4.95982349e-01 1.43987763e+00 1.05536981e-02 -2.33905569e-01 9.07518387e-01 9.64996159e-01 -3.11445538e-02 -3.52113634e-01 -6.81966618e-02 4.49458271e-01 -4.12761718e-01 3.69666457e-01 -7.63545632e-01 -2.48329982e-01 8.70086133e-01 3.26504558e-01 4.63597029e-01 6.00122273e-01 -3.78743261e-02 5.22209525e-01 2.21002236e-01 4.38800812e-01 -1.38091791e+00 -1.02983885e-01 4.86783862e-01 1.08745039e+00 -9.46079850e-01 6.04288392e-02 1.31980563e-02 -7.84651577e-01 1.15602839e+00 1.95796385e-01 2.75454149e-02 6.39299572e-01 -9.60997716e-02 -3.86651345e-02 -8.20438087e-01 -5.12547076e-01 -3.39973599e-01 3.90775502e-01 9.03145671e-01 6.26519382e-01 8.52229744e-02 -7.39816308e-01 1.02584577e+00 -5.19928515e-01 -9.38784108e-02 6.26416564e-01 9.59603131e-01 -5.19944906e-01 -9.94080663e-01 -2.09489912e-01 9.89307538e-02 -5.43499172e-01 -4.35524464e-01 -4.87143129e-01 1.13024628e+00 4.45587248e-01 1.09362590e+00 4.15712669e-02 -3.58677745e-01 5.60941756e-01 3.61021847e-01 3.27811569e-01 -7.68968582e-01 -5.70351183e-01 -5.71777463e-01 5.06212234e-01 -3.44753638e-02 -8.08816075e-01 -5.45797706e-01 -9.78407264e-01 -5.04694223e-01 -2.97196239e-01 5.32217920e-01 8.13304543e-01 1.23411644e+00 1.26108080e-01 4.75068092e-01 -5.62315434e-02 -2.52623826e-01 -5.49289703e-01 -1.12393641e+00 -6.27902985e-01 4.94991690e-01 8.94673821e-03 -1.04575777e+00 -5.80990076e-01 -1.69941112e-02]
[10.275096893310547, 8.877296447753906]
7a221578-1882-406a-91e7-d37578a0825e
line-as-object-datasets-and-framework-for
1909.06591
null
https://arxiv.org/abs/1909.06591v2
https://arxiv.org/pdf/1909.06591v2.pdf
Sem-LSD: A Learning-based Semantic Line Segment Detector
In this paper, we introduces a new type of line-shaped image representation, named semantic line segment (Sem-LS) and focus on solving its detection problem. Sem-LS contains high-level semantics and is a compact scene representation where only visually salient line segments with stable semantics are preserved. Combined with high-level semantics, Sem-LS is more robust under cluttered environment compared with existing line-shaped representations. The compactness of Sem-LS facilitates its use in large-scale applications, such as city-scale SLAM (simultaneously localization and mapping) and LCD (loop closure detection). Sem-LS detection is a challenging task due to its significantly different appearance from existing learning-based image representations such as wireframes and objects. For further investigation, we first label Sem-LS on two well-known datasets, KITTI and KAIST URBAN, as new benchmarks. Then, we propose a learning-based Sem-LS detector (Sem-LSD) and devise new module as well as metrics to address unique challenges in Sem-LS detection. Experimental results have shown both the efficacy and efficiency of Sem-LSD. Finally, the effectiveness of the proposed Sem-LS is supported by two experiments on detector repeatability and a city-scale LCD problem. Labeled datasets and code will be released shortly.
['Mingyang Li', 'Boren Li', 'Yi Sun', 'Kai Sun', 'Yongjiang Chen', 'Xushen Han']
2019-09-14
null
null
null
null
['line-segment-detection', 'loop-closure-detection']
['computer-vision', 'computer-vision']
[ 1.08078077e-01 -7.67748728e-02 -1.57448635e-01 -2.98473656e-01 -5.09680629e-01 -3.22191298e-01 6.03918672e-01 4.45391268e-01 -3.14653724e-01 2.75555074e-01 -1.90169550e-02 -2.47403756e-01 -5.38319983e-02 -6.39906406e-01 -8.22747707e-01 -1.58248097e-01 -3.51531029e-01 8.36093649e-02 9.26118255e-01 -2.46393174e-01 3.64756107e-01 6.84359133e-01 -1.43072939e+00 3.79267149e-02 7.66171575e-01 1.00977850e+00 6.10211670e-01 1.20563835e-01 -1.63476616e-01 6.65843546e-01 -2.07950890e-01 1.52878270e-01 4.05970752e-01 -2.56945699e-01 -4.82657462e-01 2.70053655e-01 6.74344182e-01 1.39777333e-01 -2.72086948e-01 1.03843284e+00 3.55688453e-01 1.65250733e-01 5.52207649e-01 -1.45849407e+00 -2.26491764e-01 3.16969842e-01 -1.03836215e+00 8.62393677e-02 6.79284871e-01 -1.36158377e-01 8.54927301e-01 -1.21887350e+00 8.84379804e-01 1.44486701e+00 1.00395715e+00 -1.20488539e-01 -9.79928195e-01 -5.66935658e-01 3.82674396e-01 2.02712178e-01 -1.75704873e+00 -1.23958282e-01 8.56418192e-01 -4.25335914e-01 8.82837236e-01 1.43996492e-01 5.88888526e-01 6.84665203e-01 1.60402089e-01 1.23058486e+00 1.19652379e+00 -6.36474192e-01 1.90586329e-01 1.32971615e-01 6.17433786e-02 1.17710543e+00 5.53588152e-01 -1.18889473e-02 -5.17609119e-01 1.76329508e-01 9.08595860e-01 -1.18809089e-01 -3.32768977e-01 -1.47771323e+00 -1.57670510e+00 6.55832469e-01 7.41327703e-01 2.08098650e-01 4.71129380e-02 1.52756795e-01 4.91675496e-01 4.56240326e-02 1.58564731e-01 3.57411265e-01 -7.04044104e-02 1.74322709e-01 -1.07373607e+00 1.19427681e-01 3.63980681e-01 1.44732285e+00 9.12889719e-01 2.43269047e-03 -1.39212430e-01 7.70398438e-01 2.94819295e-01 7.29721725e-01 4.84274477e-01 -5.07639170e-01 5.09364665e-01 7.55735219e-01 1.83169559e-01 -1.46689522e+00 -1.01151657e+00 -4.90087420e-01 -7.05090344e-01 2.14083940e-01 9.72928181e-02 2.88089871e-01 -8.83844078e-01 1.46944785e+00 3.02172869e-01 3.73567522e-01 5.17679304e-02 9.77451324e-01 9.16279078e-01 5.48676670e-01 -1.78226218e-01 8.61917902e-03 1.23187649e+00 -1.35371351e+00 -4.12338823e-01 -5.06275237e-01 8.32395554e-01 -7.25338399e-01 1.19938481e+00 9.47740376e-02 -6.95690989e-01 -6.18877232e-01 -1.24491251e+00 -8.14960301e-02 -5.33560574e-01 4.40224409e-01 7.28744686e-01 3.70462090e-01 -8.40617299e-01 7.38141164e-02 -6.37287855e-01 -1.10757244e+00 2.43953377e-01 -5.55633940e-02 -5.02795398e-01 3.36851887e-02 -7.98644125e-01 7.53562927e-01 6.87212586e-01 -1.40314177e-01 -5.74864089e-01 -3.97356838e-01 -1.35163569e+00 -1.13373131e-01 5.86904526e-01 -4.57094342e-01 9.44756031e-01 -5.73254108e-01 -1.02605700e+00 1.14221907e+00 -4.48434539e-02 -5.48221231e-01 7.91681528e-01 -3.01944137e-01 -5.90918660e-01 1.38114750e-01 5.97082496e-01 6.55957580e-01 7.46712506e-01 -1.65696418e+00 -1.01605511e+00 -7.94218928e-02 -1.06748521e-01 2.87104517e-01 1.59670692e-02 -1.52702212e-01 -7.99335599e-01 -6.48168504e-01 7.53145814e-01 -9.28649187e-01 -1.55328602e-01 1.82040781e-01 -6.29389346e-01 -1.36491165e-01 1.12073398e+00 -1.10250697e-01 1.35810959e+00 -2.22541189e+00 -3.28005522e-01 3.23339611e-01 -3.01793944e-02 2.99374759e-01 -1.48281425e-01 6.37354314e-01 5.28346375e-02 -2.20591664e-01 -3.34186941e-01 -2.96355069e-01 -1.27929717e-01 -8.09816793e-02 -3.37510198e-01 6.91524267e-01 3.12328953e-02 8.65931809e-01 -1.12114561e+00 -7.56344676e-01 7.16830432e-01 -4.86823265e-03 -1.56756744e-01 -1.29104242e-01 -8.50648433e-02 1.69403210e-01 -5.21280050e-01 9.55377162e-01 8.56584311e-01 -1.51927710e-01 -2.28706151e-01 -4.46106255e-01 -4.71110106e-01 -3.71417224e-01 -1.51577973e+00 2.02813888e+00 -4.20457035e-01 7.08536386e-01 -3.27578098e-01 -8.58415484e-01 1.30904496e+00 -4.67568547e-01 3.87930781e-01 -9.99951303e-01 3.20217423e-02 3.32447648e-01 -5.04487932e-01 -2.98487306e-01 7.75721371e-01 4.58204478e-01 -4.17946965e-01 1.07136317e-01 -1.92602336e-01 -3.08215052e-01 2.23468661e-01 4.11201775e-01 8.96815836e-01 3.49704742e-01 5.85509181e-01 -5.64010441e-01 6.89852417e-01 3.15184593e-01 5.22322059e-01 9.40804422e-01 -4.43820983e-01 6.82271600e-01 3.10540870e-02 -4.62326080e-01 -8.57246816e-01 -1.12229240e+00 -2.64884323e-01 7.34102845e-01 1.10841668e+00 -7.65529692e-01 -4.42185640e-01 -5.88895321e-01 3.43753099e-01 6.21096790e-01 -4.31423724e-01 -2.67850589e-02 -5.43462873e-01 -2.72282690e-01 5.38017213e-01 5.14294505e-01 9.05871212e-01 -8.79893064e-01 -1.07274151e+00 1.44699039e-02 -4.49507385e-02 -1.43883240e+00 -3.15039843e-01 7.76759088e-02 -6.13499105e-01 -1.14231038e+00 -4.02736753e-01 -1.11386299e+00 6.24715686e-01 1.11510479e+00 9.54009652e-01 -1.76370442e-01 -4.82045442e-01 4.46549594e-01 -3.75784695e-01 -4.73113269e-01 1.14364296e-01 -1.37637272e-01 1.35800570e-01 7.05654994e-02 1.75894037e-01 -1.58293635e-01 -6.09758914e-01 4.72336680e-01 -5.76191604e-01 4.46357340e-01 7.03168988e-01 6.35318518e-01 9.66640055e-01 -1.11860577e-02 1.02483615e-01 -8.83618057e-01 4.03923810e-01 -2.39249542e-01 -7.29995906e-01 3.35059166e-01 -4.99928206e-01 -9.31209177e-02 3.46112132e-01 -3.29051986e-02 -8.83908391e-01 2.48870015e-01 3.55925500e-01 -5.06508708e-01 -4.00606692e-02 4.51067179e-01 9.39278230e-02 -3.18114072e-01 5.88943779e-01 2.77291179e-01 -3.36997598e-01 -5.21632910e-01 4.94379133e-01 4.87301648e-01 8.92202437e-01 -4.23028022e-01 9.90005553e-01 8.29377234e-01 1.56973749e-01 -1.13986623e+00 -9.47584629e-01 -1.01864707e+00 -7.96705782e-01 -2.23823845e-01 5.36967635e-01 -1.18794382e+00 -2.57070035e-01 4.17891771e-01 -8.27979386e-01 -1.88140228e-01 -3.00462723e-01 3.96474868e-01 -7.99233735e-01 5.41973293e-01 -1.17569298e-01 -7.20837653e-01 -8.78013745e-02 -8.95286500e-01 1.46335685e+00 1.89489320e-01 5.86214056e-03 -9.23940301e-01 7.45638460e-02 -6.82462156e-02 -2.94483476e-03 7.41731942e-01 5.87549150e-01 -2.76615053e-01 -6.36466682e-01 -2.68835515e-01 -6.24262750e-01 -9.33903307e-02 -8.15887675e-02 -2.43263900e-01 -6.61412299e-01 -6.08043611e-01 -4.53485012e-01 -4.09711570e-01 1.11901653e+00 1.47487715e-01 1.01098859e+00 3.96612316e-01 -8.66675138e-01 9.01870787e-01 1.78331411e+00 -2.76849885e-02 3.29143018e-01 9.00013208e-01 8.38763952e-01 3.70179415e-01 1.27701700e+00 4.39575613e-01 4.96209711e-01 8.83882403e-01 4.25957024e-01 -5.59579849e-01 -3.53898406e-01 -7.36790359e-01 3.37644637e-01 6.13973498e-01 2.88358986e-01 -8.12391862e-02 -9.98343289e-01 4.75226581e-01 -2.34621072e+00 -8.25408220e-01 -5.38815022e-01 2.07334542e+00 1.33757770e-01 3.38020116e-01 1.64760277e-01 1.07309677e-01 5.88277996e-01 2.05123782e-01 -4.35773104e-01 -1.18430831e-01 -5.42062879e-01 -2.62216777e-01 8.79420340e-01 2.50152528e-01 -1.65114927e+00 1.49565220e+00 5.76116943e+00 1.03337359e+00 -1.22378850e+00 6.57100454e-02 -5.27666183e-04 5.06895661e-01 -2.40245857e-03 1.66637097e-02 -9.10593212e-01 2.65240371e-01 9.88745764e-02 -1.00177675e-01 -2.78105706e-01 1.15432668e+00 3.45372945e-01 -5.07457912e-01 -9.17675316e-01 1.47213137e+00 3.08118314e-01 -1.27151978e+00 -2.29400955e-03 -4.56933349e-01 8.46189916e-01 2.17044085e-01 -1.28425121e-01 2.72929519e-01 6.00722879e-02 -6.53577924e-01 1.34704936e+00 2.63754368e-01 9.49103057e-01 -6.14654422e-01 5.04988790e-01 3.37259203e-01 -1.86282015e+00 -9.12487656e-02 -6.33694947e-01 9.51792821e-02 4.64840196e-02 3.61451924e-01 -8.42895269e-01 7.64588237e-01 7.60730684e-01 1.18731177e+00 -1.06770349e+00 1.51829422e+00 -2.29858249e-01 2.81470150e-01 -4.95834291e-01 9.48143750e-02 6.40331805e-01 -1.71894744e-01 5.83853602e-01 1.73266721e+00 3.67778629e-01 -3.67922693e-01 7.67440021e-01 7.52204001e-01 3.87997895e-01 3.77992302e-01 -9.43007708e-01 4.35871154e-01 6.43059373e-01 1.20322561e+00 -1.27804065e+00 -9.18349922e-02 -5.85302591e-01 1.08077896e+00 4.50488240e-01 3.67762506e-01 -8.23844731e-01 -4.50309366e-01 4.22713727e-01 2.60760427e-01 3.97583723e-01 -5.75248897e-01 -6.74806014e-02 -1.11523807e+00 -5.99038340e-02 -2.54026324e-01 2.92223155e-01 -9.07253802e-01 -6.70251250e-01 5.09481311e-01 6.55103754e-03 -1.72082794e+00 2.06001565e-01 -5.10999322e-01 -3.52954596e-01 1.54441863e-01 -1.84046626e+00 -1.63241851e+00 -7.89616048e-01 8.54626119e-01 6.97594106e-01 -1.31131634e-01 5.19431233e-01 7.48014897e-02 -6.07924402e-01 5.09019256e-01 4.04742479e-01 7.95147121e-02 6.94949329e-01 -1.18562746e+00 4.16549712e-01 1.00042868e+00 4.36665267e-01 2.98065752e-01 5.70153356e-01 -6.03068531e-01 -1.39033151e+00 -1.47359717e+00 5.97768426e-01 -2.62280107e-01 5.06228030e-01 -6.26583040e-01 -5.53973615e-01 7.16099203e-01 -2.74203330e-01 2.03968912e-01 1.25288397e-01 -2.81723887e-01 -3.46864939e-01 -1.22077450e-01 -9.78734434e-01 5.53340912e-01 1.44075966e+00 -2.89388806e-01 -5.62948406e-01 3.81380409e-01 6.54784620e-01 -5.01738906e-01 -2.93308973e-01 6.17018640e-01 2.90090084e-01 -1.30019784e+00 1.21980953e+00 -4.08423766e-02 -2.10618615e-01 -7.71525443e-01 -2.09635884e-01 -1.07539129e+00 -5.23106635e-01 -3.40481609e-01 2.39039660e-02 1.11417508e+00 -8.19618255e-02 -3.21344405e-01 6.32968426e-01 -2.88657606e-01 -4.59881663e-01 -6.86686039e-01 -6.40486002e-01 -1.24995375e+00 -4.74437237e-01 -6.44666970e-01 3.60466421e-01 8.78543258e-01 -1.54446423e-01 1.89883336e-01 -4.28994387e-01 4.23155338e-01 7.96581745e-01 6.07012391e-01 1.04518235e+00 -1.09290445e+00 3.35348219e-01 -3.82924110e-01 -9.46764112e-01 -1.26575387e+00 -2.93619707e-02 -1.05594969e+00 1.00143157e-01 -1.79501855e+00 7.21729733e-03 -6.67082727e-01 -4.12084043e-01 3.05144519e-01 1.48041010e-01 3.85919780e-01 4.05909240e-01 3.49196345e-01 -1.19385982e+00 6.02996826e-01 9.67663467e-01 -1.51840493e-01 -2.17895731e-01 -3.42112988e-01 -1.99442148e-01 1.08511639e+00 5.40510178e-01 -7.53020346e-02 -5.07836759e-01 -2.99541593e-01 -9.25312862e-02 -2.95034617e-01 3.86497378e-01 -1.50894976e+00 3.93870145e-01 -7.77334571e-02 3.01964223e-01 -1.03350759e+00 2.03644902e-01 -8.75146925e-01 -1.23244129e-01 6.15635693e-01 -3.55277807e-02 -1.82878617e-02 2.17578724e-01 8.08296978e-01 -3.80092889e-01 -2.77787559e-02 9.43417668e-01 -4.24318714e-04 -1.90638280e+00 6.50058687e-02 -7.08638802e-02 8.09662417e-02 1.60904098e+00 -6.32105291e-01 -2.64222231e-02 -2.19833717e-01 -4.02601331e-01 5.49002528e-01 7.82039404e-01 7.60731637e-01 1.02950490e+00 -1.49308479e+00 -4.39232469e-01 4.92932588e-01 9.26873863e-01 2.26148181e-02 -4.82068509e-02 8.44527245e-01 -1.10215592e+00 6.88807666e-01 -2.81677365e-01 -9.36662912e-01 -1.15195107e+00 8.35367084e-01 9.25450996e-02 9.47766975e-02 -1.04436791e+00 8.08343649e-01 5.05379379e-01 -2.90469795e-01 3.91369700e-01 -6.88959420e-01 -5.52961640e-02 -1.24084502e-01 2.53478527e-01 3.00673693e-01 -3.94597910e-02 -9.95852470e-01 -6.33302212e-01 1.08502901e+00 2.68842012e-01 4.46930051e-01 9.86645639e-01 -4.58203197e-01 1.23828731e-01 5.15676022e-01 9.62591350e-01 3.26457694e-02 -1.09956443e+00 -5.19091427e-01 3.61536562e-01 -6.34491801e-01 -3.73043604e-02 -3.88313681e-01 -7.11967170e-01 7.18196571e-01 7.83646047e-01 -2.20173344e-01 8.55462551e-01 5.61016686e-02 7.34887719e-01 5.08733690e-01 1.07687271e+00 -1.12297618e+00 2.85938203e-01 5.18852949e-01 9.38800037e-01 -1.45242178e+00 7.01951683e-02 -8.25951457e-01 -5.09166360e-01 1.05097270e+00 7.67564714e-01 -2.70614088e-01 4.14469659e-01 7.99797848e-02 -5.33294007e-02 -2.93630987e-01 -1.09398045e-01 -5.99877179e-01 2.98726082e-01 6.91341758e-01 -3.40908729e-02 1.43210962e-01 -1.89045593e-01 1.93686977e-01 -1.60596102e-01 -1.81280658e-01 3.50090235e-01 1.10888302e+00 -8.12864721e-01 -6.51506424e-01 -4.97611791e-01 2.17668906e-01 4.52775687e-01 2.08562165e-01 -3.33205819e-01 1.22618496e+00 2.98224270e-01 8.21328819e-01 4.61914018e-02 -4.52844113e-01 6.05253398e-01 -3.63537133e-01 4.29916143e-01 -5.56500793e-01 8.98824185e-02 7.83188175e-03 -1.68723743e-02 -1.02047253e+00 -4.48339254e-01 -6.51500642e-01 -1.60847902e+00 4.90588397e-02 -2.75720805e-01 -9.61880907e-02 4.79864061e-01 5.66327929e-01 3.03239584e-01 1.80096403e-01 4.61640775e-01 -7.92241037e-01 -2.06658781e-01 -4.97678638e-01 -7.97130883e-01 7.96367049e-01 2.89445639e-01 -1.17400658e+00 3.41061130e-02 -2.32939079e-01]
[7.7646613121032715, -1.941810965538025]
98f8bb64-7c4a-4a68-ae35-73c84d5d05d1
multimodal-prototype-enhanced-network-for-few
2212.04873
null
https://arxiv.org/abs/2212.04873v1
https://arxiv.org/pdf/2212.04873v1.pdf
Multimodal Prototype-Enhanced Network for Few-Shot Action Recognition
Current methods for few-shot action recognition mainly fall into the metric learning framework following ProtoNet. However, they either ignore the effect of representative prototypes or fail to enhance the prototypes with multimodal information adequately. In this work, we propose a novel Multimodal Prototype-Enhanced Network (MORN) to use the semantic information of label texts as multimodal information to enhance prototypes, including two modality flows. A CLIP visual encoder is introduced in the visual flow, and visual prototypes are computed by the Temporal-Relational CrossTransformer (TRX) module. A frozen CLIP text encoder is introduced in the text flow, and a semantic-enhanced module is used to enhance text features. After inflating, text prototypes are obtained. The final multimodal prototypes are then computed by a multimodal prototype-enhanced module. Besides, there exist no evaluation metrics to evaluate the quality of prototypes. To the best of our knowledge, we are the first to propose a prototype evaluation metric called Prototype Similarity Difference (PRIDE), which is used to evaluate the performance of prototypes in discriminating different categories. We conduct extensive experiments on four popular datasets. MORN achieves state-of-the-art results on HMDB51, UCF101, Kinetics and SSv2. MORN also performs well on PRIDE, and we explore the correlation between PRIDE and accuracy.
['Yujiu Yang', 'Yatai Ji', 'Yong liu', 'Hao Wen', 'Xinzhe Ni']
2022-12-09
null
null
null
null
['few-shot-action-recognition', 'metric-learning', 'metric-learning']
['computer-vision', 'computer-vision', 'methodology']
[ 1.35873944e-01 -5.39341390e-01 -3.50121170e-01 -3.94979954e-01 -7.92157769e-01 -2.30002627e-01 8.77424002e-01 1.24755176e-02 -6.99018359e-01 3.72393191e-01 3.99354607e-01 2.80028611e-01 -2.59129882e-01 -4.61767405e-01 -2.90596396e-01 -5.75599194e-01 2.09321752e-01 2.61661083e-01 4.08813357e-01 -1.78296149e-01 3.28604966e-01 7.47440010e-02 -1.78983223e+00 9.01911557e-01 8.02118778e-01 1.29114497e+00 1.86041132e-01 4.26205635e-01 -4.02298480e-01 1.14801407e+00 -5.56833029e-01 -5.07143259e-01 -1.31819353e-01 -7.26022720e-01 -6.94133937e-01 1.79809406e-01 1.84409961e-01 -2.73406625e-01 -5.00793517e-01 9.24407899e-01 7.87814379e-01 4.49393153e-01 8.00687909e-01 -1.50707996e+00 -7.17139304e-01 8.37996006e-01 -4.16665614e-01 -3.00469398e-02 4.63585913e-01 1.63219810e-01 1.02674186e+00 -1.34642625e+00 7.65735447e-01 1.36398017e+00 4.79297876e-01 5.38537085e-01 -8.24269891e-01 -4.04623389e-01 8.79619494e-02 8.82210195e-01 -1.48220706e+00 -3.52642715e-01 9.13866460e-01 -3.32765311e-01 7.03879416e-01 2.44602948e-01 5.32017708e-01 1.47938216e+00 -1.08144052e-01 1.25963700e+00 8.60005498e-01 -3.52380514e-01 3.99454623e-01 3.59362185e-01 2.35914990e-01 5.55868924e-01 -3.22722912e-01 3.03773955e-02 -7.20867097e-01 1.80394098e-01 2.36354440e-01 3.11159790e-01 -4.13428962e-01 -3.41999799e-01 -1.21910846e+00 7.77516723e-01 5.13423622e-01 5.61941266e-01 -2.27734953e-01 -2.35121325e-01 7.71444857e-01 2.17646495e-01 9.72800702e-02 2.68972963e-01 -3.96669321e-02 -3.70346814e-01 -9.61932719e-01 -1.47801921e-01 4.23238724e-01 8.29640985e-01 5.34865379e-01 -3.08761030e-01 -8.93533945e-01 1.17872095e+00 2.70556420e-01 4.42973167e-01 8.61342072e-01 -9.57604468e-01 5.20817041e-01 8.85984719e-01 -1.88306183e-01 -8.14555287e-01 -4.63618100e-01 -9.71841277e-04 -8.99441063e-01 -2.64781378e-02 4.11151126e-02 5.07908836e-02 -9.28287566e-01 1.38978124e+00 1.40871629e-01 1.22876242e-01 1.56212643e-01 1.05858088e+00 1.25674510e+00 6.70935631e-01 1.54295459e-01 -2.79973209e-01 1.23851573e+00 -1.25705159e+00 -7.78077185e-01 2.49473408e-01 8.10611844e-01 -7.03716159e-01 1.28747904e+00 3.13404173e-01 -9.27757025e-01 -8.87072206e-01 -1.08004141e+00 -3.40398327e-02 -6.15939677e-01 4.23888117e-01 1.22664511e-01 3.59237075e-01 -6.32921576e-01 6.67811155e-01 -6.73870385e-01 -5.16362488e-01 5.19580185e-01 -1.68642193e-01 -3.49661827e-01 -2.27391973e-01 -1.40235603e+00 1.01821518e+00 6.92878067e-01 4.25277539e-02 -8.38247299e-01 -5.65267026e-01 -9.69422281e-01 -4.58465628e-02 4.23292518e-01 -5.36845088e-01 1.17077482e+00 -1.09891891e+00 -1.44999135e+00 4.09083992e-01 1.00474626e-01 -2.78976649e-01 5.82300782e-01 1.46906644e-01 -7.69548416e-01 4.59255904e-01 -1.15808181e-01 8.95612895e-01 6.07364714e-01 -1.28922260e+00 -8.33174825e-01 -1.91057771e-01 8.14157575e-02 4.76855427e-01 -7.83129990e-01 4.49720360e-02 -6.15530789e-01 -5.24741888e-01 -1.92442134e-01 -6.34131134e-01 1.47268832e-01 6.38614669e-02 -3.22441667e-01 -4.19064373e-01 9.83667076e-01 -4.98747587e-01 1.53773844e+00 -2.30303359e+00 2.89792806e-01 -3.57690267e-02 -6.01711683e-02 4.88242418e-01 -4.97217894e-01 4.73686785e-01 6.86256289e-02 -6.87202215e-02 -3.37265104e-01 -5.48811555e-01 3.12318772e-01 1.64344445e-01 -3.51673253e-02 1.60312578e-01 1.30056605e-01 1.12908673e+00 -8.18686426e-01 -9.22785819e-01 4.19206977e-01 6.42854929e-01 -1.69069171e-01 1.99936375e-01 -5.48096299e-02 8.73562694e-02 -2.47513607e-01 7.34315455e-01 3.82648587e-01 -2.03841627e-01 -3.36461850e-02 -5.30580759e-01 -3.05936150e-02 -3.12559515e-01 -1.25472343e+00 1.93040216e+00 -2.59503782e-01 5.21096051e-01 -5.25384843e-01 -9.29459214e-01 8.50768268e-01 2.68867552e-01 5.04766226e-01 -1.03072536e+00 4.26623106e-01 7.35505819e-02 -2.39901930e-01 -8.13916326e-01 7.53137529e-01 -4.24305946e-02 -6.08856417e-02 2.73995817e-01 2.10777253e-01 2.87608147e-01 6.09808147e-01 3.45503747e-01 9.47093189e-01 2.46648639e-01 9.43454131e-02 3.29755723e-01 7.10515320e-01 -1.17525144e-03 3.32803637e-01 5.48325062e-01 -3.57874274e-01 7.43704498e-01 4.50745642e-01 -2.14729846e-01 -8.96396339e-01 -1.02209377e+00 -1.40290186e-01 1.26556075e+00 3.76756966e-01 -6.40062630e-01 -5.63775361e-01 -1.05799758e+00 -2.00735122e-01 7.89631546e-01 -7.26961136e-01 -4.52282429e-01 -2.02268660e-01 -4.56568360e-01 5.74099183e-01 8.15190971e-01 7.25581467e-01 -1.21213281e+00 -5.02171516e-01 4.64845523e-02 -5.40324926e-01 -9.20441806e-01 -6.53243005e-01 -1.97991103e-01 -5.26724041e-01 -1.05187857e+00 -9.70712364e-01 -6.59146190e-01 3.58892322e-01 2.25875556e-01 6.67668700e-01 -2.09206879e-01 -2.54344136e-01 4.95338619e-01 -9.08878624e-01 -5.77837341e-02 -1.98179170e-01 -4.38166223e-02 -2.05551311e-01 2.96369195e-01 4.01746333e-01 -1.62165970e-01 -4.86954808e-01 5.67547739e-01 -1.03406191e+00 5.70937991e-02 7.10452020e-01 9.77901638e-01 7.00250685e-01 -7.67870899e-03 3.30713123e-01 -4.22092319e-01 6.91012383e-01 -3.30250710e-01 -5.76458909e-02 6.86255932e-01 -4.64345902e-01 7.41944239e-02 6.48024619e-01 -8.16064596e-01 -1.25968897e+00 5.53238988e-02 -5.41534908e-02 -8.06718528e-01 -2.40799353e-01 7.35464811e-01 -3.00463349e-01 2.36969247e-01 6.63175285e-01 2.93594390e-01 -1.82961911e-01 -4.81742293e-01 6.39280736e-01 1.02903521e+00 5.48776567e-01 -1.81781709e-01 2.78447270e-01 3.35521877e-01 -1.58381328e-01 -7.44731069e-01 -8.06767106e-01 -6.30249500e-01 -6.40978575e-01 -5.07235169e-01 9.56942677e-01 -6.62319005e-01 -5.84335744e-01 4.22498286e-01 -1.00413072e+00 -1.14924863e-01 -4.52576280e-01 7.32843637e-01 -6.65871501e-01 5.66832900e-01 -4.81161773e-01 -6.77511275e-01 -3.56924772e-01 -9.89210188e-01 1.03104019e+00 2.55471021e-01 2.64517497e-02 -6.56880677e-01 1.28846467e-01 3.42580020e-01 2.55469650e-01 8.73483438e-03 6.37449265e-01 -6.39718592e-01 -2.64212750e-02 -1.03948668e-01 -3.73805463e-01 4.99932259e-01 -7.22650960e-02 8.10259581e-02 -9.27572072e-01 -1.55819818e-01 -2.93520629e-01 -6.01792634e-01 1.26367056e+00 1.93969473e-01 1.10127914e+00 -2.29306072e-02 -3.16166341e-01 3.79396766e-01 1.18648911e+00 4.50438917e-01 7.85205126e-01 1.61508739e-01 7.74357021e-01 5.17440081e-01 8.45000744e-01 5.74523687e-01 3.49422634e-01 7.05719352e-01 1.49684966e-01 1.28605738e-01 -2.23021030e-01 -2.36362219e-01 6.08069241e-01 9.29935157e-01 -3.29311341e-02 -3.80283594e-01 -7.89413571e-01 2.60371149e-01 -2.14794779e+00 -1.22301245e+00 5.13102673e-02 1.98418009e+00 6.06976151e-01 -6.38423720e-03 2.01510504e-01 1.93188459e-01 8.93305540e-01 2.02275813e-01 -4.71122056e-01 8.45925137e-03 -2.42780536e-01 -1.37700781e-01 -9.25910473e-02 1.78449512e-01 -1.33010912e+00 8.30139637e-01 4.77398825e+00 1.13145864e+00 -9.08749163e-01 2.53567070e-01 2.46274173e-01 -2.47820124e-01 1.94754824e-02 -1.79175228e-01 -6.04459524e-01 6.32232904e-01 8.32274914e-01 -1.58526212e-01 1.58996999e-01 9.04938877e-01 -3.43879461e-02 8.57776671e-04 -1.09801424e+00 1.36673141e+00 6.13938510e-01 -1.05765831e+00 2.32389316e-01 -2.98812687e-01 6.62682056e-01 -2.25908607e-01 2.37375516e-02 6.49825037e-01 -1.95718110e-01 -6.75842285e-01 6.33362174e-01 1.07862079e+00 6.12384975e-01 -1.00165427e+00 8.54019523e-01 1.01780295e-01 -1.38235271e+00 -3.30996066e-01 -5.20980060e-01 5.19483626e-01 3.44960213e-01 1.49142131e-01 -4.99349236e-01 7.90555954e-01 5.43518543e-01 1.06888175e+00 -8.15349102e-01 1.32750893e+00 -1.57691807e-01 1.83057889e-01 1.57138884e-01 -2.36801490e-01 3.75091970e-01 1.78176671e-01 3.76087397e-01 1.27572453e+00 3.28426331e-01 7.37168640e-02 2.25119665e-01 5.52938879e-01 -1.22375816e-01 4.09632266e-01 -2.05890849e-01 -1.60386413e-01 3.20896059e-01 1.32687926e+00 -5.58523357e-01 -6.24912620e-01 -5.25878191e-01 1.20355523e+00 1.61459073e-01 3.28264773e-01 -1.03481877e+00 -7.07267404e-01 2.59177446e-01 -3.69562775e-01 4.39197034e-01 1.76107600e-01 1.84343234e-01 -1.26047492e+00 3.77623513e-02 -7.63957679e-01 7.78365910e-01 -9.46186185e-01 -1.60338187e+00 6.99804485e-01 1.71078265e-01 -1.75275052e+00 5.19258762e-03 -5.37204802e-01 -4.57826614e-01 2.94690043e-01 -1.17332673e+00 -1.12236392e+00 -5.98777354e-01 8.26928079e-01 6.58632636e-01 -3.26896548e-01 5.03788292e-01 4.36792582e-01 -6.81734502e-01 6.93657815e-01 2.69153774e-01 1.56851873e-01 9.24932122e-01 -1.01882553e+00 -1.39100388e-01 4.92583305e-01 2.52979964e-01 1.50282472e-01 3.12674791e-01 -5.64808428e-01 -1.09676325e+00 -1.24936330e+00 7.50052512e-01 -2.60822088e-01 5.19826412e-01 1.78554244e-02 -9.92965579e-01 2.23319709e-01 2.80302048e-01 -9.90037993e-03 6.87864304e-01 -2.65329182e-01 -4.55540091e-01 -6.22985102e-02 -7.70255864e-01 5.20734131e-01 1.09928536e+00 -5.76008260e-01 -7.43608713e-01 2.00920895e-01 6.28264129e-01 -7.85765126e-02 -9.68827069e-01 5.40452421e-01 6.99197352e-01 -8.85546923e-01 8.12717497e-01 -6.75136387e-01 5.59608817e-01 -3.83641750e-01 -3.83352488e-01 -1.34397376e+00 -3.15605551e-01 3.40101011e-02 -1.61410660e-01 1.42341089e+00 4.18283999e-01 -2.15372920e-01 4.08684283e-01 1.78694338e-01 -3.05943280e-01 -8.23180497e-01 -8.50425541e-01 -1.02645457e+00 -2.33809575e-01 -5.33436358e-01 5.80796778e-01 9.93057489e-01 3.96870732e-01 4.99472082e-01 -4.73522305e-01 -4.37911779e-01 4.57606256e-01 1.09801002e-01 4.66926634e-01 -8.85146737e-01 -2.02311561e-01 -7.89443970e-01 -5.55219054e-01 -8.00565124e-01 5.10453098e-02 -9.41201389e-01 1.58869863e-01 -1.72213566e+00 6.76294327e-01 1.59815311e-01 -7.01837063e-01 4.89169508e-01 -1.76809490e-01 2.74456471e-01 5.57808697e-01 2.54396349e-01 -1.27864003e+00 1.08027315e+00 1.21307576e+00 -4.68716621e-01 -3.40021014e-01 -3.16522151e-01 -3.43342483e-01 6.91822052e-01 8.14174473e-01 -2.75989920e-01 -5.71526468e-01 -2.15814680e-01 -1.97100401e-01 -1.08358435e-01 3.29850852e-01 -1.16961670e+00 4.72798198e-01 -7.60133788e-02 5.54601669e-01 -9.32115853e-01 4.50616121e-01 -6.93299711e-01 -1.09729446e-01 2.45331526e-01 -6.19153380e-01 -8.26909542e-02 -1.38535649e-01 4.32764828e-01 -3.59790325e-01 -3.45872849e-01 7.58620620e-01 5.39013073e-02 -1.10928893e+00 3.56463701e-01 -1.72759235e-01 6.00996874e-02 1.04198039e+00 -2.34760508e-01 -6.52430832e-01 -1.65423989e-01 -7.08294749e-01 3.44847172e-01 3.15977514e-01 7.16137946e-01 9.48925912e-01 -1.87016082e+00 -6.13943934e-01 -8.09767842e-02 6.41559184e-01 -6.85319364e-01 6.26271367e-01 1.16402042e+00 -1.03691511e-01 3.45223159e-01 -3.01238716e-01 -5.80992997e-01 -1.36041665e+00 9.28498805e-01 3.79544675e-01 -5.24357595e-02 -4.09729749e-01 6.00876093e-01 3.14700641e-02 -2.51461834e-01 5.76844037e-01 -5.82683049e-02 -6.09990060e-01 6.42072558e-01 9.17587280e-01 6.96740627e-01 -1.40896499e-01 -9.99784887e-01 -3.55741471e-01 5.85240483e-01 2.30970117e-03 -2.91064948e-01 9.99278307e-01 -8.79834443e-02 1.66299939e-01 6.36682212e-01 1.48217475e+00 -7.11700797e-01 -1.00324523e+00 -3.83377492e-01 -1.35274425e-01 -3.54999363e-01 -9.18092579e-02 -9.00622189e-01 -1.01301885e+00 1.03101385e+00 9.40892339e-01 -8.89464319e-02 1.31466782e+00 6.90659434e-02 8.24884236e-01 4.88846749e-01 9.74953622e-02 -1.48970938e+00 6.20960891e-01 5.23161948e-01 9.68555272e-01 -1.26645482e+00 -2.16515779e-01 6.54282644e-02 -1.32224953e+00 1.07504010e+00 7.50422120e-01 4.54329729e-01 4.47091728e-01 -1.40134022e-01 -3.43746543e-02 4.51561389e-03 -6.66375339e-01 -5.67109942e-01 6.01124287e-01 3.97996247e-01 1.54260471e-01 -5.68009019e-02 -5.33497393e-01 8.13622296e-01 2.39307746e-01 8.31865966e-02 1.25544563e-01 9.98253882e-01 -5.98669469e-01 -9.69976425e-01 -1.59676164e-01 2.85473436e-01 2.18978934e-02 1.34978220e-02 -5.98241210e-01 6.85548604e-01 3.44846815e-01 1.12778509e+00 4.72077169e-02 -9.23453569e-01 6.22911453e-01 2.49942273e-01 4.09451991e-01 -2.76293963e-01 -4.42221850e-01 6.69345818e-03 6.56222478e-02 -6.47482932e-01 -6.05424464e-01 -5.71437359e-01 -1.27401865e+00 -5.96841685e-02 -3.06210518e-01 1.06077366e-01 4.64690685e-01 9.37641203e-01 3.02586913e-01 4.18080240e-01 7.04627514e-01 -7.13126659e-01 -4.92093056e-01 -1.10375154e+00 -5.39083898e-01 7.44683743e-01 -2.09957719e-01 -8.62105846e-01 -3.62955272e-01 -6.55347202e-03]
[8.622791290283203, 0.7958241105079651]
270977e9-0fd2-4ed3-ad14-490b11eea889
visual-enhanced-3d-point-cloud-reconstruction
2108.07685
null
https://arxiv.org/abs/2108.07685v1
https://arxiv.org/pdf/2108.07685v1.pdf
Visual Enhanced 3D Point Cloud Reconstruction from A Single Image
Solving the challenging problem of 3D object reconstruction from a single image appropriately gives existing technologies the ability to perform with a single monocular camera rather than requiring depth sensors. In recent years, thanks to the development of deep learning, 3D reconstruction of a single image has demonstrated impressive progress. Existing researches use Chamfer distance as a loss function to guide the training of the neural network. However, the Chamfer loss will give equal weights to all points inside the 3D point clouds. It tends to sacrifice fine-grained and thin structures to avoid incurring a high loss, which will lead to visually unsatisfactory results. This paper proposes a framework that can recover a detailed three-dimensional point cloud from a single image by focusing more on boundaries (edge and corner points). Experimental results demonstrate that the proposed method outperforms existing techniques significantly, both qualitatively and quantitatively, and has fewer training parameters.
['Han Wang', 'Mahdi Abolfazli Esfahani', 'Guiju Ping']
2021-08-17
null
null
null
null
['3d-point-cloud-reconstruction', 'point-cloud-reconstruction', '3d-object-reconstruction', '3d-object-reconstruction-from-a-single-image', 'object-reconstruction']
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
[-1.40897874e-02 -7.21731782e-02 -2.72161309e-02 -3.74406606e-01 -4.13395166e-01 -1.99073419e-01 4.51396972e-01 -5.30255167e-03 -3.89539599e-01 3.37426454e-01 -4.09079373e-01 -1.05654575e-01 6.72314391e-02 -8.48459721e-01 -9.13452804e-01 -4.13705230e-01 -1.91182103e-02 6.29473865e-01 3.92110199e-01 2.01347202e-01 3.07984680e-01 1.18363702e+00 -1.65046644e+00 -2.61775758e-02 7.03203559e-01 1.34777820e+00 3.90144259e-01 5.99762499e-02 -1.90059900e-01 3.25467944e-01 -3.55678916e-01 -3.65316659e-01 6.05377614e-01 2.96437562e-01 -4.43052083e-01 2.99867332e-01 7.56987929e-01 -8.10719848e-01 -3.77854139e-01 1.15714073e+00 2.72718638e-01 -1.66768432e-01 5.75621307e-01 -1.10548520e+00 -4.76686537e-01 3.55628971e-03 -9.58619833e-01 -2.82699943e-01 2.15902135e-01 -1.43549610e-02 7.66826153e-01 -1.13931489e+00 4.57511991e-01 1.19563079e+00 8.94683421e-01 3.67472351e-01 -9.39083874e-01 -6.18494987e-01 -5.21964394e-03 7.54122138e-02 -1.58058429e+00 -7.34047070e-02 1.26257527e+00 -3.07384014e-01 8.87991726e-01 -8.62748595e-04 8.55396569e-01 6.57232106e-01 4.20425534e-02 9.34538126e-01 9.65510607e-01 -2.55674034e-01 2.53276397e-02 -1.86177328e-01 -1.70002505e-01 7.36097753e-01 4.12418514e-01 4.27463591e-01 -2.38603145e-01 7.10422695e-02 1.28427553e+00 4.01484251e-01 -3.33266228e-01 -9.42062616e-01 -9.47927415e-01 6.76334083e-01 8.98268044e-01 1.13672309e-01 -3.73457938e-01 7.09264949e-02 9.68051329e-03 2.51607206e-02 4.46432263e-01 7.20601007e-02 -2.48030409e-01 1.08701348e-01 -8.91932189e-01 1.71757147e-01 4.08048719e-01 1.06987846e+00 9.98461306e-01 -6.55216426e-02 4.92669195e-01 6.10348642e-01 4.50673491e-01 3.83974135e-01 -2.39338130e-02 -1.14202166e+00 4.09213632e-01 1.04252470e+00 3.42452586e-01 -1.21086049e+00 -5.22873998e-01 -3.26120168e-01 -1.02771986e+00 7.92064369e-01 3.28590155e-01 1.75558209e-01 -1.21412599e+00 1.11501896e+00 5.46639204e-01 6.79825395e-02 -4.23641324e-01 1.23786545e+00 7.39473403e-01 4.57195967e-01 -5.41228116e-01 1.71156779e-01 7.59621441e-01 -6.42629743e-01 -3.65502447e-01 -3.67252678e-01 1.15403369e-01 -6.27731383e-01 1.02609849e+00 4.32613641e-01 -1.37790632e+00 -5.28999805e-01 -1.31487370e+00 -3.13420177e-01 -1.73476189e-01 -1.14262387e-01 6.26875460e-01 2.63605118e-01 -8.47494483e-01 6.98302209e-01 -1.00641775e+00 3.85560542e-02 8.30768704e-01 5.01526713e-01 -5.20082951e-01 -3.80400926e-01 -6.01261258e-01 8.61204803e-01 2.85306036e-01 3.15357685e-01 -6.28102899e-01 -8.96131635e-01 -7.99462259e-01 1.27277635e-02 2.33235925e-01 -7.55392194e-01 9.58120286e-01 -4.14772332e-01 -1.36183155e+00 1.16446388e+00 8.17473903e-02 -2.64338762e-01 7.12960064e-01 -4.26310301e-01 2.92670369e-01 2.20362753e-01 -1.99522853e-01 8.41546476e-01 7.70915329e-01 -1.71788609e+00 -6.52692735e-01 -7.17966497e-01 2.21900985e-01 2.74693847e-01 -4.09523174e-02 -3.18172723e-01 -5.30881882e-01 -2.46995285e-01 7.88428545e-01 -6.86425745e-01 -1.91857085e-01 6.26676679e-01 -2.85233617e-01 -2.01169476e-01 1.05835009e+00 -2.94476092e-01 5.30060649e-01 -2.10120988e+00 7.47928396e-02 7.66201913e-02 3.40417653e-01 2.38686368e-01 1.92737564e-01 -2.06217207e-02 1.39741868e-01 5.35665601e-02 -3.73448998e-01 -4.93656307e-01 -1.55134007e-01 1.72794506e-01 1.33396238e-01 7.11536705e-01 1.02498703e-01 8.86054695e-01 -6.76087558e-01 -4.50507820e-01 6.69830084e-01 7.50373960e-01 -4.43679720e-01 1.71331868e-01 -7.29950666e-02 1.06292069e-01 -4.19658929e-01 8.56528997e-01 1.29421210e+00 -4.56257850e-01 -4.16524380e-01 -2.90992528e-01 -9.59667489e-02 -5.90397380e-02 -1.10676324e+00 1.99390996e+00 -4.21326667e-01 4.52935755e-01 3.62751901e-01 -8.55653107e-01 1.13348472e+00 3.45067233e-02 6.87706292e-01 -5.63908994e-01 2.67747641e-01 2.16855511e-01 -3.39148432e-01 -2.22992316e-01 7.90801197e-02 -3.86116147e-01 2.14070529e-01 1.35241941e-01 -2.80985594e-01 -5.25502086e-01 -3.67503226e-01 -2.14447424e-01 6.86013460e-01 2.35222578e-01 1.15327232e-01 1.95739910e-01 4.41586912e-01 9.03839171e-02 6.05636358e-01 3.51131737e-01 -1.66015223e-01 9.71828043e-01 1.22617334e-02 -7.42591560e-01 -1.19562757e+00 -1.25040925e+00 -3.64753813e-01 -3.09131574e-02 7.04342306e-01 1.41854569e-01 -3.99535805e-01 -4.91805851e-01 2.90739417e-01 3.30219209e-01 -3.57903063e-01 -7.97857940e-02 -7.26015866e-01 -2.83171475e-01 9.27759632e-02 6.20339513e-01 7.26884425e-01 -8.91193986e-01 -9.48548973e-01 -1.49963815e-02 6.47939071e-02 -1.13385332e+00 -1.64480850e-01 1.79459125e-01 -1.37945700e+00 -1.02965200e+00 -9.46159065e-01 -8.99324715e-01 9.20842469e-01 6.10217273e-01 1.21878791e+00 2.74305910e-01 -3.29933137e-01 4.36206274e-02 -1.88103393e-01 -3.04885864e-01 6.32133707e-02 -1.69327885e-01 -1.54665127e-01 -3.12311113e-01 4.97412920e-01 -7.50739932e-01 -7.36111999e-01 3.29150915e-01 -8.19615304e-01 1.37103021e-01 8.58101010e-01 7.91106164e-01 7.08778322e-01 4.73422371e-02 1.45030111e-01 -3.82389039e-01 4.48913686e-02 -1.23876013e-01 -9.22261000e-01 -7.01934025e-02 -6.31743014e-01 -3.42844367e-01 4.62209791e-01 -1.74713746e-01 -9.78606999e-01 4.08267498e-01 -8.25910196e-02 -9.66586351e-01 -3.16220731e-01 2.67406255e-01 -1.19387254e-01 -4.22791958e-01 3.43158007e-01 2.46320397e-01 1.57990709e-01 -7.26464689e-01 1.07193552e-01 6.47940516e-01 3.75467122e-01 -2.19050407e-01 9.53582764e-01 9.47589338e-01 1.88701108e-01 -7.76619494e-01 -8.20160389e-01 -4.66757923e-01 -1.01720905e+00 -3.41924310e-01 5.89708626e-01 -9.27004814e-01 -8.81972313e-01 5.13939917e-01 -1.27303863e+00 1.76517814e-02 -1.92370653e-01 4.48512971e-01 -6.89188302e-01 5.28594017e-01 -5.96521676e-01 -5.76421261e-01 -2.41632491e-01 -9.92873549e-01 1.23996186e+00 3.19696128e-01 1.38143197e-01 -7.28283465e-01 -2.31350571e-01 5.25864720e-01 -1.24675790e-02 4.33786184e-01 8.36448193e-01 2.13795155e-01 -1.14458919e+00 -5.54932117e-01 -4.97696638e-01 4.52559680e-01 9.04945880e-02 -1.17466018e-01 -9.05073464e-01 -1.93296522e-01 3.40962678e-01 -3.11970055e-01 6.19631171e-01 5.62105775e-01 1.31884336e+00 -4.64730971e-02 -3.76471549e-01 1.05847526e+00 1.65866625e+00 2.54471511e-01 6.28710151e-01 4.23070967e-01 8.35287690e-01 4.79685336e-01 5.46325326e-01 2.18888745e-01 2.95962602e-01 4.75447923e-01 1.09885490e+00 -2.52562881e-01 -1.58223882e-01 -3.76876742e-01 -1.48421809e-01 5.49249947e-01 -1.68136477e-01 1.80564355e-02 -9.91146863e-01 5.27149439e-01 -1.53728604e+00 -6.82534635e-01 -6.17618039e-02 2.31720352e+00 5.64506173e-01 2.08346337e-01 -2.50642687e-01 3.05698276e-01 5.25081277e-01 -2.79545020e-02 -7.79375434e-01 -3.37369218e-02 -1.22792818e-01 -4.51426432e-02 4.11543906e-01 5.53599834e-01 -1.07501209e+00 7.70406127e-01 6.52390623e+00 6.23882055e-01 -1.18520522e+00 -2.89004743e-01 3.63341182e-01 -1.16210684e-01 -1.76357865e-01 -1.51079580e-01 -7.41992772e-01 3.77177358e-01 1.93411976e-01 -8.44074413e-02 1.65275156e-01 8.13895106e-01 2.56405398e-02 -1.84664816e-01 -1.17977428e+00 1.32814562e+00 9.83525589e-02 -1.42309093e+00 1.91103607e-01 1.82720840e-01 7.93171346e-01 2.34238133e-01 -4.93990444e-03 -1.77721769e-01 -1.57715436e-02 -1.13905466e+00 7.05693603e-01 4.35643733e-01 6.43248916e-01 -9.28882718e-01 7.59516835e-01 7.83348620e-01 -9.11977887e-01 -6.21846244e-02 -7.12980330e-01 -1.07740536e-01 1.90682054e-01 7.42911816e-01 -7.77533174e-01 5.29050946e-01 1.01428545e+00 8.17550600e-01 -2.21266389e-01 1.48400879e+00 -7.94818345e-03 -1.37372181e-01 -6.76778913e-01 1.45383567e-01 2.59338349e-01 -4.16311115e-01 5.35832405e-01 5.46936095e-01 4.00683910e-01 1.18364215e-01 1.77338824e-01 9.88120735e-01 -1.87928930e-01 -1.60414740e-01 -8.34101737e-01 3.97817552e-01 4.08747673e-01 1.01548791e+00 -7.74907589e-01 -6.81777149e-02 -6.05794847e-01 8.78091037e-01 3.93241584e-01 1.11990742e-01 -4.38951999e-01 -2.21943662e-01 4.67863292e-01 5.43948352e-01 3.82791370e-01 -4.73989129e-01 -7.98809350e-01 -9.13974822e-01 2.72546053e-01 -3.55603725e-01 -8.81515816e-02 -9.20662403e-01 -1.37571931e+00 4.92543101e-01 5.94580211e-02 -1.52479339e+00 1.08509317e-01 -7.16561317e-01 -4.03168201e-01 8.40850890e-01 -1.74111164e+00 -9.14998770e-01 -5.79549491e-01 4.57135916e-01 3.70852053e-01 1.96768090e-01 4.56635773e-01 4.46989000e-01 -1.78912625e-01 2.40806773e-01 -1.35078445e-01 4.85296249e-02 3.45104694e-01 -1.02358854e+00 2.13990539e-01 4.04159129e-01 -6.24658167e-02 4.33894277e-01 4.00699496e-01 -4.99884248e-01 -1.37042320e+00 -7.28179455e-01 6.22702420e-01 -3.64005655e-01 1.71935320e-01 -1.90798506e-01 -1.06324148e+00 4.43371147e-01 -1.12615071e-01 2.54936248e-01 1.49415016e-01 -9.51361284e-02 -1.00760393e-01 -1.51817039e-01 -1.41239595e+00 3.36708456e-01 1.10583246e+00 -3.20520550e-01 -5.61391354e-01 1.19585119e-01 4.63735342e-01 -6.27456248e-01 -9.20145869e-01 7.44085550e-01 5.38162529e-01 -1.31172335e+00 1.36482835e+00 -7.57782459e-02 4.63590324e-01 -4.86306548e-01 -6.18500262e-02 -1.08502507e+00 -1.95293918e-01 -2.86057234e-01 -2.92354763e-01 6.89127266e-01 1.14024421e-02 -1.71769813e-01 1.39082301e+00 5.81422389e-01 -2.77807772e-01 -9.98124123e-01 -1.11525059e+00 -8.67368758e-01 2.19408095e-01 -1.86877340e-01 7.24899888e-01 6.72768891e-01 -4.51098412e-01 1.57077581e-01 -1.52939603e-01 3.67010981e-01 1.07370198e+00 4.25537437e-01 6.74888551e-01 -1.72514987e+00 2.64413178e-01 -5.22049546e-01 -6.65437341e-01 -1.53951752e+00 -6.88578486e-02 -7.75221944e-01 1.50053367e-01 -1.65082967e+00 4.34341393e-02 -7.33086884e-01 3.93374041e-02 2.71381170e-01 1.62219077e-01 4.89326090e-01 1.17009565e-01 2.54873633e-01 -1.41994223e-01 7.84525573e-01 1.63205922e+00 -1.62390620e-01 -1.10379800e-01 1.31129459e-01 -3.32846910e-01 1.15321708e+00 5.96835136e-01 -3.94285917e-01 -2.45342672e-01 -9.11520243e-01 2.43032992e-01 7.75137097e-02 6.13719642e-01 -1.09126699e+00 3.59447420e-01 -7.39960000e-02 8.21628153e-01 -1.26924646e+00 7.24944592e-01 -1.34291244e+00 4.68078740e-02 3.12000424e-01 1.77969411e-01 -2.24529684e-01 1.39439732e-01 4.82539564e-01 -3.60732794e-01 -2.66791195e-01 1.06568551e+00 -2.87678540e-01 -5.69433510e-01 7.24225104e-01 3.24064910e-01 -2.69541711e-01 1.14399767e+00 -8.63231242e-01 1.54973775e-01 -1.49730712e-01 -5.20974517e-01 1.89137012e-01 1.01477623e+00 3.01078975e-01 1.15657091e+00 -1.34791946e+00 -4.32925820e-01 4.59292173e-01 -9.20080021e-02 8.29770923e-01 3.63603085e-01 3.42763603e-01 -9.44902778e-01 3.42841744e-01 -4.31691796e-01 -1.07153451e+00 -1.03762639e+00 5.26492238e-01 5.34886837e-01 1.54414460e-01 -1.14087403e+00 9.48080182e-01 1.67792276e-01 -5.09620130e-01 5.39473236e-01 -3.22200656e-01 -4.71753702e-02 -2.58077264e-01 2.07030743e-01 2.61644870e-01 1.59219295e-01 -5.60269952e-01 -2.98313677e-01 1.18922591e+00 -1.31181732e-01 2.74720758e-01 1.57273853e+00 -1.23392209e-01 1.62819222e-01 3.10704768e-01 1.32496560e+00 -4.22326863e-01 -1.86743367e+00 -2.11685792e-01 -2.62686849e-01 -9.27494526e-01 5.03395855e-01 -5.30532479e-01 -1.31236827e+00 1.24120498e+00 6.01635993e-01 8.76022577e-02 8.97582650e-01 -7.56743085e-03 9.81272161e-01 3.42757970e-01 6.12405658e-01 -7.60087311e-01 -2.38984413e-02 4.26731348e-01 9.05817866e-01 -1.44509029e+00 3.01818132e-01 -4.77150559e-01 -1.39841527e-01 1.14522994e+00 6.91642106e-01 -5.17114639e-01 9.31765079e-01 3.04997861e-01 -6.31640330e-02 -3.52835506e-01 -3.14508885e-01 1.29148522e-02 3.41044694e-01 7.57685781e-01 8.46199617e-02 -2.23664403e-01 1.06879935e-01 -5.62492898e-03 -6.90610707e-02 8.63149092e-02 4.03809428e-01 9.30144906e-01 -5.30704796e-01 -8.96765411e-01 -4.37435389e-01 4.39855516e-01 -2.31476426e-01 3.34536642e-01 -3.10506880e-01 9.48905289e-01 2.34542955e-02 5.23390055e-01 4.59820777e-01 -2.58554608e-01 4.93266165e-01 -3.58493626e-01 7.56123066e-01 -4.79631454e-01 -9.84406471e-02 2.92785969e-02 -4.62580293e-01 -6.51713192e-01 -5.20559967e-01 -5.36438763e-01 -1.34400880e+00 -3.98791164e-01 -2.66352355e-01 -1.72136903e-01 7.37180114e-01 6.59765601e-01 3.35171819e-01 1.16667449e-01 8.23344350e-01 -1.47011101e+00 -6.39253378e-01 -6.29316032e-01 -6.35085881e-01 2.39725187e-01 5.40547907e-01 -9.65992868e-01 -4.92302984e-01 -1.61752120e-01]
[8.428468704223633, -3.1254990100860596]
557835d8-fcc8-4d2c-86d8-002067b56dca
astronomical-image-time-series-classification
2304.01236
null
https://arxiv.org/abs/2304.01236v1
https://arxiv.org/pdf/2304.01236v1.pdf
Astronomical image time series classification using CONVolutional attENTION (ConvEntion)
Aims. The treatment of astronomical image time series has won increasing attention in recent years. Indeed, numerous surveys following up on transient objects are in progress or under construction, such as the Vera Rubin Observatory Legacy Survey for Space and Time (LSST), which is poised to produce huge amounts of these time series. The associated scientific topics are extensive, ranging from the study of objects in our galaxy to the observation of the most distant supernovae for measuring the expansion of the universe. With such a large amount of data available, the need for robust automatic tools to detect and classify celestial objects is growing steadily. Methods. This study is based on the assumption that astronomical images contain more information than light curves. In this paper, we propose a novel approach based on deep learning for classifying different types of space objects directly using images. We named our approach ConvEntion, which stands for CONVolutional attENTION. It is based on convolutions and transformers, which are new approaches for the treatment of astronomical image time series. Our solution integrates spatio-temporal features and can be applied to various types of image datasets with any number of bands. Results. In this work, we solved various problems the datasets tend to suffer from and we present new results for classifications using astronomical image time series with an increase in accuracy of 13%, compared to state-of-the-art approaches that use image time series, and a 12% increase, compared to approaches that use light curves.
['Julian Bautista', 'Frédéric Comby', 'Jerome Paquet', 'Dominique Fouchez', 'Marc Chaumont', 'Anass Bairouk']
2023-04-03
null
null
null
null
['time-series-classification']
['time-series']
[-1.13103837e-01 -7.35798120e-01 2.54519075e-01 -5.41642308e-02 6.98256269e-02 -6.19830728e-01 1.18811154e+00 -1.20049767e-01 -7.68948019e-01 3.93495142e-01 -3.65354955e-01 -4.90595490e-01 -3.53825182e-01 -9.41214859e-01 -3.32825601e-01 -9.57402885e-01 -2.61157393e-01 4.28781450e-01 3.91280562e-01 -4.00334388e-01 3.71072590e-01 8.59214664e-01 -1.73915565e+00 9.97307673e-02 3.09177250e-01 1.49450862e+00 2.71332502e-01 6.51969373e-01 -3.06430519e-01 6.95265234e-01 -4.70586687e-01 4.52360250e-02 4.17106122e-01 -2.82879263e-01 -6.58529162e-01 -1.33344918e-01 2.62583971e-01 9.30948555e-02 -6.30110860e-01 9.04141784e-01 2.80017525e-01 4.13789213e-01 5.65048993e-01 -1.14244151e+00 -4.87628400e-01 -8.63315091e-02 -5.00684857e-01 9.35863256e-01 -1.07525133e-01 -3.31845507e-03 6.82044983e-01 -5.46407402e-01 5.01647294e-01 8.74979138e-01 7.31433868e-01 8.15028325e-03 -8.32255721e-01 -2.82674164e-01 -4.11256850e-01 8.75985742e-01 -8.96556795e-01 3.25848870e-02 7.67434001e-01 -8.17991972e-01 9.17409062e-01 4.72637415e-01 8.20790589e-01 9.01545167e-01 2.01530799e-01 4.91576552e-01 1.16222095e+00 -5.67662179e-01 1.52831033e-01 -4.52731289e-02 1.59266889e-01 3.77656490e-01 -9.35958326e-02 4.86066967e-01 6.36596233e-03 8.44147056e-02 7.42106915e-01 2.28978649e-01 -1.62152663e-01 -2.80457109e-01 -1.36644077e+00 8.36615980e-01 4.89177644e-01 1.21559405e+00 -2.47622326e-01 -1.05208475e-02 5.74192822e-01 5.06497622e-01 6.55497253e-01 6.25964046e-01 -4.22072262e-01 -2.14758798e-01 -9.35658455e-01 4.33338225e-01 4.34706926e-01 2.48423561e-01 2.80403912e-01 6.84440136e-02 1.87542841e-01 6.45160556e-01 -1.72524020e-01 5.27541637e-01 1.00115430e+00 -5.63813925e-01 -8.19612890e-02 5.10431349e-01 1.35532200e-01 -8.95115197e-01 -8.22739780e-01 -4.75817025e-01 -9.05756474e-01 5.27465224e-01 6.67627037e-01 2.34953597e-01 -9.28980768e-01 1.24493253e+00 2.29068711e-01 2.73808658e-01 -1.40967995e-01 9.71107781e-01 6.52244568e-01 1.01778686e+00 -3.28941166e-01 -2.07887813e-01 1.53692389e+00 -7.27513731e-01 -4.94471192e-01 1.06488153e-01 2.23076999e-01 -8.50926340e-01 7.51024425e-01 4.76283759e-01 -6.53372467e-01 -7.46892631e-01 -9.71325159e-01 8.33676755e-02 -7.35370576e-01 6.49246573e-02 8.67872000e-01 4.98526990e-01 -9.20380652e-01 9.14060473e-01 -8.64365876e-01 -5.56534231e-01 2.07125738e-01 5.21957614e-02 -2.66878933e-01 4.47676659e-01 -9.95851219e-01 1.01785874e+00 4.04659659e-01 -1.41382543e-02 -6.36464715e-01 -4.78732407e-01 -4.66906309e-01 1.25552148e-01 1.74096420e-01 -2.04692230e-01 1.24435461e+00 -1.06210518e+00 -1.18486619e+00 9.63774621e-01 3.00194532e-01 -7.97531664e-01 4.56344157e-01 1.93368614e-01 -7.66665995e-01 8.54810700e-02 -1.90587491e-01 1.02482781e-01 7.91761816e-01 -5.64886510e-01 -7.41175592e-01 -4.40727860e-01 -4.06999290e-02 -3.81668985e-01 -4.20618296e-01 2.87383616e-01 -2.22736016e-01 -7.53862798e-01 2.54934609e-01 -1.06833994e+00 -1.66167364e-01 -8.09901133e-02 2.34303594e-01 -5.17535210e-01 1.13578475e+00 -6.43877447e-01 8.19292307e-01 -2.19868040e+00 1.54987067e-01 -1.37513634e-02 8.85504559e-02 4.91210192e-01 1.22251667e-01 2.36898825e-01 -4.66743052e-01 -2.61877030e-01 -3.67995530e-01 -2.11383346e-02 -1.82102188e-01 1.74653437e-02 -6.87756777e-01 7.41305351e-01 -2.80819088e-01 5.70798874e-01 -7.06210554e-01 -5.17195947e-02 6.32958651e-01 4.65187162e-01 -1.31577060e-01 2.82753985e-02 -2.08071604e-01 6.47926092e-01 -3.02602917e-01 2.14930519e-01 5.52971363e-01 -2.65779942e-01 -4.01073843e-01 -8.45825970e-02 -9.15574610e-01 6.09455667e-02 -6.33011043e-01 1.50185788e+00 -5.02039015e-01 1.11111605e+00 -4.07255977e-01 -1.43422651e+00 9.74491060e-01 4.66481537e-01 8.43200564e-01 -1.16847122e+00 4.40151900e-01 3.89836401e-01 1.97458774e-01 -7.92335331e-01 3.95540863e-01 -4.12149638e-01 2.51873314e-01 3.27798218e-01 7.62138590e-02 -2.23408163e-01 3.52838218e-01 -2.40785524e-01 9.68808234e-01 1.68178350e-01 1.74412712e-01 -3.25154662e-01 7.22744882e-01 1.24313377e-01 9.25597623e-02 3.94216776e-01 -2.25235030e-01 4.59561646e-01 2.49001354e-01 -1.25575960e+00 -1.51792228e+00 -6.01526856e-01 -5.06211340e-01 6.76169693e-01 -9.72164273e-02 9.38074589e-02 -4.10864711e-01 -2.21792892e-01 -1.94930717e-01 5.11478662e-01 -7.93715596e-01 1.50902539e-01 -6.53815925e-01 -1.03613555e+00 2.61432499e-01 2.82109261e-01 6.89793348e-01 -1.40949380e+00 -1.06718349e+00 1.88388720e-01 -8.89286771e-02 -9.33031559e-01 1.46667361e-01 2.51438003e-02 -9.43335831e-01 -1.11255801e+00 -9.58546042e-01 -6.02151930e-01 8.07484835e-02 3.51704568e-01 8.62230361e-01 -9.00254995e-02 -7.66933143e-01 2.76544511e-01 -3.92654687e-01 -5.50895810e-01 -1.78469419e-01 -1.29588112e-01 -5.35651576e-03 3.50522995e-01 2.92386144e-01 -6.42089963e-01 -4.75318879e-01 2.28845909e-01 -9.62216914e-01 -1.30016491e-01 2.99230844e-01 5.70509970e-01 2.29271457e-01 2.89537340e-01 2.32056722e-01 -3.61287594e-01 6.16564117e-02 -5.98681867e-01 -1.22328150e+00 1.72871411e-01 -3.66011322e-01 9.55577940e-02 8.57081950e-01 -3.52402449e-01 -1.04690957e+00 -3.08999509e-01 -2.28129163e-01 -5.73252618e-01 -2.47719690e-01 2.21713424e-01 7.50661552e-01 -2.13541612e-01 8.71592283e-01 3.57645333e-01 -1.08809642e-01 -7.12099314e-01 1.14994638e-01 5.69317579e-01 9.09490883e-01 -2.62029052e-01 8.32543314e-01 9.16557014e-01 2.33432382e-01 -1.08225667e+00 -7.23535538e-01 -9.38938558e-01 -4.59195822e-01 -4.97683674e-01 9.82648909e-01 -3.40295970e-01 -7.95836926e-01 6.83113813e-01 -1.11536014e+00 -5.67300171e-02 -2.19899088e-01 7.66651213e-01 -4.99747008e-01 3.54907811e-01 -4.75252509e-01 -8.40430140e-01 -1.24096543e-01 -7.62330532e-01 6.66747510e-01 1.71345875e-01 2.96506912e-01 -1.17553997e+00 4.92197573e-01 5.52000180e-02 7.31645584e-01 4.69754964e-01 6.67591989e-01 -4.57261652e-01 -5.74500740e-01 -2.88208157e-01 -2.77399361e-01 2.22147822e-01 -1.01466492e-01 1.03393998e-02 -1.03706241e+00 -4.09921080e-01 7.50327826e-01 -5.47356568e-02 1.00431192e+00 5.25006115e-01 1.31677568e+00 1.34334967e-01 -2.60612935e-01 7.21452594e-01 1.49636149e+00 6.22041881e-01 7.79034376e-01 8.42321992e-01 3.31520259e-01 6.25780344e-01 5.24062037e-01 5.29028893e-01 -2.58701026e-01 9.34784234e-01 5.69781065e-01 6.21729121e-02 2.50311978e-02 4.79586363e-01 -2.76150238e-02 7.13850856e-01 -6.95721626e-01 -5.27956039e-02 -1.02785313e+00 7.43299425e-01 -1.68174744e+00 -1.43157411e+00 -5.04044354e-01 2.30623651e+00 1.85116008e-01 3.29412855e-02 5.47117852e-02 3.36304218e-01 5.37119031e-01 3.02735925e-01 -2.26041391e-01 -9.84038189e-02 -9.87297893e-02 2.71868974e-01 4.34307426e-01 8.99850205e-02 -1.36958504e+00 3.25847059e-01 5.83016491e+00 6.94757760e-01 -1.74587858e+00 2.63063073e-01 4.03535485e-01 7.50888735e-02 -2.18435414e-02 -8.99910033e-02 -2.62671232e-01 6.37077510e-01 9.88512516e-01 -2.98089534e-01 5.47828257e-01 8.02018285e-01 4.75408062e-02 -2.18786094e-02 -6.07995272e-01 1.38881874e+00 5.43814003e-02 -1.41527188e+00 -5.40111601e-01 -3.73053141e-02 7.09949732e-01 4.70209211e-01 2.79481970e-02 2.07963601e-01 -2.52012402e-01 -6.33507371e-01 8.96313667e-01 7.05541432e-01 4.72230375e-01 -4.55623925e-01 6.70254529e-01 2.44016722e-01 -1.12734628e+00 -2.36584902e-01 -4.84598190e-01 -2.52846390e-01 6.37309477e-02 7.53379524e-01 -5.30625224e-01 6.60280406e-01 9.85126078e-01 6.58629656e-01 -5.33379197e-01 1.56189764e+00 2.53600717e-01 5.52654147e-01 -4.04770702e-01 -1.50629461e-01 5.45743048e-01 -4.10890579e-01 5.54541886e-01 7.73268998e-01 7.72178233e-01 8.68050382e-02 -1.13125123e-01 7.09292471e-01 3.53352726e-01 5.49323075e-02 -6.61368668e-01 -5.98051548e-02 -2.15730727e-01 1.42004597e+00 -1.03598213e+00 -5.34810781e-01 -6.13385916e-01 6.73725009e-01 -3.04795485e-02 1.98628247e-01 -9.36280727e-01 -3.03001970e-01 2.67186016e-01 1.06710933e-01 3.67234617e-01 -5.39873898e-01 -1.39094293e-01 -1.11809075e+00 -1.22079439e-01 -3.07228327e-01 5.34815669e-01 -9.17260289e-01 -1.13385081e+00 8.72559071e-01 1.01646464e-02 -1.48994863e+00 -1.38437837e-01 -8.89915526e-01 -6.92766726e-01 6.38515770e-01 -1.56320298e+00 -6.94520772e-01 -4.35226500e-01 5.43404639e-01 5.46452761e-01 -3.82283270e-01 4.94841158e-01 5.50016880e-01 -1.90901130e-01 -3.34529728e-01 5.07716238e-01 -8.81993920e-02 4.19798225e-01 -1.08563232e+00 5.41277170e-01 7.65548229e-01 3.70188862e-01 1.72191456e-01 8.61928582e-01 -1.18230112e-01 -1.06678021e+00 -7.64986455e-01 1.01505613e+00 -2.51052350e-01 8.56838524e-01 -1.19923547e-01 -9.49181259e-01 5.24099290e-01 2.48852849e-01 3.07206064e-01 1.76485926e-01 -1.76344141e-01 -3.03599387e-01 -1.70203000e-01 -8.89489591e-01 2.25294843e-01 5.98973632e-01 -4.11260217e-01 -7.09038913e-01 6.37970805e-01 3.15709949e-01 -4.63569239e-02 -4.88961458e-01 3.72980207e-01 3.42325717e-01 -1.30870330e+00 1.01187265e+00 -3.03966165e-01 1.94090590e-01 -4.47755218e-01 2.18518808e-01 -1.12007654e+00 -4.94217873e-01 -2.80138135e-01 1.71437830e-01 4.88223046e-01 -2.26434737e-01 -6.54437661e-01 4.19737101e-01 -9.88550708e-02 -1.95394084e-01 -1.64189607e-01 -1.06445920e+00 -1.06778729e+00 -1.02346689e-01 -4.62094158e-01 4.05462533e-01 1.28371775e+00 -2.79138893e-01 7.95915052e-02 -3.27408195e-01 -1.05752880e-02 5.99114835e-01 5.42497814e-01 4.19619322e-01 -1.54843438e+00 -4.12725329e-01 -7.45058954e-01 -6.92693233e-01 -5.63818395e-01 -9.15326253e-02 -8.62077951e-01 -2.93319046e-01 -1.22519994e+00 -1.94368213e-01 -4.52634364e-01 -3.93022776e-01 3.22022945e-01 3.15779537e-01 5.38377106e-01 7.35363886e-02 4.77376312e-01 -2.60444254e-01 3.70792240e-01 1.06289637e+00 -2.11175501e-01 2.35837325e-01 5.13661876e-02 3.76864702e-01 6.29107833e-01 8.28945875e-01 -1.40225559e-01 -1.67838290e-01 -1.93864092e-01 2.79528528e-01 1.57366961e-01 6.86810613e-01 -1.44433224e+00 1.38185307e-01 8.51485133e-03 3.57634634e-01 -6.56293511e-01 2.26560682e-01 -8.04792762e-01 4.57286119e-01 6.65915787e-01 9.57588330e-02 4.48116064e-02 2.59740531e-01 2.68685132e-01 -5.24155378e-01 -3.50662798e-01 1.14031625e+00 -3.31323594e-01 -1.00226450e+00 3.13525259e-01 -3.19458634e-01 -3.72473657e-01 1.18292260e+00 1.34136245e-01 -3.13497752e-01 -2.20252469e-01 -6.75082445e-01 -3.05780530e-01 2.04425365e-01 6.56504035e-01 1.19809441e-01 -1.32818079e+00 -5.29762566e-01 2.66663700e-01 2.65311748e-02 -4.72007632e-01 4.49369520e-01 1.08462620e+00 -8.67302537e-01 8.17169070e-01 -6.12903535e-01 -5.70017695e-01 -1.08696175e+00 1.08652163e+00 5.38994014e-01 -1.98281363e-01 -8.58085811e-01 5.65228760e-01 2.19965041e-01 -1.51615918e-01 -1.53099641e-01 -4.75894541e-01 -6.44507766e-01 2.81511724e-01 6.67722166e-01 4.10276324e-01 3.50398540e-01 -5.89275777e-01 -7.65992925e-02 5.50384939e-01 2.01287895e-01 1.34263309e-02 1.70394766e+00 2.47971013e-01 -4.50801700e-01 7.20020354e-01 1.20131052e+00 -4.10495013e-01 -1.01227355e+00 -2.92925596e-01 1.93586722e-01 -5.96186399e-01 2.20984280e-01 -5.51529586e-01 -1.07855773e+00 1.02267826e+00 9.65142906e-01 9.46274936e-01 1.12365317e+00 -2.47262847e-02 6.90074980e-01 3.24306637e-01 6.11660779e-01 -7.89622486e-01 -5.67093417e-02 5.72205603e-01 1.06486952e+00 -1.08442914e+00 -1.47687629e-01 2.82081496e-02 2.09969864e-03 1.55105805e+00 1.19172581e-01 -1.66393563e-01 6.73449755e-01 9.41883847e-02 -4.97995354e-02 -3.23177546e-01 -4.89945769e-01 -3.00777137e-01 3.21652144e-01 1.50352761e-01 3.41432273e-01 -2.15782505e-02 -5.45567036e-01 -4.22675312e-02 -2.48791188e-01 -3.25857103e-02 3.85633260e-01 6.50568366e-01 -6.04216337e-01 -1.00144637e+00 -8.66839230e-01 3.03907096e-01 -5.40311217e-01 3.26711033e-03 1.85197905e-01 6.48424983e-01 2.15642676e-01 5.25086939e-01 4.34302300e-01 2.14650556e-02 1.69770315e-01 1.19662650e-01 5.33648193e-01 4.22459729e-02 -3.60501021e-01 -1.09746724e-01 -1.91200048e-01 -3.08690280e-01 -7.47845888e-01 -8.51075411e-01 -1.08233297e+00 -2.58170605e-01 -1.32372364e-01 1.09012969e-01 1.08608508e+00 9.36177850e-01 -2.21130162e-01 5.74281096e-01 8.16253066e-01 -1.20850456e+00 -1.72665507e-01 -9.90362108e-01 -6.05679572e-01 5.95421493e-01 4.32991296e-01 -7.50733256e-01 -4.80449200e-01 4.79613319e-02]
[7.607814311981201, 3.0508198738098145]
ec870853-ddb1-4683-a799-b2da22238058
difftalk-crafting-diffusion-models-for
2301.03786
null
https://arxiv.org/abs/2301.03786v2
https://arxiv.org/pdf/2301.03786v2.pdf
DiffTalk: Crafting Diffusion Models for Generalized Audio-Driven Portraits Animation
Talking head synthesis is a promising approach for the video production industry. Recently, a lot of effort has been devoted in this research area to improve the generation quality or enhance the model generalization. However, there are few works able to address both issues simultaneously, which is essential for practical applications. To this end, in this paper, we turn attention to the emerging powerful Latent Diffusion Models, and model the Talking head generation as an audio-driven temporally coherent denoising process (DiffTalk). More specifically, instead of employing audio signals as the single driving factor, we investigate the control mechanism of the talking face, and incorporate reference face images and landmarks as conditions for personality-aware generalized synthesis. In this way, the proposed DiffTalk is capable of producing high-quality talking head videos in synchronization with the source audio, and more importantly, it can be naturally generalized across different identities without any further fine-tuning. Additionally, our DiffTalk can be gracefully tailored for higher-resolution synthesis with negligible extra computational cost. Extensive experiments show that the proposed DiffTalk efficiently synthesizes high-fidelity audio-driven talking head videos for generalized novel identities. For more video results, please refer to \url{https://sstzal.github.io/DiffTalk/}.
['Jiwen Lu', 'Jie zhou', 'Zheng Zhu', 'Wanhua Li', 'Zibin Meng', 'Wenliang Zhao', 'Shuai Shen']
2023-01-10
null
http://openaccess.thecvf.com//content/CVPR2023/html/Shen_DiffTalk_Crafting_Diffusion_Models_for_Generalized_Audio-Driven_Portraits_Animation_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Shen_DiffTalk_Crafting_Diffusion_Models_for_Generalized_Audio-Driven_Portraits_Animation_CVPR_2023_paper.pdf
cvpr-2023-1
['talking-head-generation']
['computer-vision']
[ 3.58583103e-03 1.61669105e-02 1.42137364e-01 -4.23793525e-01 -7.43955970e-01 -3.09790015e-01 5.49650192e-01 -5.44028938e-01 2.61381358e-01 3.95997465e-01 4.69717115e-01 2.96202868e-01 -4.06812541e-02 -4.82448637e-01 -5.23464561e-01 -1.03299522e+00 2.78687716e-01 -1.95985317e-01 -2.31154803e-02 -9.09530744e-02 -7.68683897e-03 3.91695410e-01 -1.75942254e+00 5.16260415e-02 9.63369489e-01 9.77140903e-01 5.10984182e-01 5.20777941e-01 2.01039478e-01 6.23062849e-01 -4.24163699e-01 -5.95590293e-01 6.38231589e-03 -7.40507722e-01 -2.46978357e-01 5.56777358e-01 5.12961388e-01 -3.07807297e-01 -3.88613790e-01 1.14875925e+00 8.35010946e-01 2.61134475e-01 3.79358053e-01 -1.19287956e+00 -6.95822299e-01 6.48639083e-01 -4.81063545e-01 9.64935403e-03 5.10753572e-01 2.98902065e-01 7.14370012e-01 -1.14538276e+00 5.07923782e-01 1.48389637e+00 2.65028208e-01 6.06892049e-01 -1.06964207e+00 -1.12480235e+00 3.47400874e-01 4.32205915e-01 -1.62620687e+00 -1.15313864e+00 1.24734581e+00 -3.15167189e-01 1.21572971e-01 2.74120510e-01 5.17334759e-01 1.42479289e+00 -6.88814223e-02 8.26261997e-01 9.50990021e-01 -2.27687657e-01 1.49708360e-01 1.59842849e-01 -2.87600547e-01 4.84277368e-01 -1.97760031e-01 -1.37217879e-01 -7.32120812e-01 1.51696401e-02 7.70630598e-01 -1.85158402e-01 -8.12448442e-01 -5.07144518e-02 -1.06842589e+00 6.04499638e-01 3.83403711e-02 3.50410461e-01 -4.03220296e-01 -1.28691057e-02 2.18644425e-01 2.99847741e-02 6.62573397e-01 -8.65952019e-03 1.99001193e-01 -2.91817844e-01 -1.08047771e+00 2.32392266e-01 5.30570745e-01 1.05483568e+00 3.99303824e-01 3.15659821e-01 -7.18744397e-02 1.03740489e+00 2.54573435e-01 4.75918710e-01 4.30648685e-01 -1.23982584e+00 3.22722703e-01 5.34940101e-02 -6.12556152e-02 -1.30622005e+00 -2.69363314e-01 -5.44956863e-01 -1.09855998e+00 -2.61892706e-01 6.56559691e-02 -2.51362175e-01 -3.35487336e-01 1.89412713e+00 5.56925297e-01 6.20244205e-01 -1.18259907e-01 1.04546154e+00 7.96191037e-01 8.30086529e-01 -2.76845515e-01 -7.53731251e-01 1.33758378e+00 -8.42026412e-01 -1.20698917e+00 2.11440492e-02 1.06240459e-01 -1.06175470e+00 1.08916199e+00 6.15598440e-01 -1.36125767e+00 -8.30989182e-01 -8.35113049e-01 2.54262220e-02 2.67054528e-01 2.69615352e-01 3.31040025e-01 6.05838180e-01 -1.16400170e+00 2.75328815e-01 -7.24545062e-01 -2.79356897e-01 2.76793599e-01 2.04687968e-01 -1.93819538e-01 -3.03435355e-01 -1.12137675e+00 2.76569396e-01 -2.87641119e-02 3.37014377e-01 -7.83874989e-01 -4.88456279e-01 -7.85706758e-01 3.88894789e-02 5.80128849e-01 -6.70807958e-01 1.28680837e+00 -1.01396322e+00 -2.02789617e+00 4.29283798e-01 -4.94287103e-01 -3.79258320e-02 5.19953787e-01 -1.51355699e-01 -5.21281421e-01 3.02880466e-01 -1.24371327e-01 4.47927952e-01 1.44588780e+00 -1.38082469e+00 -5.14482617e-01 -3.48256737e-01 -9.47881863e-02 2.12271616e-01 -7.91266620e-01 2.20625117e-01 -8.28550339e-01 -1.02173817e+00 -5.28528430e-02 -1.03408623e+00 2.65929222e-01 -1.22657940e-02 -3.58118683e-01 -2.04348832e-01 8.34425807e-01 -7.80668855e-01 1.22656357e+00 -2.47756886e+00 4.99734581e-01 -1.89123034e-01 1.36886194e-01 1.67347878e-01 -9.03788060e-02 2.24104315e-01 1.16207831e-01 -1.56610698e-01 -1.49865255e-01 -7.61410713e-01 1.23902899e-03 3.54929045e-02 -3.14551115e-01 5.27463198e-01 1.05908930e-01 5.64923763e-01 -7.10656881e-01 -5.52157521e-01 8.62089023e-02 9.46493566e-01 -7.43888378e-01 2.16191232e-01 7.63289481e-02 8.35837722e-01 -4.33632076e-01 5.79589605e-01 8.37383270e-01 -1.60547160e-02 1.24160543e-01 -3.25607359e-01 -1.15116253e-01 -1.10649347e-01 -1.43059695e+00 1.63727534e+00 -7.07281828e-01 5.55957317e-01 6.77832186e-01 -7.67253995e-01 1.09200430e+00 4.78261739e-01 3.14004600e-01 -4.00879622e-01 2.16516733e-01 1.93469957e-01 -2.35240906e-01 -5.80397964e-01 4.58713621e-01 -1.14049166e-01 1.67343676e-01 8.65630507e-02 5.77326082e-02 -1.04419194e-01 5.25155552e-02 -1.44773824e-02 6.19541168e-01 6.25327677e-02 6.01429865e-02 -2.37776339e-03 8.44403803e-01 -7.32818127e-01 7.53205538e-01 2.49944910e-01 -9.35403705e-02 7.17767596e-01 1.73410922e-01 2.96398491e-01 -6.96515262e-01 -9.63100672e-01 -1.20987043e-01 1.03172171e+00 2.42187962e-01 -4.05812472e-01 -1.01815915e+00 -9.09699276e-02 -3.78849357e-01 4.86728519e-01 -2.48383075e-01 -1.41591534e-01 -6.05775714e-01 -3.86198401e-01 4.46964294e-01 3.25227857e-01 7.45783389e-01 -9.12076890e-01 8.46838281e-02 2.47366875e-01 -6.32690907e-01 -1.21221054e+00 -9.06291187e-01 -6.40166402e-01 -6.28990710e-01 -6.17816210e-01 -1.23579073e+00 -9.14566278e-01 4.35222656e-01 4.03111488e-01 5.81049204e-01 -1.41372010e-01 -7.23546967e-02 3.78860444e-01 -4.16366279e-01 2.03886572e-02 -3.35728884e-01 -6.15578517e-02 4.22802955e-01 8.15005302e-01 1.52774656e-03 -1.00613177e+00 -6.93371713e-01 5.04576802e-01 -8.86721969e-01 1.76804826e-01 4.02854830e-01 7.93659210e-01 4.34777766e-01 2.19119862e-01 8.50053728e-01 -4.29888874e-01 9.14119661e-01 -4.98117149e-01 -5.71255386e-01 -5.41404299e-02 -8.07595253e-02 -3.49597245e-01 7.61626184e-01 -7.27747977e-01 -1.54256427e+00 4.84116040e-02 -4.40930039e-01 -6.86623633e-01 -1.96879193e-01 2.99292833e-01 -6.79533005e-01 4.12872583e-02 2.13916659e-01 5.48498213e-01 -5.58031872e-02 -7.15771139e-01 3.98914009e-01 9.52063739e-01 6.21059656e-01 -5.86023450e-01 7.88810194e-01 5.50769448e-01 -2.17457905e-01 -1.20845330e+00 -5.33989608e-01 -2.51139730e-01 -3.94935548e-01 -5.02468348e-01 7.02045500e-01 -9.85463202e-01 -8.08679640e-01 7.10417390e-01 -1.03558517e+00 -2.46798515e-01 9.81795415e-02 6.15250826e-01 -6.56143486e-01 5.51435947e-01 -6.93244278e-01 -8.50015700e-01 -1.65649548e-01 -1.37934101e+00 1.07671571e+00 3.09810817e-01 -1.25312477e-01 -7.55997181e-01 -1.95530549e-01 5.71585238e-01 3.54431868e-01 -4.84867245e-02 3.84060770e-01 -1.71659842e-01 -6.39357924e-01 1.11353649e-02 -3.97606790e-02 4.72733557e-01 3.83897185e-01 8.20447430e-02 -1.17574668e+00 -4.35552835e-01 3.74196440e-01 -9.89824161e-02 5.17523408e-01 5.01976013e-01 1.12404335e+00 -5.08997142e-01 -3.89472619e-02 7.72125125e-01 7.87871480e-01 2.07353145e-01 5.27993500e-01 -1.96993887e-01 5.98272383e-01 9.51299369e-01 6.09142303e-01 7.55484164e-01 4.64897931e-01 1.04235709e+00 1.93828493e-01 4.25978787e-02 -2.82616675e-01 -2.67468065e-01 6.48046792e-01 1.33299482e+00 -1.95132345e-01 -3.79919350e-01 -5.57823002e-01 4.67373282e-01 -1.62320364e+00 -1.07281089e+00 1.24777593e-01 1.94989264e+00 7.68771172e-01 -2.04800531e-01 9.16377828e-02 2.52800733e-01 8.98431957e-01 2.69734204e-01 -5.99861562e-01 -4.26419191e-02 -1.14838667e-01 -1.67442970e-02 -1.99022129e-01 5.00262320e-01 -7.52394378e-01 7.66960561e-01 4.69108391e+00 1.24467719e+00 -1.32800126e+00 2.25989059e-01 4.44600075e-01 -3.65477443e-01 -1.41220212e-01 -2.59071887e-01 -8.79557550e-01 7.16356516e-01 7.02642977e-01 -4.68079299e-01 4.95349258e-01 6.53141320e-01 7.93856502e-01 2.65751928e-01 -8.27692688e-01 1.30725241e+00 3.24687868e-01 -1.02556729e+00 3.81524712e-02 2.13957325e-01 5.45687497e-01 -6.73557580e-01 4.78352636e-01 1.22588664e-01 -2.97227114e-01 -6.32155597e-01 7.82075167e-01 6.93845153e-01 7.20884979e-01 -8.77571344e-01 5.14834404e-01 3.77761543e-01 -1.35108292e+00 -1.04911968e-01 -1.93266079e-01 1.94256693e-01 3.97517979e-01 5.47423959e-01 -4.07300502e-01 6.27099931e-01 7.58830249e-01 7.34891891e-01 -3.79789948e-01 9.25483167e-01 -3.50858808e-01 6.01639748e-01 -1.93194076e-01 3.02912861e-01 -2.40039796e-01 -1.52388513e-01 8.12472284e-01 1.10404932e+00 7.11830914e-01 2.73523539e-01 2.07676679e-01 7.16644645e-01 -2.23141491e-01 2.68282682e-01 -4.13049489e-01 9.80151594e-02 5.24189055e-01 1.31439710e+00 -3.88700664e-01 -1.92460284e-01 -2.71995366e-01 1.11522782e+00 -1.31125107e-01 4.33665425e-01 -1.01654851e+00 -3.25614154e-01 8.88618708e-01 2.34913036e-01 3.30217391e-01 -1.93690717e-01 2.33852476e-01 -1.34467125e+00 1.58746824e-01 -1.09775019e+00 -1.67518072e-02 -8.75356019e-01 -1.19205117e+00 6.67335510e-01 -1.02502793e-01 -1.28469479e+00 -2.70819843e-01 -7.24270642e-02 -4.75924999e-01 6.94386780e-01 -1.38845348e+00 -1.17589676e+00 -5.14393270e-01 9.43578660e-01 8.72996509e-01 1.18693002e-01 4.42919761e-01 6.06499195e-01 -8.15266848e-01 8.62506151e-01 1.37847081e-01 -1.55714884e-01 1.02049041e+00 -6.42881811e-01 1.35997273e-02 8.39757681e-01 -9.32928175e-02 6.94882274e-01 9.18862939e-01 -4.07733053e-01 -1.41399205e+00 -1.13006806e+00 6.42265499e-01 4.83757758e-04 5.24593830e-01 -5.66593587e-01 -1.04675329e+00 4.47205961e-01 2.78859347e-01 -1.53131306e-01 6.48668885e-01 -2.43915752e-01 3.37672904e-02 -3.47449154e-01 -1.04500914e+00 7.70275474e-01 1.20517504e+00 -5.04281163e-01 -2.72605956e-01 1.38300583e-01 7.67440915e-01 -2.24996477e-01 -1.00373113e+00 2.88860708e-01 4.15744394e-01 -1.05877483e+00 8.41943562e-01 2.36001432e-01 3.03946078e-01 -2.17896566e-01 -1.32335812e-01 -1.19908500e+00 -2.37649009e-01 -1.06933081e+00 -2.55303681e-01 2.01342726e+00 -1.78233683e-01 -5.80915451e-01 5.28435171e-01 3.04603338e-01 -1.85568869e-01 -5.38332462e-01 -8.80803883e-01 -6.93563104e-01 -1.68423474e-01 -4.23084110e-01 5.75260341e-01 8.64112079e-01 -9.31664854e-02 1.97002292e-01 -8.85016561e-01 4.19586957e-01 7.49157310e-01 1.93877339e-01 9.32148814e-01 -9.49584961e-01 -4.26108718e-01 -3.70996296e-01 -2.50821501e-01 -1.49047780e+00 2.48789445e-01 -5.27437449e-01 6.01608008e-02 -1.16525412e+00 -6.62493706e-02 -2.00021401e-01 1.21862225e-01 3.23192887e-02 -6.89727813e-02 3.95876318e-01 4.19037879e-01 3.39323878e-01 -3.53891492e-01 1.02875459e+00 1.47685730e+00 3.66386585e-02 -3.03888708e-01 2.02963516e-01 -8.03064883e-01 8.84679139e-01 9.17860091e-01 -2.13168457e-01 -5.71132720e-01 -2.64169812e-01 -2.24805042e-01 4.16048735e-01 2.50888586e-01 -9.55013216e-01 3.52097750e-01 -1.13270678e-01 1.05617743e-03 -3.05636048e-01 8.86547327e-01 -6.00715399e-01 3.11982870e-01 3.53841111e-02 -2.37730160e-01 -2.20478743e-01 5.89200146e-02 6.96810007e-01 -5.08075476e-01 4.75557074e-02 8.71965408e-01 1.23578124e-01 -3.60081881e-01 3.82440388e-01 -4.54728961e-01 -1.41332343e-01 9.43836927e-01 -1.58736378e-01 -7.36221224e-02 -8.28968525e-01 -8.40735435e-01 7.78618976e-02 4.95487362e-01 6.51751280e-01 6.87048972e-01 -1.34502029e+00 -7.16949344e-01 2.38410041e-01 -8.98395255e-02 -1.87777966e-01 8.37826908e-01 1.05760014e+00 3.31289554e-03 2.43787497e-01 1.62003841e-02 -5.88727474e-01 -1.46935725e+00 5.64811409e-01 7.45069385e-02 2.47404084e-01 -6.37934744e-01 7.86224306e-01 5.26256859e-01 -5.48933558e-02 4.29948002e-01 -4.40571010e-02 -3.05888683e-01 3.38242888e-01 7.96755433e-01 3.79414469e-01 -2.07481861e-01 -9.45036292e-01 -1.55523658e-01 7.43027210e-01 9.58169550e-02 -4.25075926e-02 1.21235669e+00 -7.03913271e-01 3.43273841e-02 3.83239806e-01 1.11553454e+00 4.24212009e-01 -1.43973517e+00 -2.08794594e-01 -4.15313661e-01 -6.25628829e-01 1.22173779e-01 -1.39318407e-01 -1.41013443e+00 9.39236403e-01 5.19464314e-01 6.61890358e-02 1.52423525e+00 -4.46025617e-02 9.11590576e-01 -2.86624879e-02 4.07770604e-01 -7.10871994e-01 2.40797430e-01 1.15894666e-02 1.11015642e+00 -8.53383958e-01 -3.06398362e-01 -6.60790503e-01 -6.71213329e-01 1.04843020e+00 4.23177391e-01 8.57183859e-02 4.89010513e-01 1.28284186e-01 -4.95929271e-02 9.61066857e-02 -6.33983731e-01 -4.90228273e-02 1.11278564e-01 6.19825006e-01 3.59077036e-01 -1.71976194e-01 -1.89603969e-01 7.75402904e-01 -3.64105284e-01 1.75535902e-01 5.66205144e-01 4.84701782e-01 -2.54308820e-01 -9.52742994e-01 -6.42589986e-01 -3.82386371e-02 -4.40549761e-01 2.11438816e-02 -1.34146139e-01 3.90259445e-01 1.14769042e-01 1.30995929e+00 -2.28229463e-01 -3.77019644e-01 3.02953720e-01 -2.17950754e-02 3.51269960e-01 -3.10430318e-01 -3.40162128e-01 4.95228976e-01 -1.01454750e-01 -4.76214439e-01 -4.65405166e-01 -7.13606715e-01 -7.98363864e-01 -5.66585183e-01 -3.84276211e-01 1.55045822e-01 4.07734036e-01 6.12283289e-01 5.76028228e-01 5.09903312e-01 9.14005339e-01 -1.21971834e+00 -3.56436282e-01 -9.37560678e-01 -6.34039044e-01 3.03860098e-01 3.52034450e-01 -7.29124308e-01 -5.93487263e-01 3.26601356e-01]
[13.220559120178223, -0.35609808564186096]
8c9e3a28-d06a-4d8e-8f0a-4f2a64380811
cross-view-image-matching-for-geo
1703.07815
null
http://arxiv.org/abs/1703.07815v1
http://arxiv.org/pdf/1703.07815v1.pdf
Cross-View Image Matching for Geo-localization in Urban Environments
In this paper, we address the problem of cross-view image geo-localization. Specifically, we aim to estimate the GPS location of a query street view image by finding the matching images in a reference database of geo-tagged bird's eye view images, or vice versa. To this end, we present a new framework for cross-view image geo-localization by taking advantage of the tremendous success of deep convolutional neural networks (CNNs) in image classification and object detection. First, we employ the Faster R-CNN to detect buildings in the query and reference images. Next, for each building in the query image, we retrieve the $k$ nearest neighbors from the reference buildings using a Siamese network trained on both positive matching image pairs and negative pairs. To find the correct NN for each query building, we develop an efficient multiple nearest neighbors matching method based on dominant sets. We evaluate the proposed framework on a new dataset that consists of pairs of street view and bird's eye view images. Experimental results show that the proposed method achieves better geo-localization accuracy than other approaches and is able to generalize to images at unseen locations.
['Yicong Tian', 'Chen Chen', 'Mubarak Shah']
2017-03-22
cross-view-image-matching-for-geo-1
http://openaccess.thecvf.com/content_cvpr_2017/html/Tian_Cross-View_Image_Matching_CVPR_2017_paper.html
http://openaccess.thecvf.com/content_cvpr_2017/papers/Tian_Cross-View_Image_Matching_CVPR_2017_paper.pdf
cvpr-2017-7
['cross-view-image-to-image-translation']
['computer-vision']
[-2.34921709e-01 -6.18723214e-01 1.30876273e-01 -4.06243503e-01 -1.07564306e+00 -6.87126994e-01 3.82124543e-01 3.75639684e-02 -5.29876888e-01 9.86807048e-02 -1.85516730e-01 1.47953540e-01 6.26457334e-02 -1.20724416e+00 -9.92661119e-01 -5.81419647e-01 6.14757687e-02 2.11484358e-01 4.51173961e-01 -1.02380194e-01 4.58705813e-01 6.47442102e-01 -1.79139006e+00 9.41605195e-02 4.44238245e-01 1.34451985e+00 4.07074034e-01 6.69395268e-01 4.17008907e-01 7.06060231e-01 -4.29931700e-01 -3.02414656e-01 2.84830272e-01 3.55853140e-02 -7.86525726e-01 9.14101303e-02 1.03442836e+00 -3.67083490e-01 -4.38096136e-01 1.07834685e+00 5.07715344e-01 2.45086968e-01 4.22439098e-01 -1.24880385e+00 -8.86354625e-01 3.75413261e-02 -8.11904073e-01 6.56437039e-01 6.37604296e-01 -2.17122987e-01 1.10417163e+00 -1.21776736e+00 4.36037391e-01 8.37384582e-01 8.37178707e-01 3.19006783e-03 -7.19706774e-01 -7.09929109e-01 1.66212797e-01 2.67416298e-01 -2.20553398e+00 -4.96661365e-01 8.39191258e-01 -4.22715157e-01 7.39004433e-01 1.49768889e-01 4.69938278e-01 4.09988463e-01 -3.05627864e-02 7.56441295e-01 7.14594543e-01 -2.21104771e-01 4.67673428e-02 7.90002644e-02 -4.23074007e-01 1.03059006e+00 -1.72108069e-01 5.13637671e-03 -4.10361081e-01 -2.03295946e-01 5.08213758e-01 4.94384259e-01 -4.22557026e-01 -7.39571452e-01 -1.36317384e+00 7.62696087e-01 1.23477554e+00 4.82423991e-01 -2.66646922e-01 3.60733360e-01 -4.45124879e-02 -2.20187336e-01 6.02815032e-01 2.86678672e-01 -2.46673524e-01 5.88874578e-01 -1.12273979e+00 1.65875509e-01 4.09392267e-01 9.93525565e-01 9.95181203e-01 -4.10257518e-01 2.80152887e-01 8.57672632e-01 2.90943891e-01 7.99593985e-01 3.89714390e-01 -6.20913804e-01 6.78336024e-01 6.64619505e-01 4.19711441e-01 -1.74183524e+00 -3.37384760e-01 -4.69425738e-01 -8.23916554e-01 -3.59476119e-01 2.98310131e-01 4.42818612e-01 -7.32011378e-01 1.47854316e+00 3.59444410e-01 3.68512362e-01 -5.34058772e-02 8.83458257e-01 9.42772746e-01 6.03468359e-01 -2.16673717e-01 6.55463815e-01 1.42051744e+00 -9.13321495e-01 6.76528215e-02 -4.81378734e-01 6.13298178e-01 -7.37894416e-01 8.19878101e-01 -1.72973216e-01 -8.55711162e-01 -7.05184758e-01 -9.71258938e-01 -4.92593832e-02 -1.03133750e+00 6.33097887e-01 1.44737691e-01 4.19832706e-01 -1.29069126e+00 2.70682365e-01 -4.82526541e-01 -6.34492099e-01 3.23640883e-01 4.16022390e-01 -6.29496098e-01 -5.21311402e-01 -1.10944629e+00 5.82278848e-01 3.35079581e-01 2.71155566e-01 -7.99469411e-01 -4.22700942e-01 -1.21249831e+00 1.37336388e-01 1.30239323e-01 -9.19895291e-01 7.74658263e-01 -6.74910486e-01 -5.40125847e-01 1.46747220e+00 -2.87005663e-01 -3.23009521e-01 1.77169830e-01 1.16325086e-02 -5.96204460e-01 1.26153320e-01 8.55827808e-01 6.78545356e-01 7.40117192e-01 -1.13326776e+00 -1.32683325e+00 -6.86598003e-01 1.94163144e-01 2.01386780e-01 4.37451079e-02 -1.87343940e-01 -7.57401645e-01 -3.70228291e-01 6.54347718e-01 -1.02265418e+00 -9.92713720e-02 -2.48097226e-01 -3.92511994e-01 -1.98938712e-01 5.84963322e-01 -4.63870466e-01 9.95884776e-01 -2.36893749e+00 -3.99433017e-01 2.68611282e-01 1.89695343e-01 -2.34732836e-01 1.22907199e-01 4.96064633e-01 -1.56247154e-01 -1.14492469e-01 8.53999928e-02 -9.57904980e-02 -2.65709639e-01 -1.09131090e-01 -2.57370472e-01 7.42939830e-01 -3.34347636e-02 9.19164062e-01 -1.03634083e+00 -4.07531381e-01 3.20931345e-01 3.47175837e-01 -3.78189087e-01 1.63733363e-01 3.72934580e-01 1.66348845e-01 -3.74395788e-01 8.36694837e-01 9.47631836e-01 -6.69360578e-01 -2.94131666e-01 -5.30142903e-01 -1.56740919e-01 -8.51496309e-02 -1.19113755e+00 1.55715370e+00 -8.62484634e-01 5.22262394e-01 -3.30119073e-01 -9.46338236e-01 9.17312980e-01 -8.16409066e-02 4.17533845e-01 -7.76622653e-01 -2.21993297e-01 3.66516382e-01 -8.34133208e-01 -4.40937191e-01 7.87070215e-01 3.04028124e-01 -3.08040470e-01 3.02588552e-01 -1.15774341e-01 -9.42048151e-03 8.95145610e-02 1.28310751e-02 7.65814364e-01 -2.58372873e-01 4.17123377e-01 5.76338507e-02 9.50505495e-01 -9.52992812e-02 4.28747050e-02 8.28127623e-01 -8.91914442e-02 9.25645411e-01 -2.10175171e-01 -9.00425494e-01 -8.80195141e-01 -1.07242215e+00 1.57636017e-01 1.24728727e+00 5.91833353e-01 -7.64432251e-02 -5.15859723e-01 -8.33704293e-01 1.78815518e-02 3.13136309e-01 -6.55537367e-01 1.71786487e-01 -6.53574467e-01 -5.29274344e-01 3.85981321e-01 6.12178147e-01 9.84865904e-01 -5.00238895e-01 -7.46509254e-01 -1.21874422e-01 -5.07142365e-01 -1.18185532e+00 -8.90212178e-01 -1.16861895e-01 -4.31463808e-01 -1.32937956e+00 -9.33567047e-01 -1.18304777e+00 9.20431972e-01 1.04136682e+00 1.31771636e+00 9.55804288e-02 -3.30867358e-02 5.17330110e-01 -5.78214712e-02 2.24050909e-01 2.60112524e-01 4.11273003e-01 -7.34993294e-02 2.89903015e-01 6.14526153e-01 -3.45359445e-01 -1.13039804e+00 6.70982182e-01 -7.17981398e-01 -3.36360902e-01 4.74812895e-01 6.99469864e-01 7.41064250e-01 1.26321420e-01 5.69223054e-02 -5.26548982e-01 9.91685241e-02 -6.12343013e-01 -9.53048110e-01 4.82540786e-01 -3.08968544e-01 -1.45119578e-01 5.15021861e-01 2.03008115e-01 -3.79271507e-01 6.68902159e-01 -2.35852659e-01 -3.73182327e-01 -3.54658902e-01 2.74113268e-01 -5.98428585e-02 -4.19508070e-01 4.93405163e-01 7.67174244e-01 -7.01766849e-01 -1.29114151e-01 5.27422965e-01 7.45850027e-01 7.17997015e-01 4.48914506e-02 9.16660666e-01 8.56777191e-01 -3.09824705e-01 -6.25498831e-01 -1.16739702e+00 -1.13002658e+00 -7.72953033e-01 -1.51485130e-01 1.03194273e+00 -1.42242908e+00 -8.47065687e-01 1.39251992e-01 -1.03574264e+00 5.40142655e-01 2.85796642e-01 3.98007989e-01 -5.64759254e-01 6.20353185e-02 -1.40534267e-01 -5.44788599e-01 -1.58180907e-01 -1.18442345e+00 1.71301603e+00 3.50742608e-01 2.82960892e-01 -8.05747449e-01 -1.62891438e-03 4.82972175e-01 2.03429133e-01 1.61345452e-01 3.49491328e-01 -6.48801088e-01 -1.04221427e+00 -6.20699108e-01 -5.58100700e-01 -6.13409057e-02 1.38718158e-01 -3.93989146e-01 -1.12141860e+00 -3.01189542e-01 -8.99087489e-02 1.59373373e-01 8.50686848e-01 3.85162771e-01 1.01287091e+00 -1.12936921e-01 -7.67497182e-01 9.50105488e-01 1.78675556e+00 3.11107755e-01 3.51870447e-01 5.57812691e-01 7.21617579e-01 2.80856311e-01 7.18563437e-01 2.79825479e-01 8.22540700e-01 7.08815873e-01 6.65872574e-01 -2.01130494e-01 4.29759294e-01 -6.50581241e-01 -1.33576632e-01 3.98391843e-01 2.75176257e-01 -3.44114631e-01 -1.10223591e+00 1.21705341e+00 -1.65099621e+00 -9.43384230e-01 1.16588227e-01 2.16747999e+00 -1.79074422e-01 -2.58508235e-01 1.91589911e-02 -2.41887316e-01 9.38553095e-01 3.86029273e-01 -3.73217732e-01 4.40891743e-01 3.94625925e-02 -1.50624663e-01 9.72271085e-01 2.20158800e-01 -1.80380952e+00 8.08509350e-01 5.81028175e+00 6.69546425e-01 -1.18351769e+00 1.00591846e-01 8.74209464e-01 1.13605475e-02 1.27719730e-01 -3.69944453e-01 -1.05032933e+00 4.70526248e-01 4.80771691e-01 2.04628915e-01 3.29768449e-01 1.29572797e+00 -6.60522282e-02 -3.32989872e-01 -9.77271855e-01 1.47514856e+00 5.71554065e-01 -1.61650050e+00 -2.24597484e-01 -1.44689217e-01 8.30635250e-01 3.97075504e-01 3.43775570e-01 -9.88559499e-02 5.99791929e-02 -8.26849401e-01 6.70933425e-01 3.19682509e-01 5.85927010e-01 -9.61512446e-01 8.54139566e-01 3.83239210e-01 -1.74920464e+00 -2.34720170e-01 -4.55576777e-01 3.97744119e-01 -9.75821242e-02 2.12357327e-01 -9.15233910e-01 6.56093478e-01 1.23359048e+00 9.63690221e-01 -9.41023111e-01 9.97887731e-01 1.78082986e-03 -1.37606218e-01 -4.56113964e-01 1.68104067e-01 5.34226418e-01 -6.71639666e-02 -7.98386261e-02 8.32337856e-01 8.10470223e-01 -1.05269343e-01 1.36732981e-01 8.91855478e-01 -9.85426456e-02 1.14494465e-01 -1.29469037e+00 4.26716745e-01 5.29113770e-01 1.27988172e+00 -9.36483026e-01 -2.46641576e-01 -3.54741931e-01 1.17236674e+00 4.47909147e-01 2.04392850e-01 -8.55852544e-01 -5.48434377e-01 3.39588374e-01 2.65629858e-01 8.93159270e-01 -1.21275812e-01 3.14125866e-01 -1.26719952e+00 1.48735521e-02 -3.57790262e-01 4.63110089e-01 -1.27727270e+00 -1.16374063e+00 7.50421226e-01 -2.73412704e-01 -1.41221535e+00 -2.61412442e-01 -3.38650316e-01 -3.73192191e-01 8.64468396e-01 -1.54015243e+00 -1.57963359e+00 -5.52938461e-01 6.75076544e-01 2.81594634e-01 1.46839293e-02 4.71093655e-01 7.91519642e-01 -2.33187318e-01 5.01735926e-01 3.85162562e-01 5.61093569e-01 5.69760561e-01 -1.05067825e+00 7.70474792e-01 8.84223998e-01 5.07680893e-01 6.66712761e-01 1.69363678e-01 -2.97856092e-01 -1.03853953e+00 -1.44440138e+00 1.09643745e+00 -7.00118840e-01 4.86518383e-01 -3.50478470e-01 -3.26695561e-01 8.38044465e-01 -8.74090120e-02 3.92299294e-01 4.75202829e-01 -3.00660491e-01 -3.10782433e-01 -3.79789084e-01 -1.14199591e+00 2.77639985e-01 9.59060192e-01 -9.75538433e-01 -2.62964338e-01 4.41642940e-01 4.17996943e-01 -5.14814317e-01 -7.75981367e-01 1.96436957e-01 4.67385441e-01 -1.17978644e+00 1.59278703e+00 -5.51869273e-02 1.93597615e-01 -8.21348369e-01 -6.17556214e-01 -1.26432967e+00 -1.17541410e-01 1.59059495e-01 5.29538453e-01 1.05176628e+00 1.25251144e-01 -5.21901309e-01 8.63505185e-01 1.00357242e-01 2.12591335e-01 -5.76753676e-01 -9.21689391e-01 -5.21700442e-01 -4.66917306e-01 -2.65015513e-01 8.39935422e-01 8.57173264e-01 -5.27929902e-01 2.40056202e-01 -3.40588719e-01 9.14496839e-01 4.98934299e-01 7.13664949e-01 8.12923312e-01 -8.91144991e-01 1.13461509e-01 4.83062491e-02 -7.12977469e-01 -1.35035765e+00 2.68442202e-02 -8.19532573e-01 1.42538786e-01 -1.41384041e+00 7.84781501e-02 -4.64709729e-01 -2.87424237e-01 -7.32182115e-02 -9.17383656e-02 1.00015700e+00 6.56876341e-02 2.61721760e-01 -1.13624775e+00 2.95348465e-01 6.25784755e-01 -3.86930525e-01 8.89866129e-02 3.54567498e-01 -4.86526579e-01 7.62596250e-01 5.47816277e-01 -6.23463333e-01 -2.00610310e-01 -5.54810166e-01 5.47265351e-01 9.37281772e-02 8.21800351e-01 -1.35691142e+00 5.09081304e-01 2.32123613e-01 5.90214610e-01 -1.34037912e+00 3.61349702e-01 -1.01494575e+00 1.81452528e-01 4.46476072e-01 -2.40210757e-01 6.42776012e-01 -1.76048607e-01 8.83650959e-01 -4.55910951e-01 -1.46482185e-01 8.12400103e-01 -5.74035108e-01 -1.01671946e+00 4.73544300e-01 -7.50361532e-02 7.17225447e-02 1.15334523e+00 -3.81003737e-01 -1.29085988e-01 -3.22468668e-01 -6.34105980e-01 1.90756589e-01 6.07809365e-01 3.48829120e-01 8.57651711e-01 -1.57617331e+00 -1.77030936e-01 3.80826801e-01 8.10818553e-01 -7.10681826e-02 3.89946699e-01 7.93288827e-01 -8.44470680e-01 6.93360567e-01 1.21692173e-01 -8.77310336e-01 -1.27338207e+00 9.55259264e-01 6.70943975e-01 -3.11346315e-02 -2.09239319e-01 9.81944203e-01 5.83351672e-01 -7.19668329e-01 -4.42623906e-03 -5.05010486e-01 -2.76393265e-01 -7.53313536e-03 5.54104567e-01 2.01350078e-01 2.76136324e-02 -1.25727010e+00 -7.42284894e-01 1.13703847e+00 1.01339579e-01 2.06291854e-01 1.24117386e+00 -5.38971782e-01 -6.93419352e-02 2.63478726e-01 1.62192249e+00 -1.47754019e-02 -7.30245292e-01 -3.61639380e-01 5.45504838e-02 -8.78281355e-01 -3.07807717e-02 -2.49323830e-01 -1.36362767e+00 8.80446315e-01 1.29236042e+00 3.70448440e-01 9.45686400e-01 2.69355386e-01 8.59029651e-01 4.25518245e-01 5.06600678e-01 -9.40808356e-01 -6.34375438e-02 2.51959652e-01 4.64047462e-01 -1.64006102e+00 -1.99045002e-01 -2.45194763e-01 -3.10344756e-01 9.73540068e-01 5.80917001e-01 -3.36558938e-01 9.73179698e-01 -4.42996413e-01 2.95859519e-02 -5.92241287e-01 -1.18603259e-01 -5.57728231e-01 3.23441416e-01 6.88329995e-01 2.02976480e-01 -1.01361394e-01 5.73228061e-01 -1.28145948e-01 -1.85496718e-01 -1.62241876e-01 1.37722775e-01 6.29130304e-01 -4.00235057e-01 -3.71997923e-01 -6.67928100e-01 1.66427374e-01 -4.31702375e-01 -2.89249480e-01 -1.94314793e-01 8.39146793e-01 4.28767085e-01 8.90273213e-01 3.47955942e-01 -4.84170228e-01 4.29414392e-01 -1.92758664e-01 1.38275817e-01 -3.76534998e-01 -4.88514423e-01 -2.47602820e-01 -3.54071945e-01 -5.95293462e-01 -6.94291353e-01 -3.13183010e-01 -7.03800201e-01 -3.53516102e-01 -3.68646264e-01 6.59162998e-02 5.19850969e-01 8.16025734e-01 5.32651007e-01 1.18071638e-01 9.94289637e-01 -9.50155914e-01 -9.11481977e-02 -3.13876331e-01 -6.24626517e-01 3.33499730e-01 4.55980837e-01 -6.19206369e-01 -2.50501633e-01 -2.58985572e-02]
[7.707897186279297, -1.9389103651046753]
9474f4ff-3333-40bd-9c6c-e66ebce70c32
adaptive-bernstein-change-detector-for-high
2306.12974
null
https://arxiv.org/abs/2306.12974v1
https://arxiv.org/pdf/2306.12974v1.pdf
Adaptive Bernstein Change Detector for High-Dimensional Data Streams
Change detection is of fundamental importance when analyzing data streams. Detecting changes both quickly and accurately enables monitoring and prediction systems to react, e.g., by issuing an alarm or by updating a learning algorithm. However, detecting changes is challenging when observations are high-dimensional. In high-dimensional data, change detectors should not only be able to identify when changes happen, but also in which subspace they occur. Ideally, one should also quantify how severe they are. Our approach, ABCD, has these properties. ABCD learns an encoder-decoder model and monitors its accuracy over a window of adaptive size. ABCD derives a change score based on Bernstein's inequality to detect deviations in terms of accuracy, which indicate changes. Our experiments demonstrate that ABCD outperforms its best competitor by at least 8% and up to 23% in F1-score on average. It can also accurately estimate changes' subspace, together with a severity measure that correlates with the ground truth.
['Klemens Böhm', 'Florian Kalinke', 'Tanja Fenn', 'Vadim Arzamasov', 'Edouard Fouché', 'Marco Heyden']
2023-06-22
null
null
null
null
['change-detection']
['computer-vision']
[ 8.44260305e-02 -3.83806407e-01 -2.08838522e-01 -2.96774387e-01 -6.95737839e-01 -7.93743908e-01 5.42205215e-01 8.33247900e-01 -2.87667006e-01 4.70070571e-01 1.67380497e-01 -2.53210533e-02 2.78297700e-02 -8.84067059e-01 -8.33166301e-01 -4.57954675e-01 -4.02486473e-01 3.41598660e-01 3.81968111e-01 9.77887362e-02 2.46504053e-01 5.05030811e-01 -1.60571492e+00 1.06897421e-01 4.49355394e-01 1.29872012e+00 -1.32163435e-01 9.09262121e-01 1.00033604e-01 6.52513981e-01 -7.56331384e-01 -1.09937020e-01 1.75935715e-01 -4.25491095e-01 -2.54243881e-01 3.87911536e-02 2.19956100e-01 -3.21380079e-01 -6.15170859e-02 9.01742399e-01 3.70445490e-01 -1.92467585e-01 5.04240811e-01 -1.25208437e+00 9.91507173e-02 3.45453560e-01 -4.89154756e-01 7.81132877e-01 6.71694994e-01 2.73933738e-01 9.23531950e-01 -7.35907555e-01 4.07981843e-01 8.42108965e-01 7.79629469e-01 2.49468148e-01 -1.24330568e+00 -6.06495261e-01 2.10212901e-01 3.88088077e-01 -1.22914267e+00 -5.08014023e-01 6.10824585e-01 -6.84074104e-01 6.47485435e-01 4.52559650e-01 5.81613541e-01 9.63649869e-01 1.91383243e-01 7.45671630e-01 6.11967683e-01 -1.14949428e-01 6.41568244e-01 -8.07246845e-03 -8.60671774e-02 3.63909572e-01 4.15523112e-01 -4.41312827e-02 -8.35985661e-01 -1.82057530e-01 4.32157874e-01 1.33896351e-01 -4.80776012e-01 -3.10516775e-01 -1.26881444e+00 3.73351663e-01 7.24129938e-03 1.99514821e-01 -7.30690360e-01 1.50530739e-02 5.79328656e-01 5.80748260e-01 4.32935774e-01 3.63109171e-01 -5.54435313e-01 -8.89757752e-01 -8.34170043e-01 -4.06195074e-02 6.82445884e-01 8.23221564e-01 3.83533686e-01 -2.69647837e-02 -3.58771712e-01 4.87561077e-01 3.82089503e-02 5.82430661e-01 4.78570461e-01 -6.52709663e-01 4.32202637e-01 6.69298649e-01 3.09914380e-01 -1.09615517e+00 -4.70787436e-01 -6.45836353e-01 -1.14650249e+00 7.53883412e-03 2.03344643e-01 -2.81069125e-03 -1.91250905e-01 1.89974749e+00 5.24963439e-01 5.74680567e-01 -1.42766207e-01 7.39716053e-01 7.46311024e-02 5.77902675e-01 -3.19252372e-01 -7.70659029e-01 1.06233740e+00 -1.12483457e-01 -7.22100437e-01 -1.37213049e-02 6.52457297e-01 -5.24732709e-01 1.07801187e+00 6.70521200e-01 -8.79427552e-01 -4.54231739e-01 -9.32244182e-01 8.02357495e-01 4.74486426e-02 -8.51743072e-02 -8.89538378e-02 5.92673540e-01 -7.90189505e-01 5.46541154e-01 -1.04079688e+00 -4.20797974e-01 2.78238088e-01 9.51162726e-02 -9.54263061e-02 1.50479704e-01 -1.07309163e+00 5.17432749e-01 1.83022216e-01 -2.63249606e-01 -9.89845872e-01 -8.34531128e-01 -5.20705879e-01 1.28543571e-01 2.24506527e-01 -4.16816890e-01 1.32826841e+00 -7.24966109e-01 -1.25989211e+00 3.79789025e-01 -2.34833091e-01 -7.38871157e-01 8.53591442e-01 -2.51311153e-01 -8.52090120e-01 -5.58577366e-02 -2.18303546e-01 -8.60337764e-02 9.22069788e-01 -7.33851552e-01 -1.01871049e+00 -3.75622898e-01 -1.45170510e-01 6.95139766e-02 -5.31188369e-01 -1.38033003e-01 -4.71368313e-01 -4.29239392e-01 1.16668858e-01 -7.15150952e-01 2.85138898e-02 4.38358858e-02 -2.77409405e-01 -7.23032048e-03 9.18709636e-01 -5.92137039e-01 1.93839538e+00 -2.19681191e+00 -3.51456612e-01 3.98476839e-01 2.49017805e-01 1.77467719e-01 5.59352152e-03 5.49656987e-01 3.31251137e-02 -3.07801478e-02 -2.47067839e-01 7.07136700e-03 4.84516881e-02 -2.56664399e-02 -3.92309070e-01 6.00109041e-01 2.37324864e-01 3.90335381e-01 -9.76455688e-01 4.44286838e-02 1.04557224e-01 1.87912568e-01 -6.36329353e-01 2.66715765e-01 -1.40906036e-01 3.99185568e-01 -1.26182079e-01 4.07355368e-01 3.84262383e-01 -2.36272961e-01 2.22712159e-02 6.52793050e-02 -8.95698145e-02 3.20790857e-01 -1.55054033e+00 9.68333960e-01 -3.76188934e-01 1.01592898e+00 -4.19400096e-01 -1.12472045e+00 9.69246328e-01 3.76213640e-01 6.90360963e-01 -6.72919273e-01 -2.43763641e-01 1.06428348e-01 -2.34514534e-01 -4.44684386e-01 2.16204569e-01 3.36132079e-01 -1.38699844e-01 4.22376454e-01 -7.44216383e-01 3.75645041e-01 2.06009060e-01 -8.14800966e-04 1.85246253e+00 -5.10234714e-01 5.79366624e-01 9.33648497e-02 5.82356870e-01 -4.18304592e-01 7.03485310e-01 7.27157056e-01 -2.44178191e-01 2.93484539e-01 7.50845790e-01 -4.61998016e-01 -7.71588564e-01 -1.07910562e+00 -1.22554690e-01 6.17172599e-01 1.12208970e-01 -4.13613111e-01 -4.53660131e-01 -6.72789514e-01 3.87422889e-01 8.78771842e-01 -3.70165378e-01 -5.61683238e-01 -3.85907620e-01 -5.90388894e-01 3.62957567e-01 4.81480092e-01 3.45442653e-01 -5.96517324e-01 -8.10848951e-01 4.01296794e-01 -2.06009760e-01 -9.98804390e-01 -5.12551665e-01 1.09130740e-01 -9.66919720e-01 -1.19817591e+00 -1.98617578e-01 -1.57124937e-01 5.12851000e-01 1.86496481e-01 1.13002872e+00 -3.87537539e-01 1.18960822e-02 5.20387352e-01 -3.81708443e-01 -4.83653963e-01 -5.49158394e-01 -1.23436965e-01 4.13541466e-01 3.87110591e-01 5.60704350e-01 -4.96840358e-01 -6.03581071e-01 5.02954781e-01 -1.00358188e+00 -5.41848779e-01 5.45520842e-01 3.95326644e-01 6.59967959e-01 3.60987127e-01 7.96211362e-01 -6.39631569e-01 7.27760613e-01 -7.42439985e-01 -7.36727774e-01 1.23550147e-01 -1.14218771e+00 2.15390585e-02 6.35170162e-01 -4.47189361e-01 -4.56473351e-01 -2.90721171e-02 -9.05508932e-04 -4.78137136e-01 -2.94664025e-01 3.98641169e-01 -6.47496134e-02 5.16486764e-01 7.65060961e-01 3.29750538e-01 -3.49150479e-01 -3.13372344e-01 2.03216206e-02 9.11261678e-01 5.94689608e-01 -1.14814594e-01 6.83851242e-01 3.40802729e-01 -1.61755890e-01 -8.53123009e-01 -5.76640844e-01 -8.30912769e-01 -5.94627678e-01 -6.52552545e-01 2.09241852e-01 -1.02179492e+00 -6.84212983e-01 4.27372336e-01 -9.53726113e-01 -1.55182406e-01 -4.44888115e-01 5.54245472e-01 -4.43734169e-01 1.81630164e-01 -3.31689119e-01 -7.50048161e-01 -1.49901286e-01 -5.75085580e-01 9.56831396e-01 -4.77356836e-02 -5.18854558e-01 -8.53132546e-01 2.95874029e-01 -2.36904785e-01 4.57598060e-01 4.70554650e-01 6.00145459e-01 -7.09179163e-01 -4.55334336e-01 -5.71016729e-01 1.36862203e-01 2.06392840e-01 3.76186818e-01 1.18255280e-01 -7.97842741e-01 -5.97079277e-01 -1.07179955e-01 3.73725384e-01 6.08289301e-01 3.87781411e-01 1.23905087e+00 -5.50968707e-01 -2.01922417e-01 2.87252396e-01 1.24319148e+00 3.34186345e-01 3.58802885e-01 3.23800802e-01 8.72753039e-02 2.08537817e-01 6.83918297e-01 1.15829897e+00 2.18145072e-01 8.88920486e-01 5.33782661e-01 3.43889266e-01 5.64114153e-02 -1.28797293e-01 7.89209187e-01 9.96178508e-01 4.12116230e-01 -4.04134750e-01 -1.01071286e+00 4.20998424e-01 -1.80077362e+00 -1.13230133e+00 -3.65830511e-01 2.78120661e+00 9.43094194e-01 5.67900419e-01 3.26790512e-01 6.18079543e-01 7.19237208e-01 3.98709439e-03 -1.03004909e+00 -3.18625093e-01 1.93041433e-02 -2.66851217e-01 3.71266425e-01 1.32573500e-01 -1.00251186e+00 3.71919811e-01 6.19638968e+00 3.37886274e-01 -1.27930582e+00 1.44662643e-02 4.35760379e-01 -3.13789636e-01 -1.16503097e-01 -3.78272891e-01 -8.18874776e-01 9.41306591e-01 1.28496611e+00 -4.71485227e-01 3.36311758e-01 6.15929008e-01 5.67501187e-01 -1.43437296e-01 -1.38119888e+00 1.22139764e+00 8.33921507e-02 -1.04038823e+00 -1.97733924e-01 -8.01831037e-02 7.13446736e-01 4.63853814e-02 -1.28517061e-01 1.55200690e-01 7.44566396e-02 -4.07036901e-01 6.83776677e-01 8.05135250e-01 7.67024338e-01 -8.17693353e-01 5.88765800e-01 6.16651833e-01 -1.18876863e+00 -1.33780181e-01 -8.67474005e-02 -1.61149338e-01 1.36535866e-02 1.28725767e+00 -1.07299805e+00 9.53314900e-02 7.00568318e-01 9.42859948e-01 -4.71740067e-01 1.39489400e+00 -1.80385157e-01 1.14260769e+00 -5.15486240e-01 -7.57249445e-02 -2.30422735e-01 7.40948552e-03 8.15702796e-01 1.31974959e+00 7.26529062e-01 -9.76378918e-02 1.52143821e-01 3.26238424e-01 -4.51245569e-02 6.13571480e-02 -4.88734841e-01 7.95611441e-02 9.06476557e-01 7.64864445e-01 -4.26507503e-01 -4.46175307e-01 -3.78132075e-01 9.96641755e-01 -1.05901033e-01 1.71415254e-01 -8.51210833e-01 -4.28839833e-01 1.11127067e+00 2.58058697e-01 4.00463134e-01 -1.95958823e-01 -1.19785868e-01 -1.03056204e+00 3.68496239e-01 -8.83330226e-01 3.61892313e-01 -4.06960785e-01 -9.51897144e-01 3.59997392e-01 -3.98214430e-01 -1.86211312e+00 -4.12295878e-01 -1.61094502e-01 -5.03472209e-01 2.89822280e-01 -1.47285604e+00 -1.03731491e-01 -6.73761904e-01 4.82990086e-01 5.45789897e-01 6.16505891e-02 6.02209866e-01 5.52699924e-01 -9.61156249e-01 6.49710000e-01 5.31551719e-01 4.02653664e-02 8.02024901e-01 -1.22209537e+00 4.76488382e-01 1.04112899e+00 3.66094083e-01 -8.57978985e-02 9.84046817e-01 -4.81512785e-01 -1.29473948e+00 -1.38693750e+00 7.54044116e-01 -3.82043183e-01 7.16148555e-01 -8.75090882e-02 -1.12671602e+00 4.58511829e-01 -4.75569993e-01 2.33710911e-02 6.17357373e-01 9.70762968e-02 -4.59553301e-01 -6.16511285e-01 -9.19055164e-01 2.77057469e-01 9.93520498e-01 -5.30555248e-01 -3.44854057e-01 2.46510416e-01 4.30749416e-01 -4.29781467e-01 -9.29163456e-01 2.97299743e-01 4.85759974e-01 -1.14406788e+00 4.96181041e-01 -6.91300750e-01 1.15617365e-01 -4.77079570e-01 -2.48327687e-01 -1.40576386e+00 -3.28355521e-01 -6.22704387e-01 -6.79669559e-01 1.33754253e+00 2.83127844e-01 -6.29546523e-01 4.68750030e-01 3.02407742e-01 1.73860461e-01 -6.30422831e-01 -8.90972137e-01 -1.20063221e+00 -5.88681161e-01 -9.55636263e-01 8.06638598e-01 9.66454029e-01 -2.38047782e-02 1.34065837e-01 -2.44095862e-01 7.30033159e-01 3.66599172e-01 -7.78003111e-02 8.71126413e-01 -1.45722425e+00 -1.63320109e-01 -6.55797601e-01 -8.71755242e-01 -1.10543704e+00 -2.72271782e-01 -6.13859951e-01 5.53124677e-03 -1.13987434e+00 -1.59342762e-03 -3.14301163e-01 -6.29169941e-01 3.11807662e-01 -3.21562625e-02 -3.53039019e-02 -2.68800020e-01 4.72051501e-01 -7.64222264e-01 3.92415732e-01 4.32414562e-01 -2.21555610e-03 -4.24438775e-01 4.29976165e-01 -3.47606599e-01 6.18350625e-01 9.73588645e-01 -5.01442432e-01 -2.41550237e-01 -1.78508043e-01 5.84908247e-01 6.17048591e-02 2.93779701e-01 -1.45618606e+00 2.44605631e-01 -1.51330069e-01 2.84251422e-01 -5.34780383e-01 -9.11262408e-02 -7.86845863e-01 3.66523385e-01 6.70268714e-01 -6.17716432e-01 3.12733084e-01 4.82527800e-02 1.02570117e+00 -4.04097557e-01 9.50191990e-02 7.63201177e-01 4.84636813e-01 -7.30336249e-01 3.22819352e-01 -5.35036683e-01 1.75355792e-01 1.35242200e+00 -9.95011535e-03 -1.57279968e-01 -7.41815150e-01 -3.14608127e-01 3.46455544e-01 4.11002994e-01 5.00993848e-01 5.79850495e-01 -1.44357026e+00 -7.81215012e-01 4.49210376e-01 5.27864575e-01 -2.63636947e-01 1.10592768e-02 1.08047271e+00 -2.74782211e-01 2.82662362e-01 2.86184043e-01 -9.24850106e-01 -1.44283426e+00 4.89132017e-01 2.08295941e-01 -1.74441889e-01 -5.04492044e-01 5.10986686e-01 -3.22189063e-01 5.60599528e-02 6.43287063e-01 -6.62777483e-01 -1.98926821e-01 3.84056211e-01 8.82732630e-01 5.41032255e-01 3.73397946e-01 -3.50061804e-01 -4.86217171e-01 4.26746517e-01 2.37114340e-01 2.54415333e-01 1.20867574e+00 -2.61497378e-01 3.04040313e-01 8.11676204e-01 1.10387516e+00 -6.23782501e-02 -1.40350795e+00 -6.28582120e-01 4.68077838e-01 -5.36256373e-01 1.22474633e-01 -7.88854778e-01 -7.93059051e-01 7.12105513e-01 9.84133899e-01 7.58510828e-01 1.56005013e+00 -2.83277668e-02 7.74234354e-01 4.23602045e-01 3.28482449e-01 -1.02574432e+00 1.36124775e-01 4.03569400e-01 8.14896464e-01 -1.14472890e+00 -1.66247413e-01 1.44402340e-01 -5.12772739e-01 1.08264601e+00 2.68711746e-01 2.49200389e-01 7.21654773e-01 3.48434329e-01 -1.76392831e-02 6.72940463e-02 -1.27513778e+00 -1.56058609e-01 2.15588480e-01 4.19493586e-01 6.81680888e-02 2.40625113e-01 -6.23654462e-02 2.83954054e-01 -8.82144794e-02 -2.09768146e-01 5.53972840e-01 6.71470046e-01 -7.67081499e-01 -6.95108354e-01 -4.26021755e-01 6.65475428e-01 -1.38613001e-01 3.50225300e-01 -2.58361369e-01 4.16946650e-01 -9.49796513e-02 9.79578614e-01 5.18631041e-01 -6.76651835e-01 7.95172691e-01 8.78932402e-02 -2.13231862e-01 -2.39626467e-01 -1.89639613e-01 -1.61102414e-01 -1.76392615e-01 -9.07196999e-01 -1.82991803e-01 -1.28087771e+00 -1.09572780e+00 -3.63167197e-01 -1.25637561e-01 3.94475609e-02 6.16665065e-01 8.01914334e-01 6.60816491e-01 5.97595394e-01 1.34715056e+00 -1.15794517e-01 -7.75695324e-01 -8.50368738e-01 -4.64040309e-01 3.84967953e-01 6.63996518e-01 -3.38765889e-01 -4.57424998e-01 2.34639391e-01]
[7.43136739730835, 3.0040297508239746]
430c7afa-62e9-4014-bffe-7237b9bffbab
fednoisy-federated-noisy-label-learning
2306.11650
null
https://arxiv.org/abs/2306.11650v1
https://arxiv.org/pdf/2306.11650v1.pdf
FedNoisy: Federated Noisy Label Learning Benchmark
Federated learning has gained popularity for distributed learning without aggregating sensitive data from clients. But meanwhile, the distributed and isolated nature of data isolation may be complicated by data quality, making it more vulnerable to noisy labels. Many efforts exist to defend against the negative impacts of noisy labels in centralized or federated settings. However, there is a lack of a benchmark that comprehensively considers the impact of noisy labels in a wide variety of typical FL settings. In this work, we serve the first standardized benchmark that can help researchers fully explore potential federated noisy settings. Also, we conduct comprehensive experiments to explore the characteristics of these data settings and unravel challenging scenarios on the federated noisy label learning, which may guide method development in the future. We highlight the 20 basic settings for more than 5 datasets proposed in our benchmark and standardized simulation pipeline for federated noisy label learning. We hope this benchmark can facilitate idea verification in federated learning with noisy labels. \texttt{FedNoisy} is available at \codeword{https://github.com/SMILELab-FL/FedNoisy}.
['Zenglin Xu', 'Jiayu Zhou', 'Junyuan Hong', 'Dun Zeng', 'Jintao Huang', 'Siqi Liang']
2023-06-20
null
null
null
null
['learning-with-noisy-labels', 'learning-with-noisy-labels']
['computer-vision', 'natural-language-processing']
[-2.92315245e-01 -5.18681407e-01 -8.39232374e-03 -5.77605784e-01 -1.18367124e+00 -1.10865593e+00 1.42898425e-01 -8.04144219e-02 -4.67760116e-02 7.78023243e-01 -7.44568184e-03 -3.68790984e-01 -2.49353364e-01 -4.63288695e-01 -4.75792617e-01 -9.38115954e-01 -2.39414588e-01 2.04324543e-01 -1.94921583e-01 1.58548698e-01 -2.29243442e-01 2.46758401e-01 -1.58845472e+00 3.22167069e-01 5.05538821e-01 1.13169277e+00 -2.58303434e-01 5.16330600e-01 -2.10727230e-02 9.61534560e-01 -1.00424743e+00 -6.70147777e-01 5.68585098e-01 -2.05008537e-01 -8.13430071e-01 -2.98039854e-01 6.95349753e-01 -4.00267541e-01 -3.45450103e-01 1.37402105e+00 1.04833019e+00 -1.86680242e-01 -7.86879584e-02 -1.71766591e+00 -6.68800116e-01 1.28645921e+00 -1.86760053e-01 1.98045671e-01 2.00033173e-01 6.30200386e-01 9.97353375e-01 -4.12119269e-01 5.46411037e-01 1.06997097e+00 8.06344211e-01 5.76094031e-01 -9.37077105e-01 -1.28069103e+00 1.42734408e-01 2.49490812e-01 -1.12773824e+00 -5.74059010e-01 6.08172059e-01 -6.45780116e-02 2.47114435e-01 5.34541368e-01 2.67094020e-02 1.36253297e+00 -4.51648593e-01 8.71086299e-01 1.20030582e+00 -1.75887316e-01 4.45025712e-01 9.09055322e-02 3.66887003e-01 5.32603681e-01 4.47407067e-01 2.85794556e-01 -7.16608644e-01 -8.22624683e-01 -2.65184075e-01 2.17972025e-01 -4.01052803e-01 -1.63363263e-01 -1.08488548e+00 7.28357673e-01 3.57347190e-01 1.26856580e-01 1.28534332e-01 2.43559614e-01 7.87961304e-01 6.49095416e-01 4.80879456e-01 1.97561964e-01 -6.78253174e-01 -1.74442932e-01 -6.81562901e-01 6.73228875e-02 1.01138926e+00 1.06346893e+00 9.38003123e-01 -3.70557681e-02 -2.80359924e-01 5.87322593e-01 2.01409429e-01 7.73701549e-01 2.83875406e-01 -1.52785873e+00 2.71313280e-01 4.76096332e-01 4.47586328e-02 -4.92942750e-01 -2.27589518e-01 -4.62729156e-01 -6.32688820e-01 -1.32349031e-02 6.03751421e-01 -6.04150712e-01 -2.27655783e-01 1.83591819e+00 7.99184203e-01 3.62719625e-01 -1.03774369e-01 1.06518912e+00 8.18369746e-01 7.81328455e-02 5.46799088e-03 -1.45772044e-02 9.96427238e-01 -9.39430058e-01 -8.17470074e-01 5.26091456e-01 1.04650319e+00 -8.24562788e-01 1.04913819e+00 6.21361017e-01 -6.87748313e-01 1.43655077e-01 -7.06659257e-01 1.19710863e-01 -4.76668328e-01 -3.94102871e-01 8.38401735e-01 1.11045599e+00 -1.02541196e+00 3.70254815e-01 -6.15038753e-01 -4.12837088e-01 7.65784442e-01 1.02242902e-01 3.65369618e-02 -4.03316885e-01 -1.25195086e+00 2.89027154e-01 -4.14078563e-01 -2.99996197e-01 -1.28843141e+00 -7.89905429e-01 -4.19477105e-01 -3.62258255e-02 4.32993561e-01 -4.83813345e-01 1.64686012e+00 -4.15277749e-01 -9.40283775e-01 6.57636821e-01 1.36813894e-01 -1.45120069e-01 6.60529315e-01 7.17081726e-02 -5.61493516e-01 -8.82683024e-02 -6.20138049e-02 -4.30069156e-02 1.99305266e-01 -1.34240723e+00 -6.46736205e-01 -5.85445762e-01 1.81442834e-02 -2.88532495e-01 -6.50977612e-01 2.44294614e-01 -3.43594290e-02 -4.19295788e-01 -3.62707525e-01 -7.45469153e-01 -1.17674939e-01 1.31901667e-01 -4.86043185e-01 -4.23385948e-01 1.35971534e+00 1.25642121e-01 1.00876367e+00 -2.33581996e+00 -8.72176170e-01 1.72742501e-01 4.25468683e-01 2.88680285e-01 -4.03189987e-01 5.26761353e-01 2.70907879e-01 4.47466731e-01 1.42336205e-01 -5.78121901e-01 4.36730802e-01 -2.45060213e-02 -3.40794981e-01 9.08201218e-01 -4.31599408e-01 6.55536532e-01 -1.03567827e+00 -4.53308046e-01 -1.43137306e-01 4.23750013e-01 -3.29646856e-01 4.28424001e-01 -3.15837294e-01 5.08597136e-01 -7.00020790e-01 1.20082855e+00 9.50227976e-01 -4.54585284e-01 3.16485226e-01 -5.47990501e-02 1.62819654e-01 2.72702783e-01 -1.37277079e+00 1.49893415e+00 -2.81171978e-01 2.29762033e-01 6.25300825e-01 -5.12474418e-01 8.17299187e-01 4.76160884e-01 6.96733296e-01 -6.18469536e-01 2.70164847e-01 2.15846613e-01 -5.54531813e-01 -5.17233431e-01 1.13740414e-01 2.43897378e-01 -6.99250996e-02 1.30865562e+00 4.03723679e-02 2.56797194e-01 1.23856831e-02 3.67148727e-01 1.54110920e+00 -4.42175835e-01 -1.99542418e-01 -5.22394404e-02 1.25765696e-01 -2.96932042e-01 8.11779976e-01 1.08839774e+00 -9.90177691e-01 4.53908473e-01 2.27606326e-01 -5.99101007e-01 -5.15974164e-01 -8.84927630e-01 -2.87569612e-01 1.56410897e+00 3.56562346e-01 -5.30552447e-01 -5.47481716e-01 -1.14391434e+00 3.03729206e-01 4.58320320e-01 -3.09184939e-01 -4.94186915e-02 3.15954909e-02 -8.11373830e-01 1.34194565e+00 -5.22330627e-02 4.43880528e-01 -8.93227220e-01 -2.35217184e-01 -2.27883384e-01 -4.27898735e-01 -9.07119334e-01 -6.56671464e-01 4.69274044e-01 -4.58527416e-01 -1.50964713e+00 2.18061116e-02 -4.75892872e-01 3.36903960e-01 6.34848595e-01 1.19018257e+00 3.91715646e-01 -3.54868561e-01 5.24148822e-01 -5.44214666e-01 -4.27336037e-01 -3.27030063e-01 -3.62545811e-03 2.50237674e-01 -1.03997923e-01 6.79739177e-01 -5.77704251e-01 -7.08081484e-01 7.78985679e-01 -1.19172788e+00 -7.97458172e-01 -9.57827494e-02 6.73791707e-01 4.00257945e-01 8.37146770e-03 8.36602747e-01 -1.30203366e+00 6.14575028e-01 -1.02019322e+00 -6.01171255e-01 5.63935935e-01 -8.93797576e-01 -2.01598108e-01 8.60491514e-01 -3.86232764e-01 -8.53853583e-01 -6.75253272e-02 1.29023110e-02 -3.91930461e-01 -5.02948701e-01 -1.35233924e-01 -3.50131333e-01 -1.91062063e-01 1.00617111e+00 -2.74655491e-01 -2.06573218e-01 -8.26541185e-01 4.84417140e-01 1.17458677e+00 2.45081812e-01 -9.77131069e-01 7.02430546e-01 6.27938509e-01 -4.29064631e-01 6.70023113e-02 -8.10782552e-01 -6.34959996e-01 7.65783787e-02 -2.19938979e-01 -9.39203706e-03 -9.32817519e-01 -1.13603365e+00 5.36851168e-01 -7.73687422e-01 -5.78874767e-01 -3.79308283e-01 1.78583622e-01 -2.41822064e-01 5.76755762e-01 -1.03602922e+00 -7.24104047e-01 -7.77317286e-01 -1.19221544e+00 7.99373567e-01 2.31528074e-01 1.29748851e-01 -8.34679008e-01 2.94186264e-01 7.75955141e-01 7.15338051e-01 1.58814088e-01 4.51371580e-01 -9.59699988e-01 -4.94326204e-01 -1.37725547e-01 -1.00645535e-01 2.95212209e-01 7.97975883e-02 5.49376383e-02 -1.63080180e+00 -6.68660343e-01 4.05198559e-02 -1.14526820e+00 5.65113604e-01 -1.87965035e-01 1.33932030e+00 -3.92693639e-01 -1.32864758e-01 8.78702939e-01 1.40887165e+00 -3.75606388e-01 -8.20796341e-02 2.60276169e-01 3.66432101e-01 2.92711705e-01 5.52147686e-01 9.94557977e-01 3.82356286e-01 1.48457646e-01 7.93022633e-01 3.17680612e-02 -1.57286644e-01 -2.81850584e-02 3.51650655e-01 4.92462307e-01 6.16095603e-01 -4.24282789e-01 -7.99523413e-01 3.69501770e-01 -1.64064121e+00 -8.43633950e-01 -1.97878793e-01 2.01712441e+00 9.59919691e-01 -5.44341505e-01 1.45439595e-01 1.47934452e-01 1.01273918e+00 1.62479714e-01 -9.53580976e-01 -9.10138618e-03 -3.21577162e-01 4.04564571e-03 5.54865301e-01 4.23527099e-02 -9.85884607e-01 6.36942983e-01 5.95432186e+00 6.59274757e-01 -1.08619750e+00 6.52033627e-01 7.69662619e-01 -6.19405985e-01 -3.35929662e-01 -1.61152333e-01 -6.84411287e-01 6.80210888e-01 9.76597965e-01 -3.88293028e-01 7.41585195e-01 1.12508547e+00 8.07264447e-02 8.39947090e-02 -1.12643230e+00 9.88946319e-01 -2.18961641e-01 -9.53605890e-01 -4.58786547e-01 1.03923284e-01 1.10080707e+00 7.67706931e-01 2.97851264e-01 3.35240275e-01 1.11574817e+00 -6.98289394e-01 5.12992978e-01 2.56717056e-01 8.93472135e-01 -7.21730649e-01 7.58083463e-01 4.53767687e-01 -8.37917089e-01 -3.39642406e-01 -4.26936805e-01 1.48304120e-01 -2.32981876e-01 1.11161256e+00 -4.39499438e-01 3.86945546e-01 1.14342165e+00 2.83272862e-01 -7.41215110e-01 1.31941986e+00 1.46997049e-01 1.08895695e+00 -6.20604515e-01 9.88248363e-02 -1.75308809e-01 2.50946611e-01 2.23457858e-01 1.11366415e+00 2.27864519e-01 -1.61119878e-01 2.53047854e-01 6.59133852e-01 -7.65896022e-01 4.33352664e-02 -6.92690969e-01 2.81659305e-01 1.03885770e+00 1.61722291e+00 -2.31283650e-01 -8.49931911e-02 -4.16719407e-01 4.71827626e-01 3.92031610e-01 4.17318404e-01 -7.95806885e-01 -1.14622258e-01 1.05154026e+00 -2.77171791e-01 -5.26621714e-02 2.12900743e-01 -2.90190518e-01 -1.31622553e+00 -1.39771953e-01 -1.33892369e+00 1.06430793e+00 -2.21055076e-01 -1.97546732e+00 5.23014843e-01 -6.19356036e-01 -9.57856059e-01 1.97535157e-01 2.07006037e-02 -6.06229901e-01 5.31119823e-01 -1.49927318e+00 -1.02389729e+00 -6.52898967e-01 1.02300274e+00 9.75494385e-02 -5.88488542e-02 9.65344667e-01 6.75933301e-01 -8.35269868e-01 1.33060324e+00 7.01895535e-01 7.87123814e-02 1.46338975e+00 -1.01509607e+00 7.69257545e-02 6.65430546e-01 1.32867202e-01 3.96930844e-01 4.18012112e-01 -4.62783605e-01 -1.55698502e+00 -1.50375879e+00 4.69634324e-01 -6.98129952e-01 6.36443436e-01 -5.62879503e-01 -7.22263098e-01 6.59133017e-01 -3.47625911e-02 7.80128121e-01 1.07954586e+00 2.54647464e-01 -8.48734438e-01 -4.91017282e-01 -1.73341084e+00 1.17885403e-01 1.11406362e+00 -6.16903543e-01 4.15400267e-01 8.26575220e-01 8.15323174e-01 2.80490778e-02 -1.01011980e+00 2.02631965e-01 2.38784552e-01 -1.23146856e+00 4.62501079e-01 -4.29673582e-01 -3.89363974e-01 -2.74414837e-01 -3.54847759e-01 -1.15243852e+00 -2.28294268e-01 -9.33518648e-01 -2.70404816e-01 1.82800269e+00 2.94788927e-01 -7.66694129e-01 9.21154618e-01 7.44381011e-01 3.36805314e-01 -5.42028844e-01 -8.39404702e-01 -8.28329206e-01 5.58587201e-02 -5.54708421e-01 1.17539334e+00 1.37062967e+00 1.27506644e-01 -3.41152847e-01 -2.22305134e-01 2.29764223e-01 1.10708582e+00 7.89314806e-02 8.05550873e-01 -8.71369243e-01 -4.63809967e-02 -1.35379568e-01 -6.46045208e-02 -5.03768563e-01 2.15963513e-01 -1.01222587e+00 -6.90683499e-02 -9.06405926e-01 3.64399374e-01 -1.01026273e+00 -6.98071063e-01 9.48193371e-01 -1.97599754e-01 4.46565866e-01 3.28915626e-01 5.06232381e-01 -1.43436205e+00 4.34333622e-01 8.25680733e-01 -1.86621606e-01 3.03101271e-01 9.58514437e-02 -9.25903440e-01 1.67913333e-01 1.08521461e+00 -8.66729438e-01 -2.43030697e-01 -5.19925416e-01 -2.82383338e-02 -2.48861060e-01 9.08494294e-02 -9.97469902e-01 5.39724231e-01 -1.77479282e-01 -5.18905111e-02 -2.12403342e-01 -3.53370428e-01 -1.02686489e+00 1.90177843e-01 1.53472558e-01 -5.44608474e-01 -2.45506503e-02 -2.79574811e-01 5.28515637e-01 3.38681377e-02 -6.27903044e-02 7.22382784e-01 -4.20393571e-02 -2.17134893e-01 5.07456958e-01 -7.86116347e-02 4.85093713e-01 1.07922649e+00 3.94610912e-01 -9.73956943e-01 -5.66333950e-01 -2.84035832e-01 7.24844098e-01 7.61086226e-01 1.69750780e-01 4.63464633e-02 -1.20528638e+00 -6.59505963e-01 3.75409275e-02 2.06476852e-01 -7.44643214e-04 3.00058812e-01 6.36118829e-01 -1.37668222e-01 1.17902681e-01 1.63704425e-01 -3.28662306e-01 -1.33551049e+00 7.82358646e-01 5.19263566e-01 -1.48810431e-01 -3.54558736e-01 7.90229380e-01 -3.06041837e-01 -9.63595629e-01 1.04841292e+00 1.19201683e-01 5.62845469e-01 -2.56586909e-01 7.22731829e-01 7.59697855e-01 4.01726335e-01 -2.80043840e-01 -3.92937660e-01 -1.94203511e-01 1.33174002e-01 3.65244746e-01 1.24322474e+00 -2.92606115e-01 -8.03470239e-02 2.73811400e-01 1.32197654e+00 4.96970192e-02 -1.30799091e+00 -6.53477550e-01 2.29385737e-02 -4.09831405e-01 -4.17763852e-02 -1.26512969e+00 -1.58880842e+00 3.75732481e-01 9.17559981e-01 2.76276588e-01 1.20849299e+00 3.35867666e-02 1.05740058e+00 5.36267340e-01 8.04039896e-01 -9.27735388e-01 -4.81382981e-02 3.75953913e-01 1.91910431e-01 -1.32846987e+00 -2.36318886e-01 -6.97542951e-02 -3.65742832e-01 7.43246615e-01 7.66017497e-01 3.97582531e-01 7.60250390e-01 8.03178787e-01 8.20725918e-01 -2.68923998e-01 -1.14642131e+00 -4.03742678e-02 -7.02225626e-01 6.25840664e-01 3.21656674e-01 8.25252980e-02 -5.25298864e-02 8.27762485e-01 -5.99660203e-02 5.52618988e-02 3.39955807e-01 1.05560815e+00 -2.01494023e-01 -1.32565618e+00 -9.11383569e-01 3.93313169e-01 -7.82478213e-01 1.54806703e-01 -5.99064350e-01 1.31093934e-01 4.39944983e-01 1.56174421e+00 -4.25528049e-01 -4.91142094e-01 1.26286998e-01 3.17101508e-01 -8.77602100e-02 -2.93130606e-01 -1.16434133e+00 -2.18819097e-01 -1.36350349e-01 -8.08979094e-01 -2.00396910e-01 -4.34648663e-01 -1.05001819e+00 -8.97282720e-01 -5.98643541e-01 4.98492271e-01 8.23427737e-01 4.54699010e-01 5.66149056e-01 -1.13911897e-01 1.18698430e+00 -2.35238820e-01 -1.16615283e+00 -8.12969863e-01 -9.12606418e-01 7.53150702e-01 2.06502169e-01 -3.28842968e-01 -9.28646386e-01 -2.34940007e-01]
[5.8401570320129395, 6.433576583862305]
978010a2-240d-437c-9514-d36a97eee06a
portfolio-optimization-using-a-consistent
2204.05611
null
https://arxiv.org/abs/2204.05611v1
https://arxiv.org/pdf/2204.05611v1.pdf
Portfolio Optimization Using a Consistent Vector-Based MSE Estimation Approach
This paper is concerned with optimizing the global minimum-variance portfolio's (GMVP) weights in high-dimensional settings where both observation and population dimensions grow at a bounded ratio. Optimizing the GMVP weights is highly influenced by the data covariance matrix estimation. In a high-dimensional setting, it is well known that the sample covariance matrix is not a proper estimator of the true covariance matrix since it is not invertible when we have fewer observations than the data dimension. Even with more observations, the sample covariance matrix may not be well-conditioned. This paper determines the GMVP weights based on a regularized covariance matrix estimator to overcome the aforementioned difficulties. Unlike other methods, the proper selection of the regularization parameter is achieved by minimizing the mean-squared error of an estimate of the noise vector that accounts for the uncertainty in the data mean estimation. Using random-matrix-theory tools, we derive a consistent estimator of the achievable mean-squared error that allows us to find the optimal regularization parameter using a simple line search. Simulation results demonstrate the effectiveness of the proposed method when the data dimension is larger than the number of data samples or of the same order.
['Ubaid Al-Saggaf', 'Tareq Y. Al-Naffouri', 'Muhammad Moinuddin', 'Tarig Ballal', 'Maaz Mahadi']
2022-04-12
null
null
null
null
['portfolio-optimization']
['time-series']
[ 3.85142607e-03 7.39462823e-02 -2.32290804e-01 -3.14214565e-02 -8.92999470e-01 -1.52181387e-01 4.11497056e-02 -1.08182937e-01 -6.33518696e-01 7.99879730e-01 7.34155923e-02 -3.07532042e-01 -6.53000474e-01 -4.09893245e-01 -4.80924398e-01 -1.17404902e+00 -5.11791445e-02 2.81461895e-01 -3.20397258e-01 1.75841004e-01 2.56536186e-01 1.59418017e-01 -1.06480813e+00 -8.07716250e-01 1.17319083e+00 1.05215108e+00 -5.82213663e-02 5.22626102e-01 2.85196215e-01 -1.53094223e-02 -4.54223752e-01 -2.99330384e-01 7.04159081e-01 -4.77465808e-01 3.63017917e-02 5.03014505e-01 3.10616672e-01 -1.45112559e-01 5.64488508e-02 1.43566298e+00 5.21991074e-01 2.22949728e-01 8.79150927e-01 -9.04001892e-01 -1.05783463e-01 2.48850361e-01 -8.31415892e-01 1.56974345e-01 -6.28338531e-02 -2.56973468e-02 7.46845484e-01 -9.56718087e-01 2.83528477e-01 1.08162868e+00 8.19247246e-01 4.06277657e-01 -1.52259040e+00 -7.82015741e-01 6.94232136e-02 -4.60819632e-01 -1.84547222e+00 -5.04542947e-01 6.00623608e-01 -8.55322003e-01 9.19719860e-02 4.66626212e-02 3.79456401e-01 7.54078567e-01 1.23621263e-01 1.29225388e-01 6.89758778e-01 -4.85998034e-01 5.03379107e-01 4.70075876e-01 3.98265302e-01 6.15968227e-01 1.00792766e+00 1.87050670e-01 -1.82164982e-01 -5.45910478e-01 8.84287596e-01 -1.15729734e-01 -3.80689949e-01 -8.44828010e-01 -8.42270911e-01 1.13151586e+00 -2.76632726e-01 2.47572049e-01 -4.50623393e-01 -4.89551984e-02 1.62553370e-01 2.77191103e-01 5.44517219e-01 3.18391025e-01 -2.46926725e-01 -5.91709912e-02 -8.74193728e-01 1.63482606e-01 8.49071801e-01 8.57316792e-01 3.86004031e-01 2.37385660e-01 -1.13515127e-02 7.57748842e-01 4.50990111e-01 1.07950258e+00 1.67359307e-01 -9.94667590e-01 9.43858981e-01 4.13804680e-01 5.70004940e-01 -1.06624460e+00 -5.13968945e-01 -7.26775050e-01 -1.19750321e+00 6.39496073e-02 9.99236405e-01 -9.30902660e-01 -3.95569175e-01 1.93568730e+00 5.38804710e-01 7.15478463e-03 1.34344205e-01 1.03067958e+00 4.40896340e-02 2.52074182e-01 -2.84958899e-01 -7.68638730e-01 9.09715593e-01 -2.20308438e-01 -1.06380296e+00 -4.72494274e-01 6.54601991e-01 -5.10214031e-01 8.32125962e-01 2.58703738e-01 -8.46287608e-01 -8.57769996e-02 -1.14350510e+00 5.80117226e-01 2.00063795e-01 3.52508724e-01 2.02974409e-01 1.09289336e+00 -4.73084092e-01 3.62896681e-01 -7.21206665e-01 -2.40601003e-01 1.34910187e-02 3.73588353e-01 -2.04271257e-01 3.32824647e-01 -9.74861443e-01 7.01350629e-01 1.69108793e-01 4.17413414e-01 -1.55388176e-01 -9.38831806e-01 -6.41141832e-01 3.08582801e-02 3.51598680e-01 -6.00062907e-01 6.72999978e-01 -6.50569797e-01 -1.37922847e+00 3.29980373e-01 -1.70925513e-01 -2.80731529e-01 6.86563969e-01 -1.91865504e-01 -2.41797686e-01 -1.38337716e-01 6.98508173e-02 -5.01092613e-01 1.23846030e+00 -8.45328987e-01 -3.65451723e-01 -6.41158223e-01 -6.33174777e-01 1.76542178e-01 -3.46748769e-01 -3.61176997e-01 -4.06643689e-01 -4.47922677e-01 3.71789366e-01 -9.75990117e-01 -6.54012024e-01 -7.52318650e-02 -5.43223441e-01 4.81539816e-01 3.82533133e-01 -4.57424104e-01 1.60164964e+00 -2.48934913e+00 3.98750566e-02 7.80285239e-01 3.04006543e-02 1.11154497e-01 9.86586437e-02 4.16882873e-01 1.16341347e-02 -1.05050787e-01 -3.67776573e-01 -2.40351558e-01 -1.49306163e-01 -1.68241173e-01 -1.16231643e-01 1.17592692e+00 -9.29751843e-02 3.08833066e-02 -6.82381690e-01 -1.88675910e-01 4.23471779e-02 4.11491156e-01 -8.04708660e-01 3.34056616e-02 5.69746792e-01 4.38013315e-01 -8.43696892e-01 7.34421238e-02 9.08993006e-01 -4.68100488e-01 4.99644369e-01 -1.77181333e-01 -9.03632343e-02 -4.28972512e-01 -2.10597968e+00 1.06571126e+00 -3.76558721e-01 5.95068991e-01 3.95161003e-01 -1.07364392e+00 8.90361071e-01 2.11714208e-01 6.65672958e-01 -2.14389488e-01 2.86713213e-01 2.75402039e-01 7.81246051e-02 -5.08010268e-01 1.51570886e-01 -4.43607420e-01 -5.21172164e-03 2.62482911e-01 -3.28601390e-01 1.86779037e-01 3.06299746e-01 -3.54596823e-02 6.56870663e-01 -7.14694619e-01 6.77498281e-01 -6.73012495e-01 6.13599241e-01 -5.44881642e-01 8.12580705e-01 8.42804313e-01 -4.56604548e-02 5.67574084e-01 7.51871467e-01 7.04164878e-02 -1.01977086e+00 -1.04183316e+00 -6.90284610e-01 8.27478319e-02 -9.90982354e-02 -1.41966209e-01 -7.07105637e-01 -3.04123312e-01 3.59122753e-01 5.10065317e-01 -7.67813742e-01 -5.50929755e-02 -1.51278108e-01 -1.30530047e+00 5.65595292e-02 -8.57388787e-03 3.01989257e-01 1.85478866e-01 -2.40157127e-01 2.18303069e-01 1.61899149e-01 -9.41135824e-01 -8.18688154e-01 -6.60272092e-02 -9.12080467e-01 -1.28850996e+00 -9.08003509e-01 -2.53770471e-01 1.23457980e+00 6.94889799e-02 5.02269447e-01 -2.91719705e-01 6.77610338e-02 4.04781908e-01 2.01175869e-01 -4.46898282e-01 -3.12560230e-01 -5.10412902e-02 4.89323735e-01 5.17717779e-01 2.06766367e-01 -2.76332796e-01 -5.14729083e-01 3.76657903e-01 -4.61612135e-01 -4.60685760e-01 3.37092847e-01 1.08901143e+00 4.90345418e-01 1.40734583e-01 7.69236028e-01 -9.04281676e-01 1.05023181e+00 -4.80806619e-01 -1.26941502e+00 1.72475114e-01 -9.37089503e-01 2.56062984e-01 5.05222261e-01 -7.24825621e-01 -9.49534059e-01 1.01814047e-02 3.64911795e-01 -3.19767177e-01 3.93958032e-01 5.87264538e-01 5.28259315e-02 -6.59156367e-02 5.97336948e-01 -7.18246475e-02 3.05811793e-01 -5.10458589e-01 2.08986819e-01 7.27042079e-01 1.11964606e-02 -3.39045048e-01 8.39515924e-01 3.04135352e-01 5.04235923e-01 -1.07959080e+00 -9.59062159e-01 -5.25252819e-01 -4.62266743e-01 3.25798094e-02 6.24303937e-01 -8.87082696e-01 -9.67788160e-01 1.50623307e-01 -4.75814879e-01 -7.61856213e-02 -3.96290362e-01 1.10127807e+00 -4.67731893e-01 3.71993929e-01 -1.71174482e-01 -1.32509208e+00 -2.53965050e-01 -1.04593861e+00 7.46008992e-01 -4.17165533e-02 -1.06338866e-01 -1.23128915e+00 2.12850764e-01 -7.73347989e-02 1.77520499e-01 1.60495803e-01 6.59747243e-01 -5.13573527e-01 -1.48179591e-01 -6.16949856e-01 -6.71821833e-02 3.70665073e-01 1.96479917e-01 -1.57884538e-01 -3.40980440e-01 -5.45329213e-01 4.78017300e-01 3.09219629e-01 2.73532391e-01 1.05778754e+00 7.24633217e-01 -5.26903510e-01 -2.08401963e-01 6.86555147e-01 1.56428635e+00 1.37988523e-01 1.30919501e-01 1.89632922e-01 4.21055704e-01 4.39802676e-01 6.01223409e-01 1.12524068e+00 -2.93097668e-03 5.98262250e-01 -1.42069042e-01 1.40919492e-01 4.27542299e-01 5.50046861e-02 1.90634593e-01 8.73940349e-01 2.15645209e-01 2.89597083e-02 -7.29646206e-01 2.72383362e-01 -1.87473047e+00 -8.25412750e-01 -3.68175924e-01 3.03148293e+00 6.54381812e-01 -1.56193078e-01 -1.73645988e-02 8.87799412e-02 6.40731096e-01 -2.03271229e-02 -7.74753451e-01 -9.17341784e-02 -4.98826094e-02 -4.59146142e-01 1.17233121e+00 7.80662239e-01 -9.91118550e-01 -4.68027452e-03 6.92519522e+00 6.43331468e-01 -5.88327110e-01 -1.47086933e-01 3.41637760e-01 -7.33997077e-02 3.94493043e-02 -1.44789919e-01 -1.21418738e+00 6.87646627e-01 9.33339298e-01 -5.58250964e-01 3.48817080e-01 6.56950712e-01 6.52589679e-01 -3.65123332e-01 -7.82920539e-01 1.15224123e+00 -5.36533538e-03 -9.80458021e-01 -4.80286002e-01 4.76259500e-01 9.05869365e-01 -3.90648633e-01 3.45452189e-01 -8.33530948e-02 -3.39333564e-02 -4.84880060e-01 2.00156793e-01 6.76248252e-01 6.75842524e-01 -8.87735367e-01 7.13428080e-01 4.75402266e-01 -9.13887978e-01 -3.70879233e-01 -6.47656322e-01 1.69410735e-01 1.33642271e-01 1.30124807e+00 -5.93362451e-01 1.00162789e-01 1.84876576e-01 5.46625197e-01 -1.27262831e-01 1.19923782e+00 1.66495532e-01 4.44633096e-01 -6.89483106e-01 5.31384721e-02 -6.59339279e-02 -1.05902874e+00 7.71882057e-01 8.04849029e-01 7.46971786e-01 8.77865776e-02 5.21897664e-03 5.36080122e-01 1.27363801e-01 4.19917792e-01 -4.96934712e-01 -1.08867168e-01 5.25033712e-01 7.41456270e-01 -2.68582433e-01 -5.94868371e-03 -5.37866831e-01 3.81407052e-01 5.47910705e-02 7.41606236e-01 -3.39977384e-01 -3.42540383e-01 9.25093114e-01 4.46219929e-02 2.51072973e-01 -2.67792284e-01 -3.32273155e-01 -1.20477152e+00 9.33397785e-02 -8.24829757e-01 3.52802068e-01 9.43292379e-02 -1.38659513e+00 1.42519131e-01 3.35173187e-04 -1.50415814e+00 -5.00103056e-01 -4.22802448e-01 -1.86824486e-01 1.08247817e+00 -7.92296171e-01 -1.18417338e-01 3.53677392e-01 3.12739879e-01 7.22815692e-02 -3.11235875e-01 6.16747320e-01 3.74943316e-01 -1.26337504e+00 6.82121038e-01 1.07540202e+00 -4.78306692e-03 4.73521501e-01 -9.84566391e-01 -2.39377320e-01 9.65339959e-01 -5.51274896e-01 8.55863214e-01 1.17923594e+00 -6.25698984e-01 -1.49868894e+00 -7.52930999e-01 5.86885810e-01 -1.49215609e-01 9.39277112e-01 -3.98039579e-01 -7.20572889e-01 5.92103124e-01 -4.86533165e-01 1.35118932e-01 7.23554611e-01 2.86314189e-01 1.33196980e-01 -2.79935986e-01 -1.18149674e+00 6.60289645e-01 5.63403249e-01 -2.40576357e-01 -1.47202894e-01 4.78921801e-01 1.27751634e-01 -3.26084882e-01 -1.03287816e+00 3.39094430e-01 5.76572418e-01 -6.63334012e-01 9.78802860e-01 -5.91289580e-01 -3.83823305e-01 -1.86924428e-01 -3.03784788e-01 -1.28054929e+00 -1.65182292e-01 -6.39994085e-01 -5.48712872e-02 1.08452809e+00 5.88077962e-01 -9.31210995e-01 9.67837512e-01 1.04630601e+00 6.62063003e-01 -6.27013803e-01 -1.05317175e+00 -1.13297844e+00 -7.03143654e-03 -2.92521566e-01 1.36483625e-01 6.31694973e-01 4.48808298e-02 3.93729657e-01 -6.99775577e-01 5.84448516e-01 1.21477056e+00 -1.69644535e-01 8.15436184e-01 -1.28774059e+00 -3.60361964e-01 -2.20347896e-01 -3.16728741e-01 -1.04331517e+00 3.93536314e-02 -3.69520068e-01 3.53002436e-02 -1.06259799e+00 1.46723926e-01 -4.68926907e-01 8.36974159e-02 -5.43289840e-01 -2.87015647e-01 -4.06377882e-01 4.00979482e-02 -2.33412981e-02 -4.44826856e-02 5.32682002e-01 1.16503406e+00 1.52030855e-01 -5.91602087e-01 6.41976833e-01 -6.06400132e-01 8.12941790e-01 6.26994491e-01 -6.61909521e-01 -7.99189866e-01 7.48305842e-02 2.86272526e-01 3.83525789e-01 -2.36886919e-01 -8.58968377e-01 1.24008030e-01 -2.89144933e-01 2.14833423e-01 -3.93938482e-01 2.62750179e-01 -1.06177688e+00 3.81543338e-01 3.45017672e-01 -4.46281344e-01 -3.65321100e-01 -1.61516756e-01 8.55210721e-01 1.24070287e-01 -6.88204408e-01 9.18893993e-01 3.36288393e-01 1.27861366e-01 1.91460788e-01 -5.54626822e-01 3.04368705e-01 8.46183538e-01 -2.92539336e-02 -1.33855924e-01 -6.26113415e-01 -6.85457826e-01 2.23228425e-01 1.65851742e-01 -1.29509687e-01 4.22071159e-01 -1.26009476e+00 -6.08959138e-01 4.20055240e-01 -1.35573000e-01 -4.16924864e-01 2.75341243e-01 1.23913491e+00 -4.48653735e-02 3.66795480e-01 6.40227735e-01 -4.68852848e-01 -1.12271523e+00 6.38182163e-01 5.64552069e-01 -1.38332158e-01 -5.35520375e-01 5.18143535e-01 2.86454529e-01 -1.80067390e-01 3.68990064e-01 -1.76329777e-01 -3.02013047e-02 1.65721253e-01 7.95163095e-01 6.25711143e-01 -1.67257622e-01 -5.05869746e-01 -1.37535110e-01 9.58893418e-01 2.50844717e-01 -1.32119536e-01 1.04803443e+00 -6.89537704e-01 2.24227890e-01 5.65571666e-01 1.56437874e+00 2.21813649e-01 -1.39419782e+00 -3.73295486e-01 7.58178532e-03 -7.06679285e-01 1.25885472e-01 -1.06441572e-01 -1.20249152e+00 6.13107681e-01 8.87286305e-01 2.18618900e-01 9.05853689e-01 -3.89080167e-01 1.19367816e-01 4.38868999e-01 1.46679431e-01 -1.28534102e+00 -3.25801700e-01 2.65456289e-01 8.38528872e-01 -1.40969968e+00 4.44377422e-01 -5.27691245e-01 -6.57799244e-01 8.58996928e-01 2.99068570e-01 -2.28411421e-01 1.09725595e+00 1.49543360e-01 4.83964533e-02 5.78797087e-02 -4.80030000e-01 -6.96493387e-02 5.38608372e-01 5.29199600e-01 1.53984621e-01 1.74753234e-01 -7.58169830e-01 5.00199437e-01 5.54396659e-02 -1.84669659e-01 7.40273297e-01 4.26740974e-01 -4.89066184e-01 -7.89107680e-01 -6.96437418e-01 7.42919981e-01 -5.12642562e-01 1.11898161e-01 2.46589914e-01 6.99596643e-01 -4.20113176e-01 1.08126295e+00 1.13018155e-01 1.31573066e-01 5.55668592e-01 -1.43850610e-01 2.58110762e-01 -3.26626658e-01 1.05843768e-01 3.12085837e-01 8.89261514e-02 -3.94082606e-01 -2.95176595e-01 -8.74559700e-01 -8.62026393e-01 -4.21996534e-01 -5.15357792e-01 4.06512558e-01 6.85211182e-01 8.90000045e-01 5.02537303e-02 1.11621715e-01 8.56120229e-01 -2.73293227e-01 -1.03266263e+00 -6.58636630e-01 -1.24628937e+00 1.35065570e-01 6.14891350e-01 -6.35289967e-01 -9.26699877e-01 -3.33447486e-01]
[7.06342077255249, 4.344584941864014]
d66c8b66-0b46-4b49-a2f7-a3f6d380b136
codesc-a-large-code-description-parallel
2105.14220
null
https://arxiv.org/abs/2105.14220v1
https://arxiv.org/pdf/2105.14220v1.pdf
CoDesc: A Large Code-Description Parallel Dataset
Translation between natural language and source code can help software development by enabling developers to comprehend, ideate, search, and write computer programs in natural language. Despite growing interest from the industry and the research community, this task is often difficult due to the lack of large standard datasets suitable for training deep neural models, standard noise removal methods, and evaluation benchmarks. This leaves researchers to collect new small-scale datasets, resulting in inconsistencies across published works. In this study, we present CoDesc -- a large parallel dataset composed of 4.2 million Java methods and natural language descriptions. With extensive analysis, we identify and remove prevailing noise patterns from the dataset. We demonstrate the proficiency of CoDesc in two complementary tasks for code-description pairs: code summarization and code search. We show that the dataset helps improve code search by up to 22\% and achieves the new state-of-the-art in code summarization. Furthermore, we show CoDesc's effectiveness in pre-training--fine-tuning setup, opening possibilities in building pretrained language models for Java. To facilitate future research, we release the dataset, a data processing tool, and a benchmark at \url{https://github.com/csebuetnlp/CoDesc}.
['Rifat Shahriyar', 'Anindya Iqbal', 'Wasi Uddin Ahmad', 'Tahmid Hasan', 'Md. Mahim Anjum Haque', 'Kazi Sajeed Mehrab', 'Abdullah Al Ishtiaq', 'Tanveer Muttaqueen', 'Masum Hasan']
2021-05-29
null
null
null
null
['code-summarization', 'code-search', 'code-search']
['computer-code', 'computer-code', 'computer-vision']
[ 1.58770978e-01 -1.12534262e-01 -4.48062778e-01 -3.56973052e-01 -1.01236403e+00 -5.87511063e-01 2.54997939e-01 3.87187630e-01 -2.03675941e-01 3.21558356e-01 3.65419954e-01 -4.46402669e-01 1.87371150e-01 -4.73296493e-01 -7.24144101e-01 -1.11040741e-01 4.76397248e-03 5.79182841e-02 -8.81655794e-03 -1.43716633e-01 6.65521026e-01 -4.34545539e-02 -1.69142950e+00 7.12935150e-01 1.13467443e+00 3.45808953e-01 5.82745910e-01 6.61843002e-01 -5.10624409e-01 8.67366135e-01 -7.36787617e-01 -5.55734098e-01 -1.04514798e-02 -3.98272872e-01 -1.00792348e+00 -4.35686737e-01 5.81653714e-01 -9.00091901e-02 2.92071197e-02 1.18307948e+00 6.41716719e-01 -2.21927077e-01 1.94549009e-01 -1.17724800e+00 -1.01717222e+00 1.02338648e+00 -7.09689319e-01 -5.61094172e-02 4.15591955e-01 1.69900224e-01 1.02939785e+00 -7.82428086e-01 8.24863613e-01 8.63595426e-01 8.76747131e-01 1.00958443e+00 -1.13673878e+00 -6.33662760e-01 -2.67536670e-01 -6.89092949e-02 -1.01067531e+00 -7.50325978e-01 6.55737817e-01 -6.33848131e-01 1.76356828e+00 3.32034022e-01 3.57621878e-01 1.24321830e+00 1.19271263e-01 7.59595394e-01 5.29714763e-01 -5.79086721e-01 -1.52670685e-02 7.26433322e-02 4.07122165e-01 9.23239291e-01 3.95223558e-01 -4.20858264e-01 -4.21365917e-01 -2.88816512e-01 1.28027618e-01 -1.55243859e-01 -3.13317269e-01 -1.68683857e-01 -1.24573779e+00 7.25604653e-01 1.27550647e-01 4.60284740e-01 -7.68777877e-02 5.61574697e-01 1.00362611e+00 4.60864067e-01 3.43052357e-01 9.36569154e-01 -6.82614148e-01 -8.33219886e-01 -8.79651308e-01 3.52418810e-01 1.11224496e+00 1.47823060e+00 5.96607924e-01 1.33629918e-01 -5.48751615e-02 1.23475087e+00 4.66044135e-02 4.85087186e-01 8.04295301e-01 -1.18685949e+00 8.44530702e-01 9.09881949e-01 -4.57832664e-01 -1.03625774e+00 -2.21014246e-01 -3.97246897e-01 -8.39095712e-01 -1.23890229e-01 1.90859996e-02 3.22853774e-02 -4.26295549e-01 1.59075034e+00 -3.44125926e-01 -4.62651521e-01 1.12120546e-01 3.33369166e-01 1.19478357e+00 5.45115829e-01 -3.98785114e-01 -1.37636438e-01 1.16632807e+00 -1.23253381e+00 -4.67275828e-01 -4.60381329e-01 1.07237267e+00 -9.90574181e-01 1.33933914e+00 2.19512537e-01 -1.14943171e+00 -4.43811327e-01 -8.85099292e-01 -3.81579936e-01 -2.84090132e-01 3.10909271e-01 6.49144292e-01 3.50424111e-01 -1.38098538e+00 4.86702114e-01 -9.80048537e-01 -5.67065358e-01 5.06385207e-01 5.33486754e-02 -2.62191355e-01 -2.93592289e-02 -5.44892132e-01 6.17156982e-01 3.55106294e-01 -4.01375145e-01 -6.63755953e-01 -1.01810992e+00 -9.09153998e-01 1.39627457e-01 2.79713422e-01 -5.70003808e-01 1.60002685e+00 -8.44731808e-01 -1.15070307e+00 1.04985893e+00 -1.70893610e-01 -5.36326468e-01 1.58538550e-01 -3.35749775e-01 -2.08646953e-01 -2.12960809e-01 4.50896740e-01 6.58003211e-01 3.54947448e-01 -1.09005725e+00 -4.45401818e-01 1.00715794e-02 9.56585631e-02 -3.38864803e-01 -4.21358436e-01 3.94273609e-01 -8.14691067e-01 -7.51773179e-01 -4.86081809e-01 -8.34893107e-01 -1.13560349e-01 -3.94197494e-01 -5.40324807e-01 -2.58658588e-01 4.25175339e-01 -7.49734581e-01 1.70426524e+00 -2.25118613e+00 1.13809943e-01 -2.63852209e-01 4.46255475e-01 4.07162964e-01 -5.37379801e-01 8.02591860e-01 -1.61737844e-01 5.32322884e-01 -3.76545221e-01 -4.41570789e-01 1.68262765e-01 2.17671413e-02 -2.80757636e-01 1.65275738e-01 2.46911705e-01 9.59420919e-01 -8.39203119e-01 -4.65541840e-01 -2.55403191e-01 1.74943045e-01 -1.01560152e+00 1.06391519e-01 -4.32508558e-01 -1.77480876e-01 -3.20618331e-01 7.12368071e-01 4.13495868e-01 -4.17536288e-01 -6.50699660e-02 2.81584859e-01 -2.50984818e-01 5.73310912e-01 -6.55780196e-01 2.15377736e+00 -7.92038918e-01 1.20398080e+00 -6.44090697e-02 -8.87595236e-01 8.29215467e-01 6.80531338e-02 2.28448138e-01 -8.77422214e-01 -7.32077286e-02 5.15613079e-01 -1.97452065e-02 -8.30415726e-01 7.12876081e-01 6.12914681e-01 -4.61328655e-01 8.14138234e-01 3.09648514e-02 -3.03840339e-01 7.86849678e-01 3.21869135e-01 1.42499828e+00 -5.83791621e-02 3.44464302e-01 -4.96033937e-01 4.28245038e-01 2.75432557e-01 4.47296441e-01 8.86785865e-01 1.59959823e-01 5.53207099e-01 8.06321442e-01 -5.25226653e-01 -1.06180036e+00 -4.23432827e-01 2.74027549e-02 1.29252648e+00 -3.35174620e-01 -9.92509067e-01 -1.08092701e+00 -6.62067175e-01 -3.33769582e-02 8.01952243e-01 -5.89836895e-01 -2.78186798e-01 -7.72685289e-01 -4.14888978e-01 1.02360308e+00 4.52158540e-01 4.48355407e-01 -1.21754694e+00 -6.91785216e-01 1.09617405e-01 -2.73551881e-01 -8.15880299e-01 -7.12359488e-01 2.39217728e-01 -7.12576807e-01 -1.14133370e+00 -4.11172807e-01 -1.15360296e+00 5.84731162e-01 3.32975298e-01 1.70321429e+00 6.95802748e-01 -3.22031021e-01 2.46470228e-01 -5.46292782e-01 -3.30465198e-01 -1.22739291e+00 6.25400245e-01 -3.19882691e-01 -1.05298555e+00 4.47761208e-01 -6.38660073e-01 -3.72587800e-01 -8.64204690e-02 -9.32877839e-01 4.13723826e-01 5.89398205e-01 6.92060351e-01 7.68762082e-02 -2.28344604e-01 4.84843165e-01 -8.92554402e-01 9.69444871e-01 -4.69012618e-01 -7.66996503e-01 3.40568811e-01 -7.20437229e-01 2.59952843e-01 7.69984782e-01 -2.11007163e-01 -8.37885499e-01 -1.70216233e-01 -2.97558814e-01 3.15219685e-02 -1.16695210e-01 9.03992414e-01 1.85274899e-01 -1.88962161e-03 1.04714060e+00 3.60600978e-01 1.23195283e-01 -6.63921356e-01 3.07125479e-01 8.69601369e-01 4.48087037e-01 -8.03520620e-01 5.25959730e-01 -9.43511873e-02 -6.43219352e-01 -6.68896735e-01 -4.88043308e-01 -3.86755407e-01 -3.05559456e-01 2.96407074e-01 5.96477866e-01 -8.64371121e-01 -3.62690836e-01 3.91219497e-01 -1.61242712e+00 -5.44128895e-01 -4.06007208e-02 -1.21620568e-02 -3.71851206e-01 4.06360269e-01 -7.79876351e-01 -1.00802839e-01 -6.48717225e-01 -1.58197892e+00 1.05601966e+00 1.24317199e-01 -4.73367929e-01 -8.35158765e-01 3.76407564e-01 3.31516594e-01 7.81809747e-01 1.19110994e-01 1.13001657e+00 -7.45236456e-01 -5.51771343e-01 -4.40381467e-02 -2.66111046e-01 4.99430984e-01 1.11096025e-01 4.80977327e-01 -6.97354972e-01 -3.60758275e-01 -2.08154842e-01 -4.60248202e-01 9.67811584e-01 -3.45972739e-02 1.52094328e+00 -4.15896446e-01 -2.88576573e-01 6.74463451e-01 1.52272797e+00 4.10330705e-02 4.24054235e-01 3.80570292e-01 7.80851245e-01 4.00822788e-01 4.30133417e-02 5.57134390e-01 4.81104910e-01 5.62106490e-01 2.90301919e-01 2.66865492e-01 -2.44848639e-01 8.61695260e-02 4.06064928e-01 1.59615278e+00 1.19040594e-01 -1.31634161e-01 -1.48910582e+00 7.22025275e-01 -1.70686376e+00 -7.81111121e-01 -2.48414665e-01 1.88748157e+00 1.42225575e+00 -4.44122702e-02 -1.65383071e-01 -3.70747954e-01 4.00393367e-01 -4.38867919e-02 -3.08577776e-01 -6.50605559e-01 1.76392913e-01 1.88321948e-01 3.00461113e-01 2.88922250e-01 -8.39929163e-01 7.76312768e-01 5.67265368e+00 9.66357350e-01 -1.12942922e+00 7.37753650e-03 2.84818202e-01 7.19150994e-03 -5.33794343e-01 -1.12922132e-01 -6.56864941e-01 5.76014340e-01 1.07370842e+00 -6.02334678e-01 6.36090755e-01 1.15156794e+00 2.16135290e-02 -3.88673097e-02 -1.49352348e+00 9.02605057e-01 2.75856763e-01 -1.58768809e+00 -5.94960824e-02 -8.88322294e-02 9.77261007e-01 4.64309752e-01 -2.03208581e-01 7.78089404e-01 4.30455625e-01 -8.67831588e-01 6.78265929e-01 2.48112291e-01 8.86873305e-01 -4.49372441e-01 7.25066423e-01 3.37945521e-01 -9.87171888e-01 -1.29760981e-01 -3.26030791e-01 -7.97931626e-02 -4.58702832e-01 5.05761385e-01 -6.80443227e-01 4.16828275e-01 7.50311732e-01 9.97653425e-01 -1.15497315e+00 9.38389301e-01 1.20679714e-01 5.66649556e-01 8.54207948e-02 -2.49708384e-01 8.69711116e-02 1.96456179e-01 3.10786456e-01 1.75272191e+00 3.53421122e-01 -4.31284994e-01 5.83020672e-02 1.35204792e+00 -4.99235719e-01 1.38435185e-01 -7.65013695e-01 -4.78614360e-01 5.80092669e-01 1.03051996e+00 -5.98720312e-01 -3.90424252e-01 -6.89079821e-01 7.10461795e-01 4.83202398e-01 2.51598686e-01 -8.37377787e-01 -8.76268387e-01 6.25173688e-01 -2.49343023e-01 8.48452821e-02 -1.34574309e-01 -4.01896387e-01 -1.27723753e+00 5.10517895e-01 -1.57727516e+00 -1.97753608e-02 -5.03867805e-01 -9.33973312e-01 6.12985969e-01 8.46942216e-02 -1.09515309e+00 -4.16518360e-01 -4.03388202e-01 -6.70346022e-01 6.26076460e-01 -1.16264451e+00 -7.51978636e-01 -4.38140690e-01 -8.03957805e-02 9.73506153e-01 -5.84180713e-01 8.11668634e-01 5.31276405e-01 -6.62864804e-01 7.20508695e-01 4.18951333e-01 3.22862357e-01 7.15427697e-01 -1.32470381e+00 1.07142460e+00 9.20865655e-01 -3.08476258e-02 1.23672485e+00 6.21563315e-01 -5.33525407e-01 -1.60352099e+00 -1.07945776e+00 9.50553358e-01 -5.38231492e-01 8.60996783e-01 -5.77789903e-01 -1.01795423e+00 4.34302539e-01 6.21227145e-01 -1.83564723e-01 6.62871242e-01 5.30288815e-02 -4.56708312e-01 -9.55344141e-02 -6.86610222e-01 5.95636606e-01 9.93781209e-01 -7.56602645e-01 -4.77577806e-01 2.49509260e-01 1.20737195e+00 -4.64133710e-01 -8.69057238e-01 1.50074899e-01 5.51253855e-01 -1.03396201e+00 6.85688734e-01 -4.67393905e-01 1.20668542e+00 1.13708459e-01 -1.48395732e-01 -1.30738223e+00 1.69867580e-03 -5.48764169e-01 1.08883232e-01 1.60740662e+00 6.65815711e-01 -3.69357646e-01 5.41408718e-01 4.49636281e-01 -4.30890322e-01 -8.29297304e-01 -4.12200630e-01 -6.34427071e-01 3.86898279e-01 -6.42679513e-01 5.84255695e-01 1.07912123e+00 2.22756773e-01 2.16056600e-01 1.75917316e-02 -3.82307798e-01 1.96868494e-01 3.62703711e-01 9.80713665e-01 -9.03171182e-01 -4.89861101e-01 -9.93857205e-01 7.51765892e-02 -8.08972478e-01 4.69624728e-01 -1.24075949e+00 1.68938175e-01 -1.67735791e+00 5.81351161e-01 -1.61480188e-01 2.46744156e-01 8.12081993e-01 -8.72929618e-02 -1.23380989e-01 1.55822426e-01 2.54071385e-01 -8.86725128e-01 2.54211873e-01 7.56282747e-01 -3.57694864e-01 -1.52857706e-01 -1.04855128e-01 -9.55482960e-01 5.10412574e-01 1.03084457e+00 -6.91802204e-01 -3.39091033e-01 -1.04493535e+00 6.54620290e-01 5.14471717e-02 1.79777905e-01 -8.88990045e-01 1.29927397e-01 5.38071021e-02 -2.95940191e-01 -3.65450233e-01 -4.81324404e-01 -4.32195276e-01 -9.12720710e-02 5.66333175e-01 -7.38240361e-01 5.15685141e-01 5.83584487e-01 4.36836388e-03 -1.79989442e-01 -6.61445022e-01 4.93077725e-01 -3.54291201e-01 -7.24681079e-01 -2.01803923e-01 -4.59275961e-01 6.26383722e-01 4.77317959e-01 3.48346345e-02 -8.28569353e-01 -7.98322260e-04 9.65387225e-02 3.36605728e-01 9.27272260e-01 7.90796578e-01 5.02363741e-01 -1.07241333e+00 -8.15082908e-01 2.73174107e-01 4.77408230e-01 1.10116892e-01 4.33712304e-02 6.98440731e-01 -9.64071929e-01 4.73885953e-01 -1.52847394e-01 -4.76800025e-01 -1.31989849e+00 4.50916946e-01 1.73026368e-01 -1.35304689e-01 -4.18372691e-01 7.53058076e-01 -8.43218192e-02 -7.67241955e-01 3.57506365e-01 -7.08098292e-01 5.06867804e-02 -2.27409214e-01 4.86504108e-01 2.77099669e-01 3.02662045e-01 -2.70391911e-01 -4.32953149e-01 3.52078110e-01 -2.84058124e-01 4.62907702e-01 1.58737791e+00 1.85177013e-01 -7.88811207e-01 2.93785512e-01 1.67365420e+00 2.80969858e-01 -5.93364298e-01 -1.32087976e-01 3.53942931e-01 -2.64643759e-01 -2.84013420e-01 -8.11161458e-01 -1.09925771e+00 7.33045340e-01 2.59668291e-01 3.34777594e-01 1.00694883e+00 7.51516521e-02 5.72571993e-01 9.18205976e-01 2.66040474e-01 -8.80211353e-01 1.67756334e-01 7.69923985e-01 1.06904531e+00 -1.34943354e+00 3.69447432e-02 2.75331512e-02 -3.45356673e-01 1.27428889e+00 7.51851857e-01 7.49921054e-02 2.04379782e-01 5.69817066e-01 1.84136555e-02 -3.14396054e-01 -1.14256632e+00 2.28998065e-01 2.07397953e-01 4.19064403e-01 9.55176592e-01 -2.40903720e-01 -2.27265954e-01 6.50334895e-01 -3.31093222e-01 -4.62503769e-02 9.24835980e-01 1.09424770e+00 -2.45785013e-01 -1.29193556e+00 5.88488439e-03 7.59753644e-01 -5.95496356e-01 -5.26642442e-01 -4.33970600e-01 8.63779783e-01 -7.11772218e-02 7.78489649e-01 -1.60963416e-01 -3.58763993e-01 3.58009279e-01 -1.60672963e-01 1.43765330e-01 -9.41168785e-01 -8.92434597e-01 -2.60056436e-01 1.93208396e-01 -7.22059608e-01 -1.91500664e-01 -5.51024675e-01 -1.07157719e+00 -4.45988208e-01 -2.11672232e-01 2.23792568e-01 6.78144693e-01 4.73346025e-01 9.12025750e-01 6.50167644e-01 2.21929207e-01 -6.69692397e-01 -5.68669498e-01 -1.01146710e+00 1.21950313e-01 2.92479992e-01 3.39672476e-01 -1.47934750e-01 -3.63670021e-01 4.12286252e-01]
[7.58885383605957, 7.944717884063721]
abcbb7c5-1a18-469f-a914-5c1a72d4ac27
sampling-from-large-graphs
null
null
https://dl.acm.org/doi/10.1145/1150402.1150479
https://cs.stanford.edu/people/jure/pubs/sampling-kdd06.pdf
Sampling From Large Graphs
Given a huge real graph, how can we derive a representative sample? There are many known algorithms to compute interesting measures (shortest paths, centrality, betweenness, etc.), but several of them become impractical for large graphs. Thus graph sampling is essential.The natural questions to ask are (a) which sampling method to use, (b) how small can the sample size be, and (c) how to scale up the measurements of the sample (e.g., the diameter), to get estimates for the large graph. The deeper, underlying question is subtle: how do we measure success?.We answer the above questions, and test our answers by thorough experiments on several, diverse datasets, spanning thousands nodes and edges. We consider several sampling methods, propose novel methods to check the goodness of sampling, and develop a set of scaling laws that describe relations between the properties of the original and the sample.In addition to the theoretical contributions, the practical conclusions from our work are: Sampling strategies based on edge selection do not perform well; simple uniform random node selection performs surprisingly well. Overall, best performing methods are the ones based on random-walks and "forest fire"; they match very accurately both static as well as evolutionary graph patterns, with sample sizes down to about 15% of the original graph.
['Christos Faloutsos', 'Jure Leskovec']
2006-08-21
null
null
null
kdd-2006-8
['graph-sampling']
['graphs']
[ 1.41018778e-01 1.29873767e-01 -2.77187914e-01 6.41128197e-02 -4.53250259e-01 -7.90784180e-01 5.00630438e-01 3.56525660e-01 -2.66565561e-01 9.80978310e-01 -8.99240226e-02 -4.05973345e-01 -4.99912918e-01 -1.26028895e+00 -5.00785708e-01 -7.41338551e-01 -4.65674818e-01 8.68906021e-01 7.87267089e-01 -3.63707930e-01 4.87775832e-01 4.18824375e-01 -1.36219990e+00 -4.72358078e-01 1.03957522e+00 5.33016682e-01 -1.28916919e-01 8.01913917e-01 -1.14034303e-01 3.95469308e-01 -5.88309228e-01 -5.04237354e-01 2.07610443e-01 -6.91421688e-01 -9.89649653e-01 3.23818952e-01 -1.36075923e-02 7.52156898e-02 -1.12750649e-01 1.18496490e+00 4.14199829e-01 8.61623138e-03 5.60754538e-01 -1.25740719e+00 -2.38442674e-01 9.46576238e-01 -7.10322082e-01 2.81032503e-01 5.87272644e-01 1.07903779e-01 1.20083785e+00 -3.48827720e-01 8.61227512e-01 1.01319432e+00 8.19356441e-01 1.08734943e-01 -1.50232863e+00 -3.70644748e-01 -6.70931563e-02 -2.68392451e-02 -1.49800408e+00 -2.87575990e-01 5.43001115e-01 -2.40487829e-01 4.96125579e-01 5.73693931e-01 8.15822959e-01 8.11566949e-01 3.52766700e-02 2.00536728e-01 1.12683821e+00 -5.18746853e-01 4.02854562e-01 1.97294801e-01 2.51598150e-01 8.96002352e-01 9.64363694e-01 -7.31273741e-02 -1.49532765e-01 -7.26197302e-01 5.83644211e-01 -2.18919426e-01 -4.57665831e-01 -5.25609910e-01 -1.12477708e+00 9.88215208e-01 3.26276124e-01 3.71580720e-01 -2.02004880e-01 1.82024524e-01 3.33452165e-01 5.97924352e-01 4.53012705e-01 8.68460417e-01 -4.55036700e-01 5.17128259e-02 -7.39680350e-01 2.59802371e-01 1.31582034e+00 1.05634463e+00 1.06194317e+00 -2.91687995e-01 2.58311480e-01 5.99157393e-01 -1.37591958e-01 3.75368983e-01 -1.70829147e-02 -7.44547069e-01 1.49850160e-01 5.77029467e-01 5.77506702e-03 -1.29141366e+00 -3.83938640e-01 -4.14926738e-01 -9.05506015e-01 -4.38506827e-02 8.99009585e-01 -2.15094849e-01 -5.22354662e-01 1.65474927e+00 4.39678907e-01 -1.52246594e-01 -6.88497543e-01 3.20486426e-01 3.63395780e-01 3.01211178e-01 -3.49948972e-01 -4.46119487e-01 1.04199314e+00 -6.48012042e-01 -1.22634128e-01 -1.54522926e-01 8.21138382e-01 -6.18649602e-01 1.02161634e+00 2.98623174e-01 -9.40817416e-01 -5.59827350e-02 -8.39677751e-01 6.34387732e-01 -2.99648404e-01 -2.81819671e-01 7.17573047e-01 9.33182716e-01 -1.31722152e+00 1.06617463e+00 -4.89873141e-01 -7.78608620e-01 -7.75367767e-02 2.58672416e-01 -1.89203098e-01 -1.13105088e-01 -7.52296269e-01 5.96763849e-01 3.12554866e-01 -3.17208201e-01 -3.56425494e-01 -2.58097202e-01 -3.89341295e-01 9.43147093e-02 9.27419126e-01 -8.40584934e-01 6.80257380e-01 -8.81953776e-01 -9.80989516e-01 6.39709830e-01 -7.42365941e-02 -2.80276746e-01 4.18718189e-01 5.72747290e-01 -1.20700493e-01 3.12033236e-01 1.03679858e-01 1.21629067e-01 4.73799855e-01 -1.18071067e+00 -2.88549721e-01 -3.85037571e-01 1.46507174e-01 -1.98805109e-01 -3.22996825e-01 -1.93586200e-02 -2.82905072e-01 -3.14853281e-01 2.52180755e-01 -1.17082441e+00 -6.50546849e-01 -1.94225490e-01 -7.56830931e-01 -2.71557540e-01 1.77799493e-01 -2.87187785e-01 1.47280061e+00 -1.89799011e+00 1.99813880e-02 8.13218415e-01 6.28363669e-01 -1.88440844e-01 -1.62547424e-01 9.28025901e-01 1.44302368e-01 6.61359787e-01 -1.13366760e-01 4.18223619e-01 -2.95629650e-02 2.95399595e-02 1.47057220e-01 5.86745441e-01 -9.17967036e-02 6.34264052e-01 -1.12248969e+00 -5.14336646e-01 -1.49395093e-01 2.50755996e-02 -5.78104138e-01 -1.03627332e-01 -7.60853961e-02 -4.19648141e-02 -5.37735999e-01 5.37850261e-01 4.79649246e-01 -6.73359156e-01 6.09169066e-01 -1.01402150e-02 2.24509582e-01 3.55521798e-01 -1.43526483e+00 7.92529643e-01 -5.40732406e-02 5.42986333e-01 1.03722803e-01 -9.31199372e-01 1.06799686e+00 -1.59575671e-01 3.56552452e-01 -2.20260203e-01 1.12870179e-01 4.23504382e-01 3.09632868e-01 -2.73229659e-01 2.89914936e-01 -2.75687501e-02 -3.03600263e-02 7.19484389e-01 -1.51375845e-01 -1.65479973e-01 7.75214434e-01 5.06172121e-01 1.86626339e+00 -5.68575263e-01 7.17137754e-01 -7.01494515e-01 2.10993230e-01 2.05788687e-02 2.63157874e-01 8.58315766e-01 -7.62508586e-02 5.14723301e-01 1.22966623e+00 -4.97765690e-01 -1.06783593e+00 -9.71296310e-01 2.55081087e-01 8.02334487e-01 2.63769001e-01 -7.40925431e-01 -8.14583182e-01 -7.41548896e-01 2.11292952e-02 3.10504794e-01 -7.08768070e-01 -6.25651777e-02 -4.15009409e-01 -8.96158874e-01 1.90664604e-01 1.31340221e-01 1.04585595e-01 -8.45444024e-01 -1.85446203e-01 7.74142966e-02 -5.21033257e-02 -9.86956775e-01 -5.25150478e-01 2.81359372e-03 -1.03872979e+00 -1.50265336e+00 -4.13224727e-01 -5.64679921e-01 8.62378538e-01 5.11158347e-01 1.53140438e+00 7.24193454e-01 -1.44183561e-01 1.99130431e-01 -4.72181737e-01 1.54531561e-02 -5.94537616e-01 3.76688570e-01 -1.05937243e-01 -3.33005130e-01 5.51481918e-02 -9.48577940e-01 -4.85035360e-01 5.26598454e-01 -5.61522603e-01 -4.69464302e-01 8.02802384e-01 7.79779017e-01 3.40851039e-01 3.52614582e-01 5.05900085e-01 -1.36216247e+00 9.31921542e-01 -5.86297452e-01 -4.19543505e-01 3.95575523e-01 -9.90490079e-01 2.69247264e-01 7.91772306e-01 -4.07964766e-01 -1.06419176e-01 -3.24475884e-01 -6.99868500e-02 1.27263665e-01 2.35021472e-01 4.18454915e-01 6.21400736e-02 -3.72490287e-01 8.36796224e-01 1.47596866e-01 8.14861879e-02 -1.45825639e-01 3.16929787e-01 1.72265559e-01 -4.56184298e-02 -5.95967174e-01 9.02837694e-01 4.11290079e-01 1.39558598e-01 -1.13315201e+00 -4.72820908e-01 -4.29429561e-01 -4.95646387e-01 -2.10160375e-01 2.04532549e-01 -1.95650116e-01 -8.13460886e-01 -7.79337361e-02 -8.16479146e-01 -3.37451339e-01 -4.15784359e-01 2.02307105e-01 -2.99630612e-01 5.88593781e-01 -4.55190480e-01 -8.77247870e-01 -2.56405264e-01 -8.96052241e-01 8.27319443e-01 2.75618415e-02 -5.06085932e-01 -1.07197094e+00 2.15234682e-01 -1.91617757e-01 4.79967505e-01 6.42301381e-01 1.16883087e+00 -6.58853173e-01 -7.14861095e-01 -3.02190989e-01 -3.23643208e-01 -9.28221270e-02 8.44785124e-02 5.02043664e-01 -1.76462919e-01 -6.15300238e-01 -4.13209528e-01 -4.10845540e-02 7.25779057e-01 4.97054994e-01 8.85424435e-01 -3.64995390e-01 -6.74261034e-01 3.54411721e-01 1.59637213e+00 -2.71479040e-01 6.59695387e-01 1.27095163e-01 3.84241402e-01 6.02814496e-01 2.59773880e-01 3.80502999e-01 2.09425569e-01 4.19622004e-01 2.61391908e-01 2.43805453e-01 -8.06234851e-02 -3.60453695e-01 1.30277589e-01 1.26223993e+00 -2.79628456e-01 -3.00051719e-01 -7.98764229e-01 4.81028229e-01 -1.49273336e+00 -9.58379745e-01 -3.73116404e-01 2.49698949e+00 5.38168788e-01 5.81069410e-01 7.77148664e-01 3.34709376e-01 8.08848321e-01 2.66084135e-01 -4.19527203e-01 -2.04144627e-01 3.45925763e-02 2.51732916e-01 7.20941544e-01 4.84263778e-01 -5.62111199e-01 5.18094540e-01 7.31347227e+00 9.55012619e-01 -8.25833082e-01 -1.30619571e-01 6.44932806e-01 2.50810653e-01 -6.63764417e-01 5.41294634e-01 -5.85701466e-01 4.63592291e-01 1.00141549e+00 -5.72631359e-01 5.49856365e-01 7.87742257e-01 -3.35569642e-02 -3.57908458e-01 -8.99280190e-01 6.61808133e-01 -1.88353837e-01 -1.32309794e+00 -1.71256065e-01 4.37529296e-01 6.35452986e-01 4.43134457e-02 -5.94277680e-01 2.23608557e-02 7.12828100e-01 -9.23345447e-01 1.84468418e-01 3.82841021e-01 7.15434670e-01 -7.03803301e-01 5.87688804e-01 5.09244204e-01 -1.43204463e+00 1.64195687e-01 -5.70041358e-01 2.41810698e-02 -1.38903394e-01 1.21162343e+00 -7.70956159e-01 5.40673673e-01 5.63985169e-01 3.77714276e-01 -9.09356058e-01 1.23204780e+00 -1.29158035e-01 6.87826037e-01 -6.73197210e-01 -8.36789250e-01 -9.20297429e-02 -4.23440546e-01 7.34401047e-01 8.48415315e-01 4.13075984e-01 -2.55182683e-01 1.15404136e-01 7.03811705e-01 -1.42514318e-01 2.56679088e-01 -8.30749154e-01 -2.43983403e-01 8.39336276e-01 1.41061532e+00 -1.38129175e+00 -1.35291547e-01 -1.36720777e-01 5.31684101e-01 5.86792409e-01 2.25848958e-01 -5.34171820e-01 -7.43818581e-01 2.89236397e-01 6.85711443e-01 2.48226270e-01 -3.66168410e-01 5.01554906e-02 -1.00143898e+00 1.95401892e-01 -9.74725246e-01 4.11365390e-01 -2.87435740e-01 -1.34404564e+00 7.14111388e-01 -6.60532117e-02 -1.00732255e+00 -2.69005030e-01 -4.55967128e-01 -8.76985013e-01 4.08676624e-01 -8.79580379e-01 -3.33426028e-01 -3.20182949e-01 1.75095811e-01 -4.33323048e-02 1.65354013e-01 6.21197999e-01 2.38494068e-01 -3.38174313e-01 5.24187267e-01 2.61468083e-01 -2.07679663e-02 2.75069952e-01 -1.33937633e+00 5.50368547e-01 7.21370101e-01 1.91251516e-01 5.72936416e-01 9.24690545e-01 -5.27341366e-01 -1.57184136e+00 -6.58726752e-01 7.54312813e-01 -3.05848211e-01 8.64626884e-01 -4.35115844e-01 -7.56860852e-01 3.56256276e-01 -1.32859573e-01 1.06837235e-01 2.40001008e-01 4.45803106e-01 -2.18646124e-01 -5.39620705e-02 -1.24885249e+00 6.47564471e-01 1.43252707e+00 -5.81182651e-02 1.68648809e-01 5.78485310e-01 7.27220058e-01 2.19510630e-01 -8.75876427e-01 2.57683575e-01 5.80704212e-01 -1.41372764e+00 8.45264673e-01 -4.89443898e-01 1.24104664e-01 -1.17607646e-01 1.43122077e-01 -1.50200117e+00 -4.70733792e-01 -9.17953432e-01 1.24549091e-01 1.23305559e+00 6.45057917e-01 -1.04501748e+00 1.07202196e+00 1.50206670e-01 6.35628164e-01 -1.22758675e+00 -7.06429124e-01 -1.09626901e+00 -8.94397646e-02 6.10331260e-02 6.36827648e-01 1.05909204e+00 4.17660959e-02 7.15328991e-01 -9.85098258e-02 -2.08677977e-01 8.16699207e-01 2.64948130e-01 1.11792850e+00 -1.65185750e+00 -5.04199743e-01 -6.22827053e-01 -4.85583574e-01 -7.95468450e-01 -2.84811318e-01 -7.26641476e-01 -3.40348661e-01 -1.45940685e+00 3.85425597e-01 -6.41417265e-01 4.15968239e-01 -1.72468752e-01 -2.93550164e-01 -5.83838411e-02 -1.61885440e-01 1.56531751e-01 -6.61679029e-01 1.76485956e-01 1.20528483e+00 2.10892752e-01 -1.35250196e-01 2.03366250e-01 -8.55429113e-01 6.71158254e-01 8.38194668e-01 -5.66039145e-01 -3.19732666e-01 1.44114658e-01 6.95164263e-01 1.49299636e-01 -8.89085699e-03 -7.89569497e-01 1.64607137e-01 -3.41648042e-01 -3.32383625e-03 -2.87693828e-01 -1.59680530e-01 -3.71310413e-01 3.07251543e-01 6.07751071e-01 -1.97331950e-01 3.64272028e-01 -4.47060674e-01 7.08585560e-01 1.37133226e-01 -4.66778427e-01 7.46695101e-01 -3.43845993e-01 -1.41210437e-01 3.90153021e-01 -6.98931143e-02 5.17274797e-01 8.67707610e-01 -2.68589199e-01 -4.57746744e-01 -6.88759744e-01 -4.65688616e-01 1.39193192e-01 7.92328656e-01 -1.69999287e-01 3.03494841e-01 -1.10351193e+00 -7.84341872e-01 -1.51389882e-01 1.17853425e-01 -2.27067590e-01 -4.24997151e-01 1.09479773e+00 -6.85206413e-01 -1.48517132e-01 2.18525395e-01 -3.85210752e-01 -1.17315924e+00 6.09649181e-01 1.89641580e-01 -4.90485728e-01 -4.57178354e-01 6.42215014e-01 -2.35028476e-01 -5.16108394e-01 -4.60337587e-02 -1.11433290e-01 2.99526840e-01 2.66884621e-02 2.90668398e-01 6.60603166e-01 -1.44325167e-01 -1.65758595e-01 -3.85930151e-01 4.92610276e-01 1.30647436e-01 1.89124241e-01 1.41642690e+00 -2.48372167e-01 -4.25482988e-01 2.56071508e-01 1.20284843e+00 4.25449789e-01 -4.35787469e-01 -1.77590281e-01 2.84601301e-01 -6.37182236e-01 -5.32142460e-01 -6.20345362e-02 -1.06747580e+00 4.85944211e-01 -1.70027375e-01 1.43207479e+00 9.87586081e-01 2.49340266e-01 5.50974667e-01 4.03330892e-01 7.76657939e-01 -8.91209066e-01 -2.77530402e-02 2.16752648e-01 6.19485676e-01 -1.03395450e+00 5.60732484e-01 -8.66330028e-01 -1.56027436e-01 1.17618024e+00 2.76832283e-01 -5.65053999e-01 7.58277833e-01 1.70349807e-01 -5.89443266e-01 -4.58655655e-01 -9.37543333e-01 -4.03866529e-01 -3.49011868e-02 3.47503960e-01 3.56824905e-01 2.05647275e-01 -6.45168245e-01 2.35896707e-02 -5.26567221e-01 -2.93096006e-01 8.11822593e-01 4.86593813e-01 -7.63357580e-01 -1.13839114e+00 -1.52789205e-01 1.03771520e+00 -9.83992144e-02 1.34086117e-01 -9.77529943e-01 1.02989101e+00 -3.46485674e-01 9.24124241e-01 -3.72025728e-01 -6.61391377e-01 2.00278357e-01 -3.58743906e-01 6.85269356e-01 -4.53139067e-01 -3.50380838e-01 -1.82920516e-01 6.29049122e-01 -4.00608033e-01 -1.96910068e-01 -6.35316908e-01 -6.35877609e-01 -1.07351899e+00 -9.16880310e-01 4.07977283e-01 2.49326780e-01 7.42512465e-01 1.73625410e-01 1.69463113e-01 8.84371281e-01 -5.99377096e-01 -7.66315401e-01 -8.70720208e-01 -9.14880991e-01 2.74304241e-01 -3.93622741e-02 -4.17387247e-01 -8.93005371e-01 -6.63177609e-01]
[6.946131229400635, 5.416998386383057]
92d90d9b-b94f-411c-9594-b4ed6ee887df
end-to-end-2d-3d-registration-between-image
2306.11346
null
https://arxiv.org/abs/2306.11346v1
https://arxiv.org/pdf/2306.11346v1.pdf
End-to-end 2D-3D Registration between Image and LiDAR Point Cloud for Vehicle Localization
Robot localization using a previously built map is essential for a variety of tasks including highly accurate navigation and mobile manipulation. A popular approach to robot localization is based on image-to-point cloud registration, which combines illumination-invariant LiDAR-based mapping with economical image-based localization. However, the recent works for image-to-point cloud registration either divide the registration into separate modules or project the point cloud to the depth image to register the RGB and depth images. In this paper, we present I2PNet, a novel end-to-end 2D-3D registration network. I2PNet directly registers the raw 3D point cloud with the 2D RGB image using differential modules with a unique target. The 2D-3D cost volume module for differential 2D-3D association is proposed to bridge feature extraction and pose regression. 2D-3D cost volume module implicitly constructs the soft point-to-pixel correspondence on the intrinsic-independent normalized plane of the pinhole camera model. Moreover, we introduce an outlier mask prediction module to filter the outliers in the 2D-3D association before pose regression. Furthermore, we propose the coarse-to-fine 2D-3D registration architecture to increase localization accuracy. We conduct extensive localization experiments on the KITTI Odometry and nuScenes datasets. The results demonstrate that I2PNet outperforms the state-of-the-art by a large margin. In addition, I2PNet has a higher efficiency than the previous works and can perform the localization in real-time. Moreover, we extend the application of I2PNet to the camera-LiDAR online calibration and demonstrate that I2PNet outperforms recent approaches on the online calibration task.
['Hesheng Wang', 'Wolfram Burgard', 'Yixiang Zhu', 'Zhe Liu', 'Yanfeng Guo', 'Yu Zheng', 'Guangming Wang']
2023-06-20
null
null
null
null
['point-cloud-registration', 'image-based-localization', 'image-to-point-cloud-registration']
['computer-vision', 'computer-vision', 'computer-vision']
[-2.23655745e-01 -2.84578383e-01 -4.06349227e-02 -4.24625099e-01 -9.49615479e-01 -3.82569015e-01 4.03374851e-01 -4.94287871e-02 -7.87438810e-01 2.28670835e-01 -6.00430369e-01 6.10634275e-02 -7.54233524e-02 -8.30586612e-01 -1.01956880e+00 -4.16230381e-01 8.88789892e-02 1.11461961e+00 5.32866120e-01 -3.57204556e-01 5.37399054e-01 8.70418727e-01 -1.49201012e+00 -6.75649703e-01 8.59335065e-01 1.17544103e+00 2.52806127e-01 2.36293942e-01 -1.03258714e-01 -2.11895257e-01 -1.53726563e-01 -6.56675473e-02 7.69393146e-01 5.29491380e-02 -4.59798314e-02 -2.10988790e-01 6.67745292e-01 -1.58649042e-01 -1.59968868e-01 1.33153343e+00 6.41835213e-01 2.68439427e-02 4.36917454e-01 -1.32877707e+00 -1.88847676e-01 -2.90372431e-01 -9.37513411e-01 -4.76441562e-01 5.98956168e-01 7.74558038e-02 2.58595824e-01 -1.41487134e+00 6.93153083e-01 1.20637107e+00 1.18558300e+00 -4.09565344e-02 -1.17208445e+00 -8.19191217e-01 -2.27470726e-01 -3.53915617e-03 -1.92296040e+00 -1.08512290e-01 8.28490257e-01 -3.96700442e-01 9.11694884e-01 -2.75645494e-01 4.89612788e-01 5.46012282e-01 4.24523503e-01 -8.08753893e-02 1.08592546e+00 -3.57440263e-01 9.48409364e-02 -2.21781135e-01 -6.66159689e-02 9.22537327e-01 4.41991270e-01 3.15410972e-01 -6.88929379e-01 -5.02213947e-02 1.23672044e+00 3.13072413e-01 -1.02152951e-01 -1.05469680e+00 -1.35064626e+00 7.04029441e-01 7.60223389e-01 -1.67162448e-01 -1.83011845e-01 3.77017081e-01 9.46897864e-02 1.82114407e-01 2.96018541e-01 2.96343595e-01 -4.02395636e-01 -9.78224948e-02 -6.42977118e-01 -9.96578392e-03 5.67295015e-01 1.60840523e+00 1.46813262e+00 -2.43554473e-01 5.98153710e-01 4.90355283e-01 6.23824537e-01 1.16619229e+00 4.67150062e-01 -1.08215022e+00 6.37441516e-01 7.87522733e-01 1.22375481e-01 -1.22255540e+00 -6.90810621e-01 -4.44314033e-02 -7.21934974e-01 4.43575531e-01 1.80209383e-01 3.24250400e-01 -8.31968486e-01 1.13349223e+00 3.78273308e-01 1.41061604e-01 3.63638774e-02 9.06925738e-01 6.01016402e-01 1.68693110e-01 -6.01679385e-01 1.60870209e-01 1.09355116e+00 -7.73929596e-01 -5.01777411e-01 -5.08852601e-01 5.87706447e-01 -8.87573242e-01 9.13088918e-01 7.50032812e-02 -7.52608478e-01 -6.92459226e-01 -1.29149866e+00 -3.38205576e-01 -2.83145875e-01 2.57166862e-01 2.96567559e-01 3.37389737e-01 -8.44270825e-01 3.37134123e-01 -9.76364315e-01 -6.02790892e-01 6.15093894e-02 7.89143264e-01 -9.34752226e-01 -2.34724730e-01 -7.15044141e-01 1.18325245e+00 5.55565953e-01 2.05420941e-01 -3.90637130e-01 -4.77294654e-01 -1.14162230e+00 -5.32270670e-01 3.59424949e-01 -7.41170049e-01 7.76563883e-01 -3.44536752e-01 -1.61751223e+00 1.05718732e+00 -3.63934934e-01 -2.13257104e-01 5.68043530e-01 -2.69453585e-01 -1.98864453e-02 2.51544192e-02 5.03369093e-01 7.46009231e-01 4.96202052e-01 -1.46760213e+00 -6.54097915e-01 -8.38678002e-01 -4.56503242e-01 3.57988536e-01 3.22072923e-01 -6.04729116e-01 -9.38762903e-01 -1.43514097e-01 1.34018505e+00 -1.15310168e+00 -3.90078068e-01 1.29191726e-01 -1.64093703e-01 1.38479725e-01 9.05424654e-01 -2.80058116e-01 3.84909362e-01 -2.22333455e+00 -7.35531524e-02 5.63073277e-01 -4.01974544e-02 -3.77746403e-01 -1.97872687e-02 1.34421930e-01 1.67037964e-01 -4.12911117e-01 -2.08759934e-01 -7.68173039e-01 1.83310322e-02 4.14904833e-01 -1.03176758e-02 1.10025620e+00 -1.23286098e-01 8.40696692e-01 -8.62520993e-01 -4.58913147e-01 5.26010871e-01 4.63039935e-01 -3.27278167e-01 -1.74214076e-02 2.58878440e-01 6.43931091e-01 -2.36603409e-01 9.24531221e-01 1.18067098e+00 2.51403689e-01 -4.06519592e-01 -4.29219544e-01 -4.68145221e-01 9.24600437e-02 -1.30018282e+00 2.48645592e+00 -3.53033453e-01 4.40264016e-01 9.38102752e-02 -5.81620336e-01 1.39733815e+00 -1.87166616e-01 5.86248577e-01 -8.40364993e-01 2.82164067e-01 7.47100055e-01 -4.49649930e-01 -1.39720395e-01 8.31384778e-01 6.91046268e-02 -3.85387391e-01 -2.03206554e-01 1.39288353e-02 -9.00601983e-01 -3.53878140e-01 -2.36102149e-01 7.44037330e-01 6.84919417e-01 3.75932604e-01 -1.10814773e-01 6.11858904e-01 4.31166738e-01 6.52968049e-01 5.11154115e-01 -7.42028952e-02 9.11525190e-01 -5.56464829e-02 -1.75188959e-01 -1.11169386e+00 -1.11618543e+00 -2.60482818e-01 1.03550218e-01 9.89885628e-01 -7.86244646e-02 -4.43766266e-01 -3.41601849e-01 4.42150235e-01 1.75275624e-01 -1.44969642e-01 -1.43229254e-02 -7.38870680e-01 -2.38499478e-01 4.27450478e-01 3.87986720e-01 9.26447392e-01 -3.99462283e-01 -5.71794808e-01 5.95747270e-02 -1.13110133e-01 -1.23595726e+00 -3.69653672e-01 3.98867458e-01 -1.10898423e+00 -1.03732038e+00 -1.86547682e-01 -9.06571746e-01 8.44930768e-01 5.81537843e-01 6.32353663e-01 -2.26055607e-01 1.67248379e-02 4.89559233e-01 -2.32264534e-01 -3.08471411e-01 5.69777526e-02 1.01832762e-01 6.34396672e-01 -4.63419437e-01 5.55514574e-01 -6.39461219e-01 -4.46976721e-01 8.35693896e-01 -3.90854716e-01 -3.22563261e-01 5.72215319e-01 5.93729496e-01 1.22529650e+00 -2.35297889e-01 -1.50628746e-01 -3.23854119e-01 -4.17181738e-02 -1.08903445e-01 -1.24538207e+00 -3.46312344e-01 -7.46511400e-01 4.50050719e-02 1.01885714e-01 -2.42937550e-01 -5.32868385e-01 8.24217856e-01 -1.68211624e-01 -7.63873041e-01 -1.05806552e-02 4.99508768e-01 -3.21681082e-01 -7.90604591e-01 7.00958848e-01 1.23457924e-01 3.29489350e-01 -3.78430516e-01 3.69151860e-01 5.41670859e-01 1.03999460e+00 -4.06139702e-01 1.34651494e+00 8.52938116e-01 4.55916911e-01 -6.18027151e-01 -3.17132950e-01 -9.36799288e-01 -1.39134622e+00 -4.50903438e-02 8.08141053e-01 -1.27835071e+00 -7.43500948e-01 4.24621195e-01 -1.38260162e+00 -3.33808772e-02 -2.59697706e-01 8.24419737e-01 -8.43897343e-01 5.84785223e-01 -8.00750703e-02 -5.13602316e-01 -9.67696086e-02 -1.50818324e+00 1.35864782e+00 1.94135606e-01 1.16644278e-01 -6.52168870e-01 3.55273277e-01 9.33023393e-02 -7.58450106e-02 3.31273556e-01 7.32223094e-02 -2.41772503e-01 -1.03306699e+00 -6.15078926e-01 -2.63266653e-01 1.36199025e-02 2.73130275e-03 -2.27708027e-01 -8.32414031e-01 -3.09820294e-01 2.15633258e-01 -2.35320684e-02 5.15819311e-01 1.51741594e-01 4.46400374e-01 4.19561684e-01 -4.94866997e-01 1.19030416e+00 1.72824395e+00 -7.94296269e-04 8.41787517e-01 9.91526723e-01 8.74776244e-01 4.24843311e-01 1.29421127e+00 5.83948642e-02 6.89537764e-01 1.08747506e+00 9.63502645e-01 -3.76474336e-02 2.20913768e-01 -4.12477493e-01 4.88227874e-01 9.00014400e-01 -8.39608312e-02 4.74223375e-01 -9.94029045e-01 3.25213104e-01 -2.00122571e+00 -3.86752188e-01 -6.00889683e-01 2.54240775e+00 4.17862594e-01 1.54739112e-01 -3.87055784e-01 -3.08438033e-01 6.48629725e-01 -1.90880120e-01 -3.91595721e-01 1.68156803e-01 -1.21027648e-01 1.34430647e-01 1.33568311e+00 7.05784440e-01 -1.00193596e+00 1.17776799e+00 4.72510815e+00 7.13779032e-01 -1.06400979e+00 2.70614743e-01 -2.99181610e-01 3.80162656e-01 1.60074577e-01 3.10722947e-01 -1.23059726e+00 2.26162612e-01 6.00224257e-01 2.22789392e-01 6.08034804e-02 1.25717795e+00 -6.73029348e-02 -4.32118624e-01 -1.10998154e+00 1.47426963e+00 2.50891745e-01 -1.02648067e+00 -3.71471554e-01 3.15074801e-01 5.96568704e-01 6.55789375e-01 -1.82325006e-01 -4.31869254e-02 -1.08079769e-01 -7.17355371e-01 1.00609100e+00 4.44638789e-01 1.15773201e+00 -7.08977461e-01 9.56319332e-01 5.88773906e-01 -1.52345037e+00 1.43573046e-01 -7.42003679e-01 5.97109012e-02 3.66419256e-01 5.45087278e-01 -7.57767141e-01 1.02583718e+00 7.86451757e-01 7.73394525e-01 -5.96876979e-01 1.15463030e+00 -2.77983367e-01 -4.17058885e-01 -7.34128237e-01 4.42860097e-01 8.35116059e-02 -7.31774390e-01 4.61218745e-01 7.74056852e-01 8.71293843e-01 -7.69368857e-02 3.61402303e-01 8.22290540e-01 1.10170230e-01 -1.42104268e-01 -9.34511900e-01 8.29610050e-01 7.62794614e-01 1.20605385e+00 -7.89820969e-01 -2.17226278e-02 -1.90279275e-01 1.11397386e+00 6.86242878e-02 -1.48202270e-01 -7.86409020e-01 -6.38901711e-01 5.13326228e-01 1.36875004e-01 2.13009596e-01 -9.23271477e-01 -4.16425377e-01 -1.06575716e+00 2.55252481e-01 -2.89485872e-01 -2.46825293e-01 -8.96889925e-01 -9.99737918e-01 5.10625362e-01 -1.79042369e-02 -1.79166543e+00 -4.75131162e-02 -7.96066344e-01 -1.69980049e-01 9.77091432e-01 -1.77142453e+00 -1.05424678e+00 -9.80589628e-01 7.15753794e-01 1.46887645e-01 9.59608927e-02 7.11460948e-01 3.42537582e-01 -1.19393766e-01 3.24442208e-01 2.29778275e-01 1.80075720e-01 1.08035028e+00 -9.03380811e-01 5.41500330e-01 8.02153409e-01 2.64984951e-03 6.84827507e-01 3.12462598e-01 -8.89651835e-01 -1.68278325e+00 -1.11404729e+00 5.94496727e-01 -8.71693611e-01 3.73361439e-01 -4.24938887e-01 -5.51361322e-01 8.46816599e-01 -4.38002437e-01 3.53629410e-01 -8.27980638e-02 -4.16902632e-01 -1.57549545e-01 -4.21274751e-01 -1.40629303e+00 3.07776958e-01 1.18376243e+00 -5.31196654e-01 -4.94244397e-01 1.75113723e-01 7.25143909e-01 -1.22493505e+00 -9.34540570e-01 5.10399699e-01 5.30355394e-01 -9.35735881e-01 1.28750384e+00 5.48684716e-01 -1.99564621e-01 -9.39673662e-01 -5.09625614e-01 -1.03904223e+00 2.53181513e-02 -5.12047946e-01 3.96915078e-01 1.03022456e+00 7.46243224e-02 -1.06150639e+00 9.56419349e-01 2.09717497e-01 -5.38902104e-01 -1.51352346e-01 -1.57521939e+00 -1.04691315e+00 -3.21837485e-01 -5.21120906e-01 4.77622807e-01 7.53154814e-01 -4.05234069e-01 1.08415209e-01 1.56474896e-02 7.71086097e-01 9.62223053e-01 3.11809126e-03 1.47042179e+00 -1.38631940e+00 3.91826421e-01 -1.50591552e-01 -9.96729851e-01 -1.52833724e+00 1.35092258e-01 -8.01223159e-01 5.33404112e-01 -1.25579047e+00 -3.13831568e-01 -7.75421202e-01 2.21245110e-01 1.40013650e-01 2.68822014e-01 5.17638266e-01 1.77032426e-02 7.57926047e-01 -4.33800995e-01 5.56749761e-01 7.99637794e-01 9.21616480e-02 -3.36660415e-01 -1.40044570e-01 -1.52801175e-03 9.02882576e-01 4.74977165e-01 -6.01550400e-01 4.70102951e-02 -4.51821059e-01 1.66405290e-01 -1.39294699e-01 5.22677422e-01 -1.36914694e+00 6.60591304e-01 7.05893850e-03 3.59593630e-01 -1.25891519e+00 7.35530198e-01 -1.32405281e+00 3.24211270e-01 5.03678143e-01 5.16234756e-01 4.86460060e-01 3.03325113e-02 6.04521513e-01 -2.96215147e-01 -1.96346611e-01 7.78851390e-01 -1.64595470e-01 -8.81144822e-01 5.15204787e-01 2.61703849e-01 -3.36850226e-01 1.06934309e+00 -8.02626610e-01 -2.00844690e-01 -1.17346816e-01 -1.40176475e-01 2.76877493e-01 1.30673850e+00 1.75849259e-01 1.00065792e+00 -1.62089288e+00 -2.31059551e-01 5.68266749e-01 5.68825960e-01 9.40150857e-01 -2.20431864e-01 1.34768367e+00 -1.13226449e+00 4.07458901e-01 -2.58465946e-01 -1.44746041e+00 -1.04375291e+00 3.64067852e-01 4.00217414e-01 2.25574061e-01 -6.79287672e-01 7.30520546e-01 -1.37402520e-01 -1.11880243e+00 1.22073427e-01 -5.57097256e-01 3.02003145e-01 -2.10522413e-01 -7.36596584e-02 4.66683358e-01 1.69322476e-01 -1.16660500e+00 -6.80686355e-01 1.50665212e+00 3.88774097e-01 -1.83788314e-01 1.16671252e+00 -4.40520495e-01 -5.01659930e-01 3.95296186e-01 1.31172216e+00 1.22814089e-01 -1.38448751e+00 -1.82116672e-01 1.46511137e-01 -7.00993836e-01 -1.23812474e-01 -3.42481732e-02 -8.90667021e-01 7.72199273e-01 9.43663061e-01 -4.44885522e-01 6.83757722e-01 -1.64507896e-01 6.48697317e-01 5.59437275e-01 1.10897672e+00 -9.56017017e-01 -3.86519060e-02 9.92969871e-01 7.30122447e-01 -1.51132572e+00 4.11510617e-01 -5.01307487e-01 -2.57720649e-01 1.26435673e+00 6.62054956e-01 -4.33739841e-01 4.64795351e-01 2.56973475e-01 2.91383326e-01 -2.80018628e-01 4.70247060e-01 -1.47115365e-01 1.78104907e-01 8.89443278e-01 -1.94964424e-01 -2.61033088e-01 1.22792296e-01 1.43408537e-01 -2.94842064e-01 -1.07601315e-01 2.33570337e-01 1.07706428e+00 -5.61577916e-01 -1.24316871e+00 -9.21315193e-01 -2.59840161e-01 4.68227804e-01 1.09319679e-01 -1.77912265e-01 1.14637744e+00 3.68363768e-01 5.36574602e-01 2.41157681e-01 -5.86072028e-01 7.58132160e-01 -1.74671218e-01 5.90392768e-01 -5.75182736e-01 -3.42139572e-01 1.89820692e-01 -4.04831618e-01 -1.09265614e+00 -4.49569434e-01 -6.22678161e-01 -1.69991624e+00 -2.00237259e-01 -5.57389259e-01 -1.33024678e-01 1.27438569e+00 5.82440376e-01 5.30917466e-01 -1.15106657e-01 6.58953547e-01 -1.63931465e+00 -3.61216098e-01 -7.51248240e-01 -5.74238122e-01 -8.11317842e-03 3.13384831e-01 -9.59445775e-01 -3.90286177e-01 -3.22960824e-01]
[7.3972883224487305, -2.2886905670166016]
fcd2497e-5616-4233-8291-f2134d7d4029
open-eye-an-open-platform-to-study-human
2205.06680
null
https://arxiv.org/abs/2205.06680v1
https://arxiv.org/pdf/2205.06680v1.pdf
Open-Eye: An Open Platform to Study Human Performance on Identifying AI-Synthesized Faces
AI-synthesized faces are visually challenging to discern from real ones. They have been used as profile images for fake social media accounts, which leads to high negative social impacts. Although progress has been made in developing automatic methods to detect AI-synthesized faces, there is no open platform to study the human performance of AI-synthesized faces detection. In this work, we develop an online platform called Open-eye to study the human performance of AI-synthesized face detection. We describe the design and workflow of the Open-eye in this paper.
['Siwei Lyu', 'Ming-Ching Chang', 'Xin Wang', 'Shu Hu', 'Hui Guo']
2022-05-13
null
null
null
null
['face-detection']
['computer-vision']
[ 1.66549464e-03 3.58588755e-01 2.00491250e-01 -1.19119547e-01 -1.41190186e-01 -4.49494451e-01 6.91395402e-01 -3.40183318e-01 -3.80526460e-03 4.32185918e-01 -1.12689331e-01 2.19063442e-02 6.56894684e-01 -4.15843189e-01 -3.37308437e-01 -4.41438764e-01 1.07059233e-01 1.06979318e-01 2.08082423e-01 -2.71804154e-01 4.43510473e-01 7.73133457e-01 -1.81130016e+00 7.15754211e-01 4.66379881e-01 6.77517891e-01 -2.26851970e-01 7.51965046e-01 2.30050758e-01 7.88984358e-01 -8.78994286e-01 -9.92382109e-01 2.38677219e-01 -7.48138428e-01 -4.30345446e-01 2.52077281e-01 8.06634247e-01 -7.62027740e-01 -1.34235933e-01 1.33040845e+00 5.39295316e-01 -6.08750284e-01 4.68514204e-01 -1.72607112e+00 -8.68162394e-01 9.69582237e-03 -6.97456896e-01 1.82264343e-01 1.00200319e+00 3.18616301e-01 3.86902660e-01 -1.09109282e+00 1.00036252e+00 1.71659207e+00 5.82643807e-01 8.45080853e-01 -1.00126803e+00 -1.03323174e+00 -4.74512666e-01 1.55914947e-01 -1.69068968e+00 -1.11652148e+00 7.24784076e-01 -6.62042022e-01 5.72982073e-01 3.17817599e-01 8.87027383e-01 1.41139901e+00 4.38863572e-05 7.53428996e-01 1.39553165e+00 -5.65816641e-01 -2.46911004e-01 7.16386318e-01 -3.52381527e-01 9.89756525e-01 3.90310347e-01 1.95833594e-01 -7.20941603e-01 -3.43982965e-01 4.78012711e-01 -3.45921963e-01 1.07347950e-01 2.08654225e-01 -5.61012208e-01 7.56594419e-01 3.55144560e-01 3.44922096e-01 -2.26573527e-01 -7.62448087e-02 3.24157417e-01 2.40501329e-01 7.88130403e-01 3.44712019e-01 4.07299459e-01 -1.20688178e-01 -9.30368781e-01 4.17462796e-01 5.85602641e-01 6.44181430e-01 4.86494422e-01 3.10769631e-03 -1.27958179e-01 5.95075786e-01 3.64444464e-01 6.75072312e-01 1.15915246e-01 -7.76450396e-01 -1.70746654e-01 7.92765439e-01 2.89964497e-01 -1.68366981e+00 1.84633229e-02 3.21275771e-01 -2.08276883e-01 5.65080523e-01 5.11157513e-01 -1.98038155e-03 -6.62078440e-01 1.10967219e+00 4.82407600e-01 1.67371422e-01 -4.64560568e-01 9.97315705e-01 1.12637341e+00 3.74641776e-01 -6.91709816e-02 -3.51258703e-02 1.43274021e+00 -6.92513347e-01 -9.99028027e-01 -2.09262103e-01 6.21849477e-01 -1.04144311e+00 1.04652476e+00 2.80893534e-01 -1.03186727e+00 -3.47141847e-02 -9.93702710e-01 7.21588209e-02 -5.75163841e-01 1.85685873e-01 4.36198086e-01 1.55639541e+00 -1.09434104e+00 3.62309486e-01 -1.56606570e-01 -4.05581415e-01 1.16495526e+00 1.31931007e-01 -6.42516911e-01 -2.34475043e-02 -8.02572370e-01 1.08631778e+00 -1.35142997e-01 2.27275670e-01 -8.58161867e-01 -3.52277458e-01 -6.11750841e-01 -5.19061685e-01 3.78311545e-01 -1.24532208e-01 1.20444345e+00 -1.53383064e+00 -1.53667438e+00 1.68915462e+00 -1.43368304e-01 -1.18088730e-01 8.77356052e-01 -7.32562467e-02 -9.48776603e-01 3.28149557e-01 4.87288833e-02 7.52944112e-01 1.54577160e+00 -1.33731651e+00 -1.78839236e-01 -5.29928327e-01 -8.83475021e-02 -2.39741996e-01 -4.22154129e-01 1.01613390e+00 -4.38300446e-02 -5.39420187e-01 -4.75268751e-01 -1.01341307e+00 2.90541202e-01 3.85566920e-01 -4.43261594e-01 -1.06971696e-01 1.32269347e+00 -6.11596704e-01 1.16026950e+00 -2.31143260e+00 -5.84805429e-01 2.42997077e-03 7.25378573e-01 8.89448702e-01 4.85507883e-02 3.05303395e-01 2.63898343e-01 3.65135461e-01 9.24300551e-02 -3.85349423e-01 -1.42840043e-01 -3.01091284e-01 -2.35064924e-01 9.51744020e-01 3.35852712e-01 8.57252896e-01 -9.80368018e-01 -6.48791134e-01 1.31539539e-01 5.93958080e-01 -3.02404583e-01 -1.51033071e-03 1.44872159e-01 3.17942053e-01 -1.56957939e-01 1.17378235e+00 1.10082531e+00 1.11797504e-01 -1.09735444e-01 5.49271628e-02 -1.79216266e-01 -1.54533371e-01 -6.24861538e-01 6.64540231e-01 1.45374998e-01 1.08212256e+00 3.32514107e-01 1.18046887e-02 7.86655426e-01 2.58312732e-01 -1.40714776e-02 -5.96685529e-01 5.20575047e-01 3.30765128e-01 8.47791508e-02 -6.41328037e-01 5.52700162e-01 -1.10246591e-01 3.32922578e-01 4.30662572e-01 -2.78432429e-01 -1.69206932e-01 5.91464601e-02 1.56482309e-01 7.85205066e-01 -1.10269137e-01 3.09663266e-01 -2.93501988e-02 7.46990681e-01 -1.38763282e-02 9.76371989e-02 2.19015688e-01 -8.16310763e-01 5.43068469e-01 6.53657138e-01 -6.10452294e-01 -8.00480664e-01 -7.00015962e-01 -1.39017016e-01 1.02883816e+00 9.27312151e-02 -4.86187667e-01 -1.23458898e+00 -7.04818845e-01 1.87972084e-01 3.62943292e-01 -7.28395700e-01 -2.16542576e-02 -2.70855039e-01 -3.65543634e-01 1.10158944e+00 -2.88082749e-01 4.77882624e-01 -1.36563385e+00 -7.82233477e-01 -2.85057455e-01 1.80920139e-01 -1.27030611e+00 -4.37818497e-01 -1.21366024e+00 -4.78949249e-02 -1.33325303e+00 -8.06709409e-01 -4.15033728e-01 8.29442918e-01 4.38979328e-01 9.56821024e-01 5.45551777e-01 -7.28574157e-01 2.10632667e-01 -1.80981100e-01 -8.89724851e-01 -7.37315536e-01 -5.64218044e-01 2.64210820e-01 5.87131679e-01 5.63968182e-01 1.80953071e-01 -5.95031738e-01 5.87154567e-01 -7.23731995e-01 -4.99799699e-02 2.30754301e-01 2.45077163e-01 -1.24012753e-01 -3.51258725e-01 5.67067385e-01 -1.08871782e+00 9.44737911e-01 -3.05016041e-01 -7.12568104e-01 1.26576811e-01 -3.67505580e-01 -6.31911695e-01 1.22599088e-01 -4.56491530e-01 -8.91030788e-01 -5.15841097e-02 4.12067585e-02 -4.95648026e-01 -2.64980584e-01 -2.17067257e-01 1.78222165e-01 -7.39505708e-01 8.30738783e-01 -3.24779958e-01 3.31323862e-01 -9.48517248e-02 1.59998983e-01 1.11258900e+00 3.41836959e-01 -8.92611407e-03 8.69561434e-01 8.56987596e-01 -1.73117965e-01 -1.33401155e+00 -6.87281609e-01 -3.12854975e-01 -2.65473753e-01 -1.01222241e+00 7.43269265e-01 -6.31004632e-01 -1.00789452e+00 1.00730538e+00 -1.34816754e+00 -7.73235038e-02 3.42214435e-01 -1.10172480e-01 -9.79249030e-02 3.65443975e-01 -3.40016872e-01 -1.27185535e+00 -2.85413861e-01 -1.12521517e+00 1.14977992e+00 2.28435934e-01 -4.40365136e-01 -4.75520909e-01 -5.37799729e-04 7.19575524e-01 4.93593246e-01 5.84829390e-01 -1.00181960e-01 -1.49549201e-01 -3.58875871e-01 -4.80382025e-01 -6.51691616e-01 2.24792466e-01 5.31485910e-03 6.38140023e-01 -1.48056090e+00 -1.95950866e-01 -2.42929667e-01 -4.07764226e-01 2.89759874e-01 -2.72985604e-02 8.24915171e-01 -3.38017821e-01 -5.61531782e-01 1.23751730e-01 8.63689780e-01 -4.94237151e-03 8.83037090e-01 8.29856023e-02 6.24970436e-01 9.61755693e-01 6.37744904e-01 3.98723006e-01 1.91256702e-02 6.31784737e-01 4.53183770e-01 4.21962962e-02 -1.81689411e-01 -2.67307937e-01 6.45043850e-01 1.08283773e-01 -1.66369215e-01 -2.70228773e-01 -9.76308167e-01 3.66863251e-01 -1.27484822e+00 -1.00750935e+00 -6.16196215e-01 1.93599761e+00 1.96286216e-01 -4.00348604e-02 5.09039462e-01 9.69454497e-02 1.17289925e+00 4.06572297e-02 -1.66183397e-01 -7.58962929e-01 -2.06703156e-01 -8.86665937e-03 3.12913507e-01 2.59614855e-01 -9.91867602e-01 1.25393343e+00 7.27460289e+00 7.29979753e-01 -1.34779835e+00 3.39768499e-01 5.92610121e-01 -5.66738769e-02 1.86830219e-02 -9.56901088e-02 -7.58020997e-01 6.06007397e-01 7.47652769e-01 -1.76746979e-01 4.26912010e-01 7.48533964e-01 2.72026628e-01 -2.65037328e-01 -7.32778788e-01 1.17225873e+00 6.13579750e-01 -1.24484694e+00 -1.15436815e-01 3.83308589e-01 6.17095888e-01 -3.61453950e-01 3.84192556e-01 -1.93454742e-01 -1.18447050e-01 -1.29826784e+00 7.90678561e-01 2.57616907e-01 1.18743229e+00 -6.97846889e-01 5.00332057e-01 -4.56340574e-02 -5.40127635e-01 3.15578096e-02 -1.75743565e-01 -1.93226486e-01 1.17768392e-01 4.38060969e-01 -1.21322072e+00 -2.52546579e-01 7.21978247e-01 4.18121606e-01 -9.04746413e-01 8.77481341e-01 -2.02736109e-01 3.16123962e-01 -5.08428179e-02 -1.22292362e-01 1.38816955e-02 -1.65395394e-01 6.75808311e-01 9.71919417e-01 9.83077809e-02 -2.40541682e-01 -2.61037439e-01 1.02906823e+00 -2.72814602e-01 1.47900999e-01 -1.18094790e+00 -7.76970208e-01 1.34100512e-01 1.39988041e+00 -8.12839389e-01 -1.04602344e-01 -4.57238495e-01 1.23009348e+00 1.47187203e-01 -1.54090434e-01 -7.64591932e-01 -2.18652934e-01 7.29831576e-01 8.06725502e-01 -2.57098645e-01 5.65862469e-02 2.05170661e-02 -8.61905336e-01 -9.38055888e-02 -1.07879758e+00 4.50008623e-02 -9.64016914e-01 -1.22478282e+00 7.30773032e-01 -1.20843820e-01 -9.70464408e-01 2.53874604e-02 -7.30945230e-01 -5.21645188e-01 5.42162180e-01 -9.93648052e-01 -1.32669139e+00 -5.08056223e-01 4.87006843e-01 1.87011376e-01 -5.00026464e-01 8.27999055e-01 2.82964647e-01 -7.43315935e-01 7.68900394e-01 -4.18131471e-01 8.03552270e-02 8.99066865e-01 -5.06253242e-01 3.57122779e-01 6.63838804e-01 -1.01179540e-01 3.72399688e-01 8.34754348e-01 -7.93568909e-01 -1.30072272e+00 -5.73926449e-01 6.76528931e-01 -7.66834736e-01 7.35086083e-01 -6.16985202e-01 -5.60912609e-01 4.63902384e-01 3.32730651e-01 2.92593807e-01 6.72906101e-01 -5.08351803e-01 -4.97223616e-01 1.79094344e-01 -1.75612569e+00 7.89756954e-01 9.54617441e-01 -7.26128817e-01 -2.04264104e-01 4.63240892e-01 -6.36800155e-02 -1.71058893e-01 -3.89498591e-01 3.13304439e-02 1.01493621e+00 -1.37301075e+00 7.82215953e-01 -2.22563952e-01 3.62762630e-01 -3.42994481e-01 4.83092219e-01 -1.01794028e+00 3.03776357e-02 -8.53696227e-01 -1.08987756e-01 1.16515255e+00 1.38563663e-01 -8.89502645e-01 9.22983229e-01 4.74336147e-01 3.48010451e-01 -2.84217864e-01 -7.13979483e-01 -5.48525393e-01 -4.26463306e-01 -4.18300740e-02 4.61756200e-01 9.55829024e-01 -5.33849448e-02 -1.06643979e-02 -5.18885791e-01 -2.34254152e-02 7.06246138e-01 -2.96308815e-01 1.05871773e+00 -1.27654922e+00 2.77502924e-01 -6.71187639e-01 -8.59779894e-01 -2.00767610e-02 2.38263056e-01 -5.07335484e-01 -4.05875981e-01 -7.59226024e-01 2.28138909e-01 4.79539558e-02 4.32964414e-01 1.43295854e-01 3.43554281e-02 1.11407316e+00 3.58020008e-01 1.84252068e-01 -4.17352170e-01 9.01499763e-02 1.40769935e+00 9.36958343e-02 7.77996406e-02 -3.41747433e-01 -6.82093263e-01 9.00365591e-01 1.03912520e+00 -5.74133635e-01 -1.97235309e-02 -5.49582653e-02 4.07204777e-01 -5.13801932e-01 5.81153810e-01 -9.73882198e-01 -7.52394795e-02 -9.40559655e-02 3.34505260e-01 -5.13547063e-01 4.75723654e-01 -5.17026544e-01 1.00947864e-01 4.67606395e-01 2.92718410e-01 2.61699017e-02 2.63167918e-01 3.56687278e-01 3.21153328e-02 -3.92435104e-01 1.18204856e+00 -2.82851398e-01 -3.92407089e-01 7.27611333e-02 -6.79791927e-01 -1.62053570e-01 1.54009759e+00 -5.24071455e-01 -6.38640583e-01 -4.97593313e-01 -3.43077630e-01 -1.76550254e-01 9.65654671e-01 4.90032673e-01 8.06592584e-01 -1.24114799e+00 -8.61071229e-01 4.08537745e-01 4.93184745e-01 -7.66313434e-01 1.53243631e-01 7.78991282e-01 -1.10463154e+00 9.46740434e-02 -6.92787290e-01 -4.08614308e-01 -1.81218147e+00 5.89712024e-01 2.23993391e-01 4.44367081e-01 -3.66107047e-01 8.15206170e-01 7.45391175e-02 -1.33499384e-01 -2.48804122e-01 5.74005187e-01 -1.59686014e-01 1.09877862e-01 1.12669361e+00 8.68497789e-01 -4.54334691e-02 -1.52614474e+00 -5.50251842e-01 1.16299838e-01 -1.18830264e-01 6.34071603e-02 8.08226287e-01 -7.21167102e-02 -5.18503547e-01 1.81107730e-01 1.07127738e+00 5.41553855e-01 -7.79334724e-01 3.28030616e-01 4.57170652e-03 -1.26624060e+00 -1.17361825e-02 -4.36863303e-01 -9.89553928e-01 5.03927946e-01 6.18578076e-01 3.70319217e-01 6.50456667e-01 -1.08909386e-03 7.30294228e-01 -7.32427761e-02 4.64190215e-01 -9.79033589e-01 2.99219847e-01 1.33359760e-01 1.19484341e+00 -1.30350256e+00 4.12988625e-02 -8.75718892e-01 -6.75915122e-01 8.54781985e-01 7.02038407e-01 -2.21330747e-01 5.46116531e-01 2.25081101e-01 3.04444820e-01 -6.23328626e-01 -1.92360476e-01 -1.46490127e-01 2.56074846e-01 7.71191180e-01 4.56026763e-01 2.14255527e-01 -2.59147167e-01 2.23139562e-02 -3.13418210e-01 1.62449375e-01 7.81362295e-01 6.88382924e-01 -2.48268828e-01 -8.94466996e-01 -9.25974071e-01 4.27299529e-01 -8.05445433e-01 2.34968215e-01 -1.44233906e+00 5.81286252e-01 2.09956989e-01 1.20487118e+00 -1.51300088e-01 -3.94378811e-01 1.12491086e-01 2.36227009e-02 6.87360823e-01 -4.71258223e-01 -6.42719865e-01 -4.20166016e-01 4.05548245e-01 -6.67660892e-01 -3.96275580e-01 -5.97321808e-01 -5.70154667e-01 -8.85687411e-01 -4.44288760e-01 -4.96076345e-01 7.99432099e-01 5.54644823e-01 4.53063339e-01 -1.39507249e-01 5.43636560e-01 -1.13485599e+00 2.28111967e-01 -8.89910340e-01 -5.08335710e-01 4.35127676e-01 3.15600663e-01 -9.63241041e-01 -6.26876831e-01 -2.58456409e-01]
[12.651317596435547, 1.0952308177947998]
db66dbaf-ae9d-43c0-920f-195865f63586
towards-contextual-spelling-correction-for
2203.00888
null
https://arxiv.org/abs/2203.00888v2
https://arxiv.org/pdf/2203.00888v2.pdf
Towards Contextual Spelling Correction for Customization of End-to-end Speech Recognition Systems
Contextual biasing is an important and challenging task for end-to-end automatic speech recognition (ASR) systems, which aims to achieve better recognition performance by biasing the ASR system to particular context phrases such as person names, music list, proper nouns, etc. Existing methods mainly include contextual LM biasing and adding bias encoder into end-to-end ASR models. In this work, we introduce a novel approach to do contextual biasing by adding a contextual spelling correction model on top of the end-to-end ASR system. We incorporate contextual information into a sequence-to-sequence spelling correction model with a shared context encoder. Our proposed model includes two different mechanisms: autoregressive (AR) and non-autoregressive (NAR). We propose filtering algorithms to handle large-size context lists, and performance balancing mechanisms to control the biasing degree of the model. We demonstrate the proposed model is a general biasing solution which is domain-insensitive and can be adopted in different scenarios. Experiments show that the proposed method achieves as much as 51% relative word error rate (WER) reduction over ASR system and outperforms traditional biasing methods. Compared to the AR solution, the proposed NAR model reduces model size by 43.2% and speeds up inference by 2.1 times.
['Hosam Khalil', 'Sheng Zhao', 'Veljko Miljanic', 'Jinyu Li', 'Yanqing Liu', 'Xiaoqiang Wang']
2022-03-02
null
null
null
null
['spelling-correction']
['natural-language-processing']
[ 4.23393756e-01 -2.94761956e-01 -2.77080655e-01 -6.96265578e-01 -1.07293725e+00 -4.25855100e-01 5.70508718e-01 -2.01826379e-01 -8.19556653e-01 6.01258337e-01 7.64548779e-01 -8.23240817e-01 2.50473768e-01 -2.99380302e-01 -6.32772505e-01 -5.54219186e-01 5.76065600e-01 4.01472300e-01 2.91242659e-01 -5.32009780e-01 5.11019826e-01 5.68890452e-01 -1.07610786e+00 4.18346852e-01 7.57941425e-01 5.21246731e-01 6.85567200e-01 9.80589688e-01 -2.04181582e-01 6.25417054e-01 -9.43029046e-01 -1.42642081e-01 6.20501004e-02 -1.99085146e-01 -5.06063282e-01 -1.17047049e-01 3.60755235e-01 -1.54017612e-01 -5.92739582e-01 9.56031024e-01 1.03108442e+00 4.68380421e-01 4.07468885e-01 -5.62691391e-01 -7.88943589e-01 1.08113301e+00 -3.58467400e-01 5.51579595e-01 3.16484272e-01 5.97359538e-02 9.28559899e-01 -1.17280900e+00 1.04124308e-01 1.44940484e+00 2.99055368e-01 1.20387387e+00 -1.08326447e+00 -7.40017295e-01 6.98607385e-01 8.40505660e-02 -1.53891599e+00 -9.68072474e-01 6.65890396e-01 -1.02477111e-01 1.25516152e+00 6.10175848e-01 1.09166391e-02 1.18651748e+00 -2.96847731e-01 9.84321952e-01 9.59467769e-01 -6.81355059e-01 2.09177524e-01 3.16560864e-02 4.54219490e-01 2.08900403e-02 1.07967488e-01 2.02424988e-01 -7.82579720e-01 -1.28254861e-01 3.67978364e-01 -2.64171153e-01 -3.70134413e-01 3.04169804e-01 -1.29445457e+00 4.70217407e-01 1.86598115e-02 1.98470071e-01 -3.69732648e-01 2.07820728e-01 3.78896475e-01 1.33695483e-01 3.34850788e-01 2.70576596e-01 -6.70531929e-01 -2.60282248e-01 -1.16398430e+00 2.65603274e-01 3.17491710e-01 1.32432008e+00 2.02508733e-01 5.48197210e-01 -8.04068148e-01 1.52370715e+00 6.01550341e-01 1.08324814e+00 1.03464186e+00 -3.23070586e-01 7.18410730e-01 4.28344235e-02 2.04390079e-01 -4.21352774e-01 -1.04396097e-01 -5.93967676e-01 -6.08193517e-01 -4.71922636e-01 9.46261883e-02 -8.54834691e-02 -1.44449329e+00 1.80668414e+00 1.02815710e-01 -5.49631380e-03 1.69317320e-01 1.02304125e+00 6.85412049e-01 9.16891754e-01 1.99701816e-01 -2.99460232e-01 1.34967732e+00 -9.93161559e-01 -1.08782995e+00 -6.13854468e-01 6.88252151e-01 -1.08265257e+00 1.47437227e+00 3.63684028e-01 -1.14508915e+00 -5.63816607e-01 -1.05245709e+00 -6.97190389e-02 -2.66175568e-01 4.69082564e-01 -1.36307985e-01 8.57111335e-01 -8.19485903e-01 2.40596116e-01 -4.70894992e-01 -1.53678954e-01 -4.65669706e-02 3.46828580e-01 6.36262447e-02 -1.58186153e-01 -1.46357548e+00 8.77515137e-01 2.82122552e-01 7.15112537e-02 -7.64204502e-01 -6.72819555e-01 -6.27111673e-01 -1.42137960e-01 2.97862709e-01 -3.20036918e-01 1.57667816e+00 -9.10283327e-01 -1.89121675e+00 6.27222419e-01 -8.62503290e-01 -4.19563621e-01 1.29340142e-01 -6.46761239e-01 -7.92424262e-01 -5.25933862e-01 -4.63422447e-01 2.39814192e-01 8.48033607e-01 -1.06142044e+00 -6.95269465e-01 -3.14565212e-01 -3.32482874e-01 3.22230726e-01 -2.42233440e-01 5.69910467e-01 -5.16463280e-01 -1.20308447e+00 1.75405547e-01 -9.69273329e-01 -1.81776509e-01 -8.31525326e-01 -2.43805334e-01 -2.66785294e-01 5.63632309e-01 -1.09429610e+00 2.01942611e+00 -2.09694076e+00 6.39691055e-02 1.98401064e-01 -4.87613291e-01 8.71084273e-01 -1.54044956e-01 3.34692329e-01 -2.97583714e-02 6.05936125e-02 -2.44013488e-01 -1.67910025e-01 8.47704783e-02 3.25752586e-01 -4.63449836e-01 2.60341704e-01 3.24032716e-02 5.17649293e-01 -8.41138780e-01 -1.63396999e-01 2.38062963e-01 5.18824875e-01 -6.20062768e-01 5.43831885e-01 -1.73218921e-01 1.75967455e-01 1.09879263e-01 4.70626324e-01 5.84487975e-01 5.83799243e-01 8.45439434e-02 1.07254542e-01 -4.72699627e-02 1.19201672e+00 -1.36720610e+00 1.53987110e+00 -8.11555982e-01 4.42342818e-01 1.27215549e-01 -6.04595125e-01 1.28660429e+00 4.69680637e-01 -3.81370306e-01 -5.15133858e-01 6.37631416e-02 3.94516617e-01 2.42697269e-01 -1.50056615e-01 8.45994413e-01 -2.18704388e-01 6.42484277e-02 -5.41989617e-02 -1.56414196e-01 -1.00620333e-02 -4.82145995e-02 -7.06565455e-02 8.37160349e-01 -1.92946851e-01 4.24786240e-01 -2.59467602e-01 1.06606328e+00 -3.79619390e-01 7.51303315e-01 8.05326223e-01 -1.62449658e-01 6.19140148e-01 -2.89545417e-01 -1.13173328e-01 -1.01485527e+00 -7.17837572e-01 5.48177958e-02 1.48314261e+00 -2.32105061e-01 -3.37509990e-01 -7.31716573e-01 -6.41946673e-01 -1.16678908e-01 1.44457328e+00 -1.13362111e-01 -2.50617653e-01 -1.01231229e+00 -6.23825848e-01 7.23484993e-01 7.12650239e-01 2.67966181e-01 -8.99226308e-01 4.55894768e-01 2.23629221e-01 -2.21442938e-01 -1.09279943e+00 -1.06710935e+00 1.51832059e-01 -7.12016702e-01 -1.75644919e-01 -7.75713265e-01 -6.09808028e-01 4.42925990e-01 4.00812954e-01 8.93226862e-01 -6.83371723e-02 3.81217003e-01 -5.02683269e-03 -5.22827506e-01 -3.93968672e-01 -6.77118123e-01 2.71324664e-01 3.03386867e-01 7.91903287e-02 8.27948093e-01 -1.85887411e-01 -5.68061292e-01 6.82471335e-01 -6.91300929e-01 -2.71906674e-01 7.84513175e-01 9.59778070e-01 6.33019328e-01 -5.73332131e-01 7.41863906e-01 -9.40728366e-01 7.10448265e-01 -2.39706293e-01 -4.63406563e-01 1.76430091e-01 -8.63910019e-01 1.94464937e-01 6.67346299e-01 -5.94329476e-01 -1.29138672e+00 -5.48100397e-02 -7.92142689e-01 -1.41112164e-01 -2.58188665e-01 1.17121145e-01 -7.99533725e-01 4.36113834e-01 6.51369393e-01 5.27817428e-01 -4.02438432e-01 -6.34235203e-01 5.88719726e-01 1.60346532e+00 4.01909828e-01 -6.39991641e-01 6.23990417e-01 -9.15245935e-02 -4.44304556e-01 -9.55849171e-01 -6.58356965e-01 -8.82116735e-01 -3.73908758e-01 4.05630609e-03 4.39576983e-01 -1.00955451e+00 -1.85830951e-01 2.30376527e-01 -1.20658696e+00 -3.58853996e-01 -4.82527949e-02 7.03052580e-01 -2.94386625e-01 2.56680697e-01 -4.65979397e-01 -1.14542043e+00 -6.64574504e-01 -1.05241120e+00 1.04588199e+00 -1.47230774e-01 -4.69429702e-01 -5.33904493e-01 -3.88863273e-02 4.17095512e-01 6.48323715e-01 -9.27621603e-01 6.05379999e-01 -1.03988230e+00 -1.23793006e-01 -1.43576384e-01 -3.66325676e-02 8.44066262e-01 -6.37267381e-02 -2.63206124e-01 -1.07906985e+00 -3.48293513e-01 -2.26804152e-01 4.01382774e-01 7.85316765e-01 2.54989326e-01 1.20785153e+00 -4.09063876e-01 -1.52840197e-01 3.98786187e-01 1.12318671e+00 7.25449622e-01 7.89477468e-01 1.46586150e-01 8.60430956e-01 3.14465255e-01 8.75234067e-01 1.59263566e-01 -2.32223216e-02 9.26133454e-01 -5.20715639e-02 2.40013123e-01 -5.53355396e-01 -4.24006790e-01 6.83707595e-01 1.56787050e+00 2.31371924e-01 -4.41309661e-01 -9.20738995e-01 8.57967556e-01 -1.67081320e+00 -8.11526299e-01 -8.05679634e-02 2.55775595e+00 1.06675720e+00 2.24691004e-01 -8.04098025e-02 2.87874460e-01 1.03885269e+00 2.18445569e-01 -2.56521434e-01 -9.12715673e-01 -1.17593534e-01 3.92632931e-02 8.73401403e-01 1.04422426e+00 -6.41242981e-01 1.37742639e+00 6.08275843e+00 1.20273638e+00 -1.12081373e+00 1.95619121e-01 2.86793947e-01 -2.78177053e-01 -3.85771453e-01 -2.61487626e-03 -1.42863393e+00 6.02849603e-01 1.24670732e+00 3.45668569e-02 6.86994970e-01 1.03136051e+00 5.95746279e-01 4.66497511e-01 -1.08650446e+00 1.15970540e+00 1.84680238e-01 -7.93786108e-01 4.31751639e-01 -2.73690104e-01 4.51951742e-01 4.37317155e-02 -1.25698417e-01 4.93834883e-01 2.32652783e-01 -8.38096857e-01 7.79756010e-01 3.32827806e-01 7.05039799e-01 -8.49021792e-01 5.72327077e-01 2.75646597e-01 -8.00279617e-01 -8.57523754e-02 -5.18300295e-01 -4.30454426e-02 1.73592374e-01 6.46110654e-01 -9.77129459e-01 2.23573908e-01 2.35484317e-01 1.38701394e-01 -1.77003890e-01 7.25818515e-01 -4.08897102e-01 1.29400969e+00 1.34963645e-02 -5.16298890e-01 -3.64101911e-03 1.88413635e-01 9.36072409e-01 1.91888249e+00 2.75174946e-01 9.89266783e-02 -2.03416035e-01 1.29439443e-01 -1.76068068e-01 5.08570790e-01 -2.96741277e-01 1.95153743e-01 1.02169156e+00 6.71239614e-01 -5.45732267e-02 -5.96031010e-01 -2.72327811e-01 9.57090676e-01 1.79164827e-01 3.38590890e-01 -6.84560955e-01 -5.47777653e-01 9.37921226e-01 8.05561543e-02 3.49371672e-01 -3.62987995e-01 -4.73521918e-01 -1.00927639e+00 -4.13202941e-02 -1.32826209e+00 1.29554063e-01 -7.61420429e-01 -1.06022906e+00 6.13878071e-01 -2.35818446e-01 -8.32589388e-01 -1.37888046e-03 -7.23954260e-01 -2.17391148e-01 1.23658681e+00 -1.78684103e+00 -8.00049484e-01 9.67452601e-02 3.60906065e-01 1.19029522e+00 -4.80085522e-01 5.96835911e-01 6.26773179e-01 -6.29578292e-01 9.98350143e-01 1.16795458e-01 1.12215824e-01 1.00982094e+00 -1.28427351e+00 5.69889665e-01 1.34272993e+00 1.77256852e-01 1.09986317e+00 8.44736636e-01 -6.67969763e-01 -1.30786180e+00 -1.29484689e+00 1.39204085e+00 -4.03761953e-01 2.92180210e-01 -4.44231004e-01 -9.27662790e-01 5.62276363e-01 1.32511318e-01 -2.12390617e-01 7.29522109e-01 3.00006300e-01 -4.45445687e-01 -5.26427686e-01 -8.19223583e-01 9.62054253e-01 1.19362569e+00 -5.70163786e-01 -9.91064727e-01 3.44742909e-02 1.10860038e+00 -5.34524739e-01 -4.33133334e-01 3.91771585e-01 4.94023323e-01 -2.53803581e-01 7.44964421e-01 -6.96758926e-01 -8.37630481e-02 -6.36774719e-01 -7.90127099e-01 -1.65066683e+00 -5.51107585e-01 -9.02377009e-01 -2.01389208e-01 1.58111572e+00 6.38139665e-01 -4.89102364e-01 4.20505762e-01 4.95846093e-01 -4.68128651e-01 -5.27928829e-01 -8.37498903e-01 -9.50263381e-01 1.12146124e-01 -8.14329326e-01 1.04181802e+00 5.43601274e-01 -2.75792599e-01 6.34860575e-01 -5.39634824e-01 5.18478572e-01 1.36670291e-01 -4.86425608e-01 6.68400764e-01 -4.90578622e-01 -3.09412211e-01 -4.98412013e-01 -9.34134275e-02 -1.79448092e+00 1.41930208e-02 -7.88262010e-01 4.24672335e-01 -1.24103439e+00 -2.36765698e-01 -4.79182452e-01 -7.00831711e-01 1.18864834e-01 -5.11802793e-01 -1.94509625e-01 8.93669426e-02 -1.28155321e-01 -1.26567990e-01 5.00532806e-01 7.51905620e-01 -5.73164709e-02 -2.97784477e-01 2.62556374e-01 -6.41081452e-01 5.10347843e-01 9.87769842e-01 -5.12773037e-01 -3.09753805e-01 -6.89450741e-01 5.03289551e-02 -2.10367188e-01 -2.45790631e-01 -7.35507965e-01 2.46952057e-01 -1.69527620e-01 3.38379554e-02 -7.34951735e-01 1.74498051e-01 -8.04315090e-01 -3.99320573e-01 1.30034640e-01 -7.48820662e-01 1.05695933e-01 2.20812976e-01 4.64085549e-01 -1.76434681e-01 -3.49687546e-01 8.56972098e-01 1.16832674e-01 -5.59008121e-01 -7.78205767e-02 -6.21443987e-01 8.37895423e-02 5.05886137e-01 1.11212373e-01 -1.53573051e-01 -3.35354656e-01 -4.77427721e-01 -5.09205693e-03 -8.41418877e-02 6.55504346e-01 6.25894964e-01 -1.35473001e+00 -9.23100054e-01 1.86800286e-01 2.34965950e-01 -2.48677418e-01 8.07782859e-02 4.93762463e-01 -1.01086900e-01 7.96119034e-01 5.08206964e-01 -4.31944907e-01 -1.65543509e+00 6.91504776e-01 2.32349455e-01 -3.69171426e-03 -1.50259495e-01 1.07774270e+00 -8.99494961e-02 -6.08588457e-01 6.64015591e-01 -4.37711358e-01 -1.02522880e-01 -3.05536687e-01 7.98763812e-01 5.68538964e-01 5.19850433e-01 -8.52578223e-01 -5.16897202e-01 3.59939188e-01 -4.04590994e-01 -4.07084167e-01 7.76706219e-01 -5.28057635e-01 2.23780140e-01 5.86827993e-01 7.69755363e-01 5.42379022e-01 -7.64875412e-01 -5.14549613e-01 1.72518522e-01 -5.65361798e-01 3.92517418e-01 -1.15647805e+00 -5.90157926e-01 9.83029604e-01 5.26381433e-01 -1.63983986e-01 1.05098510e+00 -3.80253285e-01 1.05456913e+00 3.70231092e-01 1.14458036e-02 -1.60096180e+00 -4.49080020e-01 7.86012888e-01 1.06208789e+00 -1.02410936e+00 -2.87435323e-01 -2.64303237e-01 -7.18585432e-01 7.76944637e-01 5.62164903e-01 -2.79081110e-02 5.06062388e-01 1.25125259e-01 4.24326479e-01 5.05632102e-01 -9.37794387e-01 -3.39029789e-01 4.49152529e-01 5.70976138e-01 8.04176688e-01 3.49167347e-01 -8.76694858e-01 8.32975268e-01 -2.75643975e-01 -3.46847653e-01 3.21882874e-01 6.48434222e-01 -6.19771838e-01 -1.41319704e+00 -6.13928974e-01 1.61965817e-01 -6.86634243e-01 -5.42864084e-01 -2.71450043e-01 2.25005940e-01 -1.75566345e-01 1.26835978e+00 -2.07925990e-01 -4.00187582e-01 7.16928720e-01 5.17796278e-01 1.44063577e-01 -7.47685492e-01 -6.41693711e-01 1.86499536e-01 4.10499930e-01 -3.34752083e-01 6.17938899e-02 -6.40594304e-01 -9.99531984e-01 -1.56259108e-02 -6.25446320e-01 1.19482979e-01 1.09801328e+00 9.22537386e-01 4.39882487e-01 8.07947993e-01 7.75231719e-01 -3.52293849e-01 -7.78358221e-01 -1.46406209e+00 -2.46005923e-01 2.21713290e-01 4.03948843e-01 -3.69503438e-01 -3.71705621e-01 -6.79822415e-02]
[14.393120765686035, 6.755378246307373]
d8db907c-d27f-4ae7-a779-8a9ee1cc110a
sandwich-batch-normalization-1
2102.11382
null
https://arxiv.org/abs/2102.11382v2
https://arxiv.org/pdf/2102.11382v2.pdf
Sandwich Batch Normalization: A Drop-In Replacement for Feature Distribution Heterogeneity
We present Sandwich Batch Normalization (SaBN), a frustratingly easy improvement of Batch Normalization (BN) with only a few lines of code changes. SaBN is motivated by addressing the inherent feature distribution heterogeneity that one can be identified in many tasks, which can arise from data heterogeneity (multiple input domains) or model heterogeneity (dynamic architectures, model conditioning, etc.). Our SaBN factorizes the BN affine layer into one shared sandwich affine layer, cascaded by several parallel independent affine layers. Concrete analysis reveals that, during optimization, SaBN promotes balanced gradient norms while still preserving diverse gradient directions -- a property that many application tasks seem to favor. We demonstrate the prevailing effectiveness of SaBN as a drop-in replacement in four tasks: conditional image generation, neural architecture search (NAS), adversarial training, and arbitrary style transfer. Leveraging SaBN immediately achieves better Inception Score and FID on CIFAR-10 and ImageNet conditional image generation with three state-of-the-art GANs; boosts the performance of a state-of-the-art weight-sharing NAS algorithm significantly on NAS-Bench-201; substantially improves the robust and standard accuracies for adversarial defense; and produces superior arbitrary stylized results. We also provide visualizations and analysis to help understand why SaBN works. Codes are available at: https://github.com/VITA-Group/Sandwich-Batch-Normalization.
['Zhangyang Wang', 'Tianlong Chen', 'Wuyang Chen', 'Xinyu Gong']
2021-02-22
sandwich-batch-normalization
https://openreview.net/forum?id=A-Sp6CR9-AA
https://openreview.net/pdf?id=A-Sp6CR9-AA
null
['conditional-image-generation']
['computer-vision']
[ 1.46006152e-01 -1.74161494e-01 -1.09899357e-01 -4.84071016e-01 -9.74781990e-01 -8.71430874e-01 8.78616929e-01 -7.64479280e-01 -4.95162666e-01 7.19707966e-01 2.74870783e-01 -4.93808538e-01 1.94189116e-01 -3.67609948e-01 -9.82693136e-01 -8.63562167e-01 1.90455735e-01 3.53647023e-01 -1.15854494e-01 -5.07694364e-01 -1.79535627e-01 5.78480184e-01 -8.04455817e-01 3.67607176e-01 4.55556810e-01 1.07294595e+00 -2.06152067e-01 7.77176321e-01 1.80192858e-01 9.69322920e-01 -7.55476654e-01 -9.95385647e-01 7.05760121e-01 -4.58540320e-01 -7.12052763e-01 -3.67850393e-01 9.37181830e-01 -3.03686410e-01 -3.57509911e-01 9.88871813e-01 7.48648882e-01 -1.09224301e-02 6.90917671e-01 -1.43115926e+00 -9.39171016e-01 6.91421330e-01 -5.56362867e-01 2.74403930e-01 -3.90139610e-01 7.41574109e-01 9.58058119e-01 -8.65117252e-01 4.72527802e-01 1.24040639e+00 8.49109530e-01 1.02134073e+00 -1.65677929e+00 -1.09330869e+00 2.12097123e-01 -1.47174940e-01 -1.20926952e+00 -7.07579136e-01 6.82701766e-01 -5.06696463e-01 7.96960652e-01 3.88054729e-01 2.37484396e-01 1.89622998e+00 1.93244562e-01 6.32312894e-01 9.35326040e-01 -5.26223443e-02 8.90870318e-02 -1.69594474e-02 -1.85895845e-01 4.70910758e-01 2.75493823e-02 2.59238005e-01 -4.10798430e-01 -1.49495363e-01 8.56784582e-01 -1.42726302e-01 -1.92421660e-01 -1.34167716e-01 -1.13125527e+00 8.05804372e-01 6.53752327e-01 1.18857473e-01 -2.71095484e-01 5.35319149e-01 4.95172292e-01 5.63110471e-01 5.14024675e-01 6.52481258e-01 -6.31721199e-01 -6.93795234e-02 -1.04158437e+00 4.53182012e-01 5.90534806e-01 9.53505456e-01 3.83649230e-01 5.79546154e-01 -5.54848850e-01 8.74916494e-01 -7.24347755e-02 5.82432508e-01 6.40721917e-01 -8.25272918e-01 4.60288018e-01 2.75810389e-03 -2.88682938e-01 -5.20282149e-01 -2.31585041e-01 -8.55648756e-01 -1.29078424e+00 5.76090038e-01 4.98301119e-01 -3.53936940e-01 -1.07311630e+00 2.25280690e+00 4.98885997e-02 -7.43633285e-02 -9.96490493e-02 7.92986453e-01 5.86467862e-01 4.43954557e-01 2.05257267e-01 4.78543371e-01 1.27237511e+00 -1.25405073e+00 -3.84042263e-01 -4.45527971e-01 3.31240743e-01 -8.71938348e-01 1.38118005e+00 2.81542182e-01 -1.21489251e+00 -6.26397431e-01 -9.78165686e-01 -1.02809116e-01 -4.19154555e-01 1.15856081e-01 5.53271234e-01 7.55521536e-01 -1.22670603e+00 4.82891291e-01 -7.57279992e-01 -2.45941225e-02 8.57029319e-01 3.55054349e-01 -4.71635252e-01 1.71810295e-02 -9.89828706e-01 8.13411832e-01 8.64414275e-02 -5.10170422e-02 -1.20271254e+00 -1.30824780e+00 -6.50801241e-01 1.22664664e-02 1.84789494e-01 -1.07141149e+00 1.29807281e+00 -1.40385163e+00 -1.69947696e+00 7.96865165e-01 2.76333671e-02 -7.35708714e-01 9.21027720e-01 -3.53623599e-01 -3.87029737e-01 -3.53261918e-01 -1.57177344e-01 8.79680336e-01 1.23299348e+00 -1.07942677e+00 -1.37121230e-01 -1.49212390e-01 1.40034463e-02 -4.81622070e-02 -3.24132413e-01 8.51543471e-02 -2.96049982e-01 -1.16861081e+00 -4.28665459e-01 -1.00013971e+00 -1.71112552e-01 -7.33269006e-02 -5.41538417e-01 9.63964015e-02 6.11341894e-01 -7.92459548e-01 1.04390216e+00 -2.24024272e+00 2.43653789e-01 1.53687015e-01 2.25348800e-01 2.87348360e-01 -5.91366768e-01 1.15136661e-01 -4.51891094e-01 3.89858961e-01 -4.01363015e-01 -6.87209070e-01 1.81692228e-01 3.61621454e-02 -6.80608153e-01 3.84243727e-01 3.62193137e-01 1.32472491e+00 -3.94206911e-01 -5.25767021e-02 -6.93388209e-02 5.55201769e-01 -7.62321651e-01 2.41436675e-01 -1.77015439e-01 6.70508564e-01 7.20547289e-02 5.99299967e-01 8.69703114e-01 -1.73652411e-01 -1.00709416e-01 -3.94531339e-01 1.86428174e-01 2.02551126e-01 -7.78787255e-01 1.78595078e+00 -4.48236942e-01 5.97051859e-01 1.23864241e-01 -7.49245226e-01 5.36529958e-01 8.14426020e-02 4.28816788e-02 -7.77430534e-01 9.50638503e-02 7.24510625e-02 1.90920755e-02 1.23327471e-01 2.38128021e-01 -8.27924907e-02 -2.93566789e-02 1.79306239e-01 5.11034548e-01 -1.72872931e-01 -5.62050939e-02 2.33700722e-01 1.08739853e+00 3.62863415e-04 2.23278329e-02 -5.18075168e-01 2.69933045e-01 -2.85116583e-01 5.03848851e-01 9.32231188e-01 -5.31995669e-02 8.27452004e-01 5.87636352e-01 -5.69495082e-01 -1.20757627e+00 -1.39884758e+00 3.41306962e-02 1.45062220e+00 -3.69108647e-01 -3.00321788e-01 -9.95759606e-01 -7.89687097e-01 -3.34621556e-02 7.50862300e-01 -1.02514207e+00 -2.58485764e-01 -5.76277077e-01 -1.00194037e+00 1.06771064e+00 6.28294408e-01 5.98941624e-01 -1.02002287e+00 -1.71838433e-01 -1.90104976e-01 1.11424007e-01 -9.15336490e-01 -8.84785295e-01 3.05078447e-01 -5.72514176e-01 -6.51577592e-01 -7.59171844e-01 -5.47692776e-01 5.41520119e-01 -1.93003476e-01 1.54242432e+00 -1.49373084e-01 -1.93646669e-01 2.32981369e-01 -5.07353954e-02 -5.26509583e-01 -5.50301194e-01 4.16984469e-01 6.17552400e-02 4.15460169e-02 -2.29989544e-01 -7.28429973e-01 -6.74137592e-01 3.46531391e-01 -1.09207880e+00 1.62019983e-01 6.11058295e-01 1.13445389e+00 5.05436540e-01 -4.65261310e-01 4.21964139e-01 -1.05118644e+00 6.60459280e-01 -4.42564070e-01 -7.40098298e-01 2.43084416e-01 -5.98375916e-01 1.56153426e-01 7.57942379e-01 -5.67042172e-01 -1.00804353e+00 -2.64047682e-01 -4.44207281e-01 -6.65522814e-01 -1.19900756e-01 2.34465059e-02 -2.45329469e-01 -1.54257804e-01 1.05550325e+00 1.09889008e-01 6.60507306e-02 -4.43502069e-01 6.38651371e-01 9.84727591e-02 6.91575825e-01 -7.85082102e-01 1.12195194e+00 3.12847555e-01 -1.54978752e-01 -3.78913164e-01 -8.30258489e-01 -2.23046821e-02 -4.53678757e-01 1.54106408e-01 8.63075078e-01 -1.15017557e+00 -5.44070065e-01 7.83006370e-01 -1.11466563e+00 -1.03773534e+00 -4.97063309e-01 2.63721608e-02 -3.72303873e-01 -3.11190277e-01 -7.10759342e-01 -2.64242440e-01 -7.12764859e-01 -1.21718097e+00 6.93259716e-01 7.32672438e-02 -2.41334960e-01 -9.16220784e-01 9.34036225e-02 3.22564125e-01 1.13646865e+00 2.37385422e-01 9.06717896e-01 -6.94428921e-01 -2.99894124e-01 6.84995800e-02 -1.20004319e-01 8.54337811e-01 -5.57314120e-02 -5.14973253e-02 -1.33995628e+00 -5.69780052e-01 3.53664458e-02 -4.30073649e-01 9.97477353e-01 3.57072353e-01 1.46972966e+00 -5.65039277e-01 1.52133480e-01 1.29095232e+00 1.29545879e+00 -5.08287223e-03 7.98493683e-01 2.77243912e-01 9.20625210e-01 1.33997589e-01 -1.74799278e-01 2.02269137e-01 -1.04231939e-01 6.35070801e-01 6.24024749e-01 -5.16189575e-01 -3.70860398e-01 -1.56937279e-02 5.49615562e-01 5.73396623e-01 -1.20990664e-01 -1.74617559e-01 -7.26076901e-01 3.50402743e-01 -1.47345507e+00 -1.00520015e+00 3.65741938e-01 1.98364449e+00 1.12022758e+00 1.88409150e-01 2.30574578e-01 -3.82259101e-01 3.68711621e-01 3.17668617e-01 -7.51303673e-01 -3.17668617e-01 -4.71408635e-01 6.32114410e-01 8.88243198e-01 5.14142036e-01 -1.11418974e+00 1.09691811e+00 6.02567577e+00 1.19453394e+00 -1.38735616e+00 5.37531674e-01 1.24638617e+00 -4.90859479e-01 -3.72452438e-01 -3.86351883e-01 -7.15356171e-01 4.52922404e-01 7.90426135e-01 5.04846573e-02 8.56463790e-01 9.82639909e-01 -3.24516952e-01 5.64477324e-01 -1.15476680e+00 9.51682031e-01 -3.79870087e-02 -1.66218233e+00 3.40140313e-01 6.03719801e-02 9.34840858e-01 4.99864876e-01 6.72858953e-01 4.97439563e-01 6.92058861e-01 -1.37997234e+00 1.02163172e+00 3.55523705e-01 1.06061554e+00 -6.33037865e-01 5.55600882e-01 -7.90371373e-02 -7.94313431e-01 -1.03013329e-01 -2.77435511e-01 3.15154880e-01 -5.48869222e-02 5.08876801e-01 -5.10044634e-01 3.62508357e-01 8.63084733e-01 3.03250194e-01 -8.24055552e-01 5.23058712e-01 -2.08081037e-01 8.95770311e-01 -1.66492030e-01 4.07094002e-01 3.90800685e-01 -7.74868503e-02 5.45202911e-01 1.38585865e+00 7.10540637e-02 -2.47906402e-01 -3.00363123e-01 1.11127090e+00 -6.11975312e-01 -1.41536251e-01 -4.36782330e-01 1.72209933e-01 2.84401357e-01 1.33147001e+00 -3.07276696e-01 -2.13505775e-01 -2.23417491e-01 1.24084997e+00 4.17334914e-01 6.04352355e-01 -1.21660352e+00 -1.72188655e-01 1.17605305e+00 1.04924232e-01 2.49024421e-01 9.46047809e-03 -5.96497178e-01 -1.18366933e+00 -3.46865617e-02 -1.32793939e+00 2.40456730e-01 -6.51918650e-01 -1.69301188e+00 1.02258515e+00 -1.59623250e-01 -8.75293016e-01 -1.22889183e-01 -7.43965268e-01 -8.57662380e-01 1.03534722e+00 -1.21256065e+00 -1.44897318e+00 -3.11973661e-01 8.81203651e-01 5.23145497e-01 -5.95999658e-01 8.45028460e-01 4.97159690e-01 -6.96499407e-01 1.28816819e+00 1.48914650e-01 3.54365170e-01 1.05102515e+00 -1.24062026e+00 1.09089327e+00 1.09677458e+00 2.73993671e-01 6.30277038e-01 4.62582886e-01 -3.07975382e-01 -1.05529439e+00 -1.26625204e+00 2.89343864e-01 -5.35971045e-01 7.25901604e-01 -7.34970629e-01 -7.11660624e-01 9.30371165e-01 5.35986900e-01 2.61597127e-01 5.70981324e-01 -5.31676635e-02 -8.72690439e-01 -4.72697496e-01 -1.11032736e+00 9.27257657e-01 1.12701452e+00 -5.13976097e-01 -5.30147403e-02 3.19102943e-01 1.01170015e+00 -5.56267560e-01 -6.43691421e-01 3.03394288e-01 6.15856886e-01 -9.19640601e-01 1.21893406e+00 -9.61909294e-01 4.80352342e-01 -5.08775841e-03 -3.51691902e-01 -1.50456917e+00 -6.04573131e-01 -9.37392056e-01 2.51764923e-01 1.38788807e+00 7.10503161e-01 -9.40466583e-01 6.65613830e-01 7.20145285e-01 -1.57526746e-01 -7.35165596e-01 -9.04524505e-01 -8.55830550e-01 5.71541846e-01 -3.85674715e-01 8.32180619e-01 1.04790437e+00 -7.88161278e-01 1.88846156e-01 -5.57211518e-01 -9.66958776e-02 5.61746240e-01 -4.90079314e-01 9.42236006e-01 -7.52882302e-01 -5.02683461e-01 -8.16875875e-01 -7.48541579e-02 -7.26620972e-01 2.24595100e-01 -9.63387132e-01 -2.16870233e-01 -8.22973728e-01 -1.55742928e-01 -5.38116395e-01 -5.71790516e-01 8.56942177e-01 -5.81209771e-02 6.21816933e-01 5.47772527e-01 1.09196454e-01 -2.39517063e-01 5.18437147e-01 9.20762956e-01 -3.33186150e-01 8.09484422e-02 -2.58906428e-02 -9.55035865e-01 4.90934938e-01 1.07038462e+00 -5.80376387e-01 -4.70919073e-01 -8.20221543e-01 1.76037565e-01 -5.71055174e-01 6.29802644e-01 -1.14500737e+00 4.69659083e-03 -2.30151182e-03 6.54667854e-01 1.02933630e-01 2.31303811e-01 -6.72687232e-01 3.96373451e-01 3.31437021e-01 -4.73364353e-01 5.03585994e-01 6.26014590e-01 1.20431051e-01 -7.79452128e-03 1.07271403e-01 1.09445977e+00 -2.06999928e-02 -3.52641404e-01 4.67571467e-01 1.94633659e-02 4.48842913e-01 6.44442201e-01 2.55339146e-01 -5.49721360e-01 -2.98758417e-01 -4.33311760e-01 -1.32092103e-01 3.21087688e-01 5.13997793e-01 2.36811876e-01 -1.45282316e+00 -9.71667945e-01 5.25392771e-01 -1.34184808e-01 2.95831319e-02 3.37119281e-01 8.14228654e-01 -5.36948025e-01 4.16112319e-02 -4.46100026e-01 -4.26791638e-01 -1.05107653e+00 3.24093431e-01 5.29172540e-01 -5.22594452e-01 -3.03281009e-01 1.42635918e+00 6.72341585e-01 -5.65783918e-01 1.87011644e-01 -1.98947027e-01 4.06192213e-01 -1.98046133e-01 4.02995855e-01 2.33456790e-01 1.91299587e-01 -3.36212784e-01 -3.10695559e-01 2.86827713e-01 -1.66080624e-01 -2.86349714e-01 1.31902599e+00 2.77069449e-01 -1.12468399e-01 9.59933028e-02 1.17775738e+00 8.59092027e-02 -1.59740043e+00 -1.32167146e-01 -5.27562320e-01 -7.72936046e-02 -9.20059904e-02 -1.28590274e+00 -1.45030355e+00 8.52029145e-01 7.15429366e-01 -1.67206660e-01 1.19881713e+00 -2.07382798e-01 6.09976292e-01 8.40941370e-02 7.48889148e-02 -6.15465224e-01 1.24911861e-02 5.79651713e-01 1.36679673e+00 -1.18314338e+00 -1.93108991e-01 1.42222762e-01 -8.75959635e-01 7.47128069e-01 7.48144209e-01 -2.05606148e-01 6.49174869e-01 5.64791620e-01 2.68242627e-01 6.89613223e-02 -7.43834436e-01 3.35206389e-01 4.74496335e-01 4.89061862e-01 2.61446506e-01 5.10037839e-02 2.84427881e-01 7.77807057e-01 -5.47546506e-01 -3.98273528e-01 -7.92237185e-03 3.58268291e-01 3.89898866e-01 -1.10851324e+00 -2.57027179e-01 3.91622275e-01 -7.75250554e-01 -6.18442178e-01 -3.04798543e-01 8.55931222e-01 8.83507654e-02 4.44446772e-01 7.31089041e-02 -4.83227193e-01 2.58014679e-01 2.05990121e-01 4.57417876e-01 -1.36490300e-01 -1.13696039e+00 -1.52485207e-01 -1.81472629e-01 -7.11555362e-01 1.56310126e-01 -3.61315399e-01 -6.15237594e-01 -6.31650865e-01 1.29654750e-01 -1.55231938e-01 7.76516140e-01 7.17121243e-01 5.88721097e-01 7.31644392e-01 3.56737435e-01 -1.02986503e+00 -7.96337545e-01 -1.11916733e+00 -2.08737656e-01 5.81650734e-01 4.07249600e-01 -2.60130584e-01 -4.60777193e-01 2.50806898e-01]
[10.306394577026367, 2.164268970489502]
31a3cacc-32e4-461a-af6a-8066deb2e2aa
deep-reinforcement-learning-for-dexterous
1709.06977
null
http://arxiv.org/abs/1709.06977v1
http://arxiv.org/pdf/1709.06977v1.pdf
Deep Reinforcement Learning for Dexterous Manipulation with Concept Networks
Deep reinforcement learning yields great results for a large array of problems, but models are generally retrained anew for each new problem to be solved. Prior learning and knowledge are difficult to incorporate when training new models, requiring increasingly longer training as problems become more complex. This is especially problematic for problems with sparse rewards. We provide a solution to these problems by introducing Concept Network Reinforcement Learning (CNRL), a framework which allows us to decompose problems using a multi-level hierarchy. Concepts in a concept network are reusable, and flexible enough to encapsulate feature extractors, skills, or other concept networks. With this hierarchical learning approach, deep reinforcement learning can be used to solve complex tasks in a modular way, through problem decomposition. We demonstrate the strength of CNRL by training a model to grasp a rectangular prism and precisely stack it on top of a cube using a gripper on a Kinova JACO arm, simulated in MuJoCo. Our experiments show that our use of hierarchy results in a 45x reduction in environment interactions compared to the state-of-the-art on this task.
['Marcos Campos', 'Victor Shnayder', 'Ruofan Kong', 'Ross Story', 'Matthew Brown', 'Aditya Gudimella', 'Matineh Shaker']
2017-09-20
null
null
null
null
['problem-decomposition']
['miscellaneous']
[-6.37117177e-02 2.08998352e-01 2.17509791e-01 2.39290148e-02 -1.91042349e-01 -7.16487348e-01 1.33095175e-01 2.00635210e-01 -6.43300474e-01 9.30408835e-01 -2.64251858e-01 -6.36397749e-02 -5.66740870e-01 -9.08983886e-01 -9.15704727e-01 -6.11334562e-01 -5.53018510e-01 7.84766614e-01 2.95801282e-01 -6.71810389e-01 3.19835365e-01 6.93038642e-01 -1.94866574e+00 3.46278816e-01 7.21928120e-01 6.61238313e-01 8.07517231e-01 6.04170322e-01 6.87557533e-02 8.90876293e-01 -6.87110960e-01 1.72067598e-01 3.79107654e-01 4.77963947e-02 -1.04278278e+00 7.69828865e-03 7.05653802e-02 -2.65187562e-01 1.71706900e-01 5.77641487e-01 1.65960118e-01 4.81751055e-01 4.56441581e-01 -1.16608775e+00 -2.83709407e-01 8.78135979e-01 -3.56505036e-01 -9.70091522e-02 3.25712293e-01 1.98470443e-01 9.37393129e-01 -5.18885851e-01 3.96136045e-01 1.31714284e+00 6.03679121e-01 7.65219450e-01 -1.32950759e+00 -6.10242128e-01 4.62861508e-01 3.29738781e-02 -8.27710748e-01 1.28091246e-01 3.56162220e-01 -3.11612785e-01 1.30442703e+00 -8.82856250e-02 9.22995865e-01 1.08554494e+00 1.84101954e-01 8.45900774e-01 1.04530978e+00 -3.84876430e-01 5.40216923e-01 2.26946194e-02 -6.82925209e-02 6.59315407e-01 3.35950553e-01 4.53668863e-01 -3.96649659e-01 5.27749024e-02 9.93878663e-01 1.47384822e-01 1.33649647e-01 -8.19072187e-01 -9.75765705e-01 8.27581704e-01 6.95690334e-01 3.92089635e-01 -3.97315741e-01 4.24476862e-01 1.53482959e-01 6.43301308e-01 -1.89182892e-01 1.34844112e+00 -8.39627624e-01 -7.06477240e-02 -5.92145145e-01 7.10931540e-01 1.10476971e+00 8.19377661e-01 7.81786084e-01 9.96369421e-02 2.27180675e-01 7.59386837e-01 2.28371732e-02 3.47399056e-01 4.79266822e-01 -1.29197049e+00 3.61629575e-01 5.53135514e-01 1.90678760e-01 -6.65909171e-01 -8.87925625e-01 -5.40565133e-01 -3.13254803e-01 9.92036939e-01 2.02166095e-01 -4.26122576e-01 -9.60989237e-01 1.59437132e+00 2.50943899e-01 -3.05477083e-01 2.63681114e-01 8.03943872e-01 3.82381678e-01 5.58483481e-01 1.13091702e-02 1.24439709e-01 9.38489616e-01 -1.19307268e+00 -1.46145776e-01 -3.87811273e-01 4.44081068e-01 -2.09017694e-01 1.04322791e+00 9.75387275e-01 -1.26486671e+00 -7.80677974e-01 -1.13331068e+00 2.79062569e-01 -6.46726608e-01 -3.22132587e-01 1.10478604e+00 3.73885810e-01 -1.25382781e+00 1.14061224e+00 -9.69289780e-01 -1.95735529e-01 3.10647607e-01 9.09026384e-01 -3.39905262e-01 -1.54324040e-01 -9.29661691e-01 1.22894621e+00 7.85925806e-01 1.85347686e-03 -1.29288232e+00 -5.49920499e-01 -9.02273178e-01 4.18246865e-01 7.35415518e-01 -5.83469987e-01 1.43849230e+00 -9.14609671e-01 -1.70300484e+00 2.56077468e-01 7.22676277e-01 -4.15051281e-01 1.15729339e-01 -3.80314469e-01 1.09875865e-01 1.56037226e-01 8.27325433e-02 9.08447146e-01 9.30939913e-01 -1.55743301e+00 -8.61815572e-01 -1.96218982e-01 8.73227894e-01 3.11158478e-01 -1.32386103e-01 -2.76450604e-01 8.00008997e-02 -4.42114264e-01 -9.84834358e-02 -1.08035314e+00 -6.11304760e-01 -3.94199431e-01 1.60539255e-01 -2.76721239e-01 6.29183829e-01 -2.16824666e-01 7.57193387e-01 -1.83694470e+00 8.40629101e-01 2.75314838e-01 2.55482882e-01 2.23138601e-01 -3.81313622e-01 6.24228477e-01 -1.24741673e-01 -9.39497873e-02 -2.94276506e-01 -1.23767629e-01 3.95596437e-02 6.16174817e-01 5.36084399e-02 -2.70216197e-01 2.01703444e-01 6.66367471e-01 -1.18762672e+00 -2.48779673e-02 2.42095347e-02 9.83306244e-02 -9.46697116e-01 3.35194021e-01 -6.22609973e-01 2.99127787e-01 -5.20665467e-01 6.36921406e-01 3.25497091e-01 -1.40090808e-01 4.35375571e-01 4.72914696e-01 -1.09200872e-01 -1.27094671e-01 -1.29740179e+00 2.05080771e+00 -7.38300443e-01 1.84143111e-01 3.47500592e-01 -1.21653080e+00 7.84993589e-01 1.54270425e-01 4.99900401e-01 -4.96706486e-01 5.34912832e-02 1.01760261e-01 3.35946500e-01 -6.75142229e-01 5.94540119e-01 -2.48062581e-01 -2.45871201e-01 4.93798107e-01 3.90237689e-01 -5.48141599e-01 6.07447624e-01 -1.68868601e-02 1.24253905e+00 5.97871602e-01 3.28454492e-03 -3.83497238e-01 2.82769918e-01 2.39350591e-02 3.39412928e-01 9.36763942e-01 2.67257988e-01 2.39551499e-01 5.22361636e-01 -6.39825404e-01 -7.95061350e-01 -1.18144917e+00 3.13307762e-01 1.38986301e+00 -1.81335583e-01 -2.48508215e-01 -7.68576503e-01 -6.42064214e-01 3.88883621e-01 5.24255455e-01 -7.67777205e-01 -9.50796604e-02 -7.09989309e-01 -5.21173120e-01 7.37322345e-02 7.04191267e-01 3.61166954e-01 -1.72252619e+00 -1.22224116e+00 5.96167266e-01 2.96302110e-01 -8.11176002e-01 2.54316419e-01 7.95872629e-01 -1.15172994e+00 -1.26954079e+00 -4.26929057e-01 -1.01162553e+00 7.06553578e-01 1.33986935e-01 1.26179361e+00 3.96265715e-01 -5.46002686e-01 7.51716018e-01 -6.11374497e-01 -5.37362337e-01 -2.74930686e-01 1.98671356e-01 2.96749204e-01 -6.18720174e-01 -7.28635341e-02 -9.33297157e-01 -4.02596325e-01 1.93193797e-02 -1.03556633e+00 -1.67395994e-01 8.46759081e-01 9.84635115e-01 6.08019680e-02 3.20808232e-01 6.21943116e-01 -6.02518320e-01 1.08502364e+00 -4.85173941e-01 -8.00783455e-01 1.14355855e-01 -4.51817453e-01 4.78633702e-01 7.16234863e-01 -4.00651902e-01 -8.88430059e-01 2.23441973e-01 1.20202854e-01 -1.91559747e-01 -1.99150816e-01 7.16495454e-01 3.11578453e-01 -1.94350570e-01 7.75612354e-01 -8.03545788e-02 1.99227080e-01 -3.86583805e-01 2.38238335e-01 1.33866355e-01 2.07914919e-01 -1.22642183e+00 6.30822301e-01 -1.50720589e-02 1.86787784e-01 -5.68598747e-01 -7.24707723e-01 -1.89380839e-01 -7.85867810e-01 -3.28256786e-02 6.94929719e-01 -6.48575187e-01 -1.30448639e+00 3.59780312e-01 -8.77807796e-01 -8.89911175e-01 -4.00629550e-01 2.48196632e-01 -6.88268006e-01 -5.35692312e-02 -5.67515135e-01 -6.33892059e-01 5.39418124e-02 -1.14093411e+00 6.58556640e-01 4.08532053e-01 1.45201068e-02 -9.14595425e-01 1.35621727e-01 1.65859193e-01 6.20038331e-01 2.49967933e-01 9.85102654e-01 -3.15473258e-01 -5.86572707e-01 8.69527757e-02 1.01317711e-01 4.51097846e-01 2.63774451e-02 -4.03612912e-01 -6.47129476e-01 -7.34792292e-01 -1.01662271e-01 -1.07485259e+00 7.57031739e-01 6.83643669e-02 1.29133403e+00 -1.85516402e-02 -2.82879144e-01 6.86059520e-02 1.41508901e+00 2.54710734e-01 5.09339154e-01 6.22835159e-01 4.21542108e-01 6.52476013e-01 5.95848799e-01 3.73728961e-01 2.48019397e-01 5.32364309e-01 7.22217798e-01 5.85349761e-02 2.45947778e-01 5.43972477e-02 2.95954496e-01 3.82910073e-01 -5.48142433e-01 1.39516309e-01 -1.03568935e+00 3.50883096e-01 -1.88755524e+00 -8.16643059e-01 5.35257280e-01 1.96054924e+00 8.92145216e-01 2.42695868e-01 2.34308586e-01 2.36363821e-02 4.61077318e-02 -3.61888409e-01 -5.86138666e-01 -5.54769456e-01 6.00021958e-01 6.44406259e-01 1.84898794e-01 6.13787055e-01 -8.68216038e-01 1.22776043e+00 7.06781530e+00 3.90415043e-01 -1.00558877e+00 -2.57805347e-01 -3.14948261e-02 -2.44053558e-01 1.74307793e-01 -1.25299096e-01 -6.57468975e-01 1.06698461e-01 7.81912386e-01 3.28163058e-01 1.04235661e+00 1.02037203e+00 -1.95299178e-01 -4.32528436e-01 -1.50263774e+00 7.99990714e-01 -1.01378582e-01 -1.09423769e+00 -7.90092796e-02 -9.28442851e-02 6.08623922e-01 1.88678168e-02 -4.71928380e-02 1.12992799e+00 8.53049994e-01 -1.34662879e+00 4.23382580e-01 2.67942399e-01 1.79929554e-01 -8.65228653e-01 4.62298453e-01 5.80522478e-01 -9.10225749e-01 -7.23933637e-01 -4.77963954e-01 -5.81291914e-01 -3.99449974e-01 -2.86402106e-02 -8.28514040e-01 5.27624130e-01 9.39949155e-01 3.44153404e-01 -4.64183033e-01 1.03058660e+00 -3.05516869e-01 8.51225853e-02 -1.88690826e-01 -2.77752340e-01 5.00519335e-01 -6.86527193e-02 2.17918694e-01 8.88671875e-01 -1.94805991e-02 2.57220328e-01 5.66008985e-01 7.92365849e-01 1.42870039e-01 -2.51098484e-01 -5.26187956e-01 1.10226922e-01 7.36102685e-02 1.27938259e+00 -8.15743268e-01 -9.94911715e-02 -2.30340034e-01 8.34358335e-01 1.01171362e+00 2.63076514e-01 -5.46231806e-01 -5.10998011e-01 5.09809971e-01 -1.78805798e-01 6.27740502e-01 -4.92052943e-01 1.79704905e-01 -1.06605303e+00 -1.89065877e-02 -1.20745122e+00 2.39108518e-01 -8.78594041e-01 -1.12652159e+00 6.41758442e-01 2.56096750e-01 -1.05495036e+00 -3.17243338e-01 -1.12179995e+00 -2.96260446e-01 5.36667168e-01 -1.36023498e+00 -8.66998434e-01 -2.74871916e-01 6.15004063e-01 4.96789724e-01 -4.44257259e-01 1.26778162e+00 -1.67827472e-01 -1.61736399e-01 8.37509334e-02 -9.22427047e-03 -7.43937418e-02 2.30749503e-01 -1.66326642e+00 1.26836583e-01 1.79433420e-01 -8.49369988e-02 5.82884490e-01 5.88181019e-01 -3.94617409e-01 -1.41776299e+00 -5.27839959e-01 1.33740470e-01 -5.92611730e-01 5.92984915e-01 -6.26921117e-01 -7.63452172e-01 5.86490273e-01 3.03885490e-01 -2.74025232e-01 5.49456120e-01 6.35463357e-01 -2.91745812e-01 -2.68913835e-01 -1.16632414e+00 5.83193660e-01 9.64615762e-01 -3.25277597e-02 -1.07035935e+00 2.57687122e-01 7.49238312e-01 -5.56286991e-01 -8.38709772e-01 2.72033513e-01 5.99765956e-01 -9.48446691e-01 9.56500947e-01 -9.20447826e-01 6.67226255e-01 -1.88299850e-01 9.15464014e-02 -1.84463441e+00 -5.45859575e-01 -5.27255356e-01 -1.95506766e-01 4.26182777e-01 6.03297949e-01 -4.23058063e-01 8.70726049e-01 4.83578801e-01 -1.40122473e-01 -7.45328069e-01 -5.55669844e-01 -9.49442029e-01 4.30543989e-01 -1.78467691e-01 5.32379031e-01 8.61871064e-01 4.93909627e-01 4.04455155e-01 -1.91395134e-01 8.14966485e-02 4.61065263e-01 1.77541018e-01 6.96670055e-01 -1.50781393e+00 -7.38824368e-01 -4.26616281e-01 -1.25607282e-01 -7.85392106e-01 1.46023914e-01 -8.16005707e-01 2.62273014e-01 -1.69396293e+00 2.77262926e-02 -7.49909639e-01 -4.62206185e-01 9.25513506e-01 1.10201567e-01 -3.93491387e-01 4.96921390e-01 -1.29313603e-01 -7.30261743e-01 4.34651434e-01 1.48370910e+00 -1.58996880e-01 -5.10649621e-01 1.00912169e-01 -7.41383553e-01 6.55207217e-01 1.06066501e+00 -5.38679361e-01 -6.05757475e-01 -6.21402979e-01 4.04491156e-01 1.68777898e-01 1.84915289e-01 -1.53105283e+00 2.13593662e-01 -2.27420896e-01 6.43087268e-01 -1.31927133e-01 3.76874626e-01 -1.20693374e+00 -2.58697212e-01 8.27520192e-01 -3.04415822e-01 2.73986667e-01 7.99510837e-01 4.53734070e-01 -4.95947315e-04 -5.52632630e-01 4.17900026e-01 -7.37782598e-01 -6.61953986e-01 -1.17984116e-02 -7.11338401e-01 -1.07161820e-01 1.12160444e+00 -4.38096896e-02 -4.13383916e-02 -1.32362172e-01 -1.24444079e+00 6.87756717e-01 1.29567593e-01 6.77539945e-01 6.04975343e-01 -1.01953149e+00 -4.56583649e-01 2.58791387e-01 -4.34706248e-02 3.20293665e-01 6.54454008e-02 2.33038440e-01 -6.33464158e-01 1.99393466e-01 -8.59062314e-01 -3.33116949e-01 -1.01352191e+00 8.58678401e-01 3.65552425e-01 -6.09718800e-01 -6.24981225e-01 7.21984684e-01 7.53555354e-03 -8.13081503e-01 5.27224302e-01 -6.59928799e-01 -3.89288932e-01 -1.15513414e-01 4.08565342e-01 -1.35417760e-03 -7.16909692e-02 4.43101764e-01 -1.95362940e-02 4.75999594e-01 -4.62026596e-01 -1.16200820e-01 1.78957438e+00 4.45139408e-01 -1.91604033e-01 6.55127242e-02 5.46474814e-01 -3.63204390e-01 -1.48677051e+00 8.27831179e-02 1.01080395e-01 -1.11889057e-01 -2.57734805e-01 -1.17074466e+00 -8.33703756e-01 6.83231950e-01 4.45111960e-01 2.70437539e-01 8.92367244e-01 -1.82362691e-01 2.05643818e-01 1.40820169e+00 8.23324621e-01 -1.54820454e+00 8.18180263e-01 8.98786306e-01 1.28209698e+00 -1.20606923e+00 1.47402808e-01 4.37005945e-02 -5.75196922e-01 1.41895282e+00 9.80923772e-01 -4.92506623e-01 5.91782928e-01 3.22542071e-01 -2.83534080e-01 -2.86294252e-01 -7.35507905e-01 -2.19415784e-01 -2.39589512e-01 8.55853796e-01 2.95088673e-03 3.68345529e-02 1.48046032e-01 5.66518307e-01 -2.25283697e-01 9.70581248e-02 7.11978018e-01 1.58635199e+00 -7.67134607e-01 -1.37721395e+00 -3.23273271e-01 2.24453107e-01 -1.69885769e-01 1.58945285e-02 -5.72176054e-02 1.07361925e+00 2.25728720e-01 7.35066712e-01 -7.39832595e-02 -3.74877661e-01 5.80057800e-01 -1.42266760e-02 1.09971428e+00 -1.18175828e+00 -8.89734209e-01 -3.89128715e-01 1.94822047e-02 -6.35677814e-01 -3.96609694e-01 -5.83999276e-01 -1.46871877e+00 -1.13622071e-02 6.92017525e-02 3.75176877e-01 6.76571667e-01 9.42540586e-01 2.11965933e-01 9.13438261e-01 5.30591071e-01 -1.33463490e+00 -7.42053807e-01 -8.04864526e-01 -6.45081818e-01 1.13107443e-01 3.29849541e-01 -1.02081823e+00 -1.58610418e-01 -1.99825376e-01]
[4.068114280700684, 1.4252097606658936]
f5e257cc-7843-4bb1-a9da-6f908f5de24c
gaussian-hermite-moment-invariants-of-general
2201.00877
null
https://arxiv.org/abs/2201.00877v2
https://arxiv.org/pdf/2201.00877v2.pdf
Gaussian-Hermite Moment Invariants of General Multi-Channel Functions
With the development of data acquisition technology, large amounts of multi-channel data are collected and widely used in many fields. Most of them, such as RGB images and vector fields, can be expressed as different types of multi-channel functions. Feature extraction of multi-channel data for identifying interest patterns is a critical but challenging task. This paper focuses on constructing moment-based features of general multi-channel functions. Specifically, we define two transform models, rotation-affine transform and total rotation transform, to describe real deformations of multi-channel data. Then, we design a structural framework to generate Gaussian-Hermite moment invariants for these two transform models systematically. It is the first time that a unified framework has been proposed in the literature to construct orthogonal moment invariants of general multi-channel functions. Given a specific type of multi-channel data, we demonstrate how to utilize the new method to derive all possible invariants and eliminate dependences among them. We obtain independent sets of invariants with low orders and low degrees for RGB images, 2D vector fields and color volume data. Based on synthetic and real multi-channel data, we conduct extensive experiments to evaluate the stability and discriminability of these invariants and their robustness to noise. The results show that new moment invariants significantly outperform previous moment invariants of multi-channel data in RGB image classification and vortex detection in 2D vector fields.
['Guoying Zhao', 'Hua Li', 'Hanlin Mo']
2022-01-03
null
null
null
null
['template-matching']
['computer-vision']
[ 3.86295766e-02 -8.76409292e-01 2.56147236e-02 -1.51236281e-01 -4.88981158e-01 -8.08606207e-01 3.24338078e-01 -4.31360416e-02 -2.90330797e-01 3.94983798e-01 -3.08916986e-01 -3.14429142e-02 -5.20011127e-01 -6.93148553e-01 -3.75239968e-01 -7.20846474e-01 -3.01064134e-01 3.78220640e-02 2.58048952e-01 -2.63119459e-01 4.63641018e-01 1.03058267e+00 -1.63967025e+00 -1.30793437e-01 5.22298157e-01 1.21915817e+00 -1.71772078e-01 7.50064671e-01 8.03455412e-02 1.47482082e-01 -3.83063585e-01 -1.75680697e-01 5.19307613e-01 -2.84248203e-01 -8.11740279e-01 3.92255634e-01 2.50751883e-01 -1.60870925e-01 -2.49505669e-01 9.86337423e-01 4.50555503e-01 1.80993184e-01 7.45960772e-01 -1.38954389e+00 -5.27087569e-01 -2.17929989e-01 -6.69380546e-01 8.57730135e-02 3.55730027e-01 -3.62130394e-03 9.11844015e-01 -7.69984543e-01 6.99547648e-01 1.10476780e+00 4.89293098e-01 3.21836472e-02 -1.17309201e+00 -4.48086172e-01 -4.69693601e-01 2.97259480e-01 -1.37015498e+00 1.12083331e-01 1.31549931e+00 -5.17107606e-01 5.12681842e-01 5.48444629e-01 7.56199300e-01 5.29455185e-01 3.94374937e-01 4.65787590e-01 1.53133655e+00 -3.45923156e-01 -3.18048209e-01 3.58120655e-03 2.04496503e-01 8.55299830e-01 3.01830709e-01 -2.11923942e-01 -2.63585478e-01 -2.04215765e-01 1.09926891e+00 1.52058899e-01 -3.31580251e-01 -6.32512867e-01 -1.45514846e+00 7.50366330e-01 4.63540629e-02 5.53485215e-01 -3.91222179e-01 -1.56789243e-01 2.75581270e-01 1.62874028e-01 1.10108815e-01 3.74779463e-01 -3.35075736e-01 -2.30982408e-01 -5.28238475e-01 2.06534177e-01 5.82314610e-01 7.66886175e-01 1.13219106e+00 -6.10769019e-02 4.66769412e-02 6.72558725e-01 1.37537941e-01 1.13719952e+00 4.94302094e-01 -7.71515310e-01 9.77322236e-02 6.33113563e-01 -8.89973417e-02 -1.45449972e+00 -7.53948390e-01 -1.35587886e-01 -9.84753728e-01 -7.11107254e-02 4.61614102e-01 1.07507214e-01 -4.94176358e-01 1.16828394e+00 3.66911232e-01 6.70446828e-02 -3.24221104e-02 9.23918426e-01 6.44967258e-01 4.68410254e-01 -5.18205583e-01 -3.82577688e-01 1.15182376e+00 -2.65866131e-01 -5.72776318e-01 5.19592285e-01 2.02955231e-01 -1.12495935e+00 8.77811432e-01 3.84566098e-01 -6.77665472e-01 -6.18215084e-01 -8.92567337e-01 1.68325633e-01 -3.40231150e-01 2.14681193e-01 9.41051602e-01 5.34765661e-01 -5.35496950e-01 5.97346783e-01 -7.14698493e-01 -1.16444118e-01 -1.48143232e-01 2.51147628e-01 -7.62489021e-01 2.21330792e-01 -8.46937239e-01 6.31404161e-01 3.01656090e-02 2.90615171e-01 -2.71310151e-01 -3.74780208e-01 -6.63892210e-01 -4.47829336e-01 4.48646843e-02 -1.46950677e-01 6.39822900e-01 -5.93634248e-01 -1.38145542e+00 6.43026233e-01 -1.08113214e-02 3.81773114e-01 1.02514923e-01 2.58047163e-01 -5.93989313e-01 4.63640332e-01 2.37513557e-02 -9.16386992e-02 1.01908231e+00 -1.31079590e+00 -3.67784500e-01 -4.32286382e-01 1.32932132e-02 -2.06645653e-01 -3.78982753e-01 -4.01077904e-02 -2.95943499e-01 -4.62612480e-01 5.91318250e-01 -1.03367209e+00 -1.01988763e-01 -2.22276852e-01 -5.34369648e-01 1.59381047e-01 1.03280985e+00 -5.21442771e-01 9.73289549e-01 -2.03408027e+00 2.53352195e-01 7.09115028e-01 1.80040032e-01 7.53220394e-02 -2.06139147e-01 2.57704675e-01 -7.26763606e-02 -4.85977717e-02 -1.61594659e-01 2.83492863e-01 -2.23432317e-01 3.20076644e-01 -1.93130419e-01 9.15410459e-01 2.61551619e-01 6.59289241e-01 -7.63645470e-01 -6.17091298e-01 6.19170725e-01 6.66291058e-01 -3.73850822e-01 -1.02947615e-02 5.07051826e-01 7.50554621e-01 -7.67887712e-01 1.02278352e+00 9.93770182e-01 -1.03673443e-01 -2.53896832e-01 -8.72265041e-01 -3.08809072e-01 -6.80343330e-01 -1.48909056e+00 1.32692242e+00 -4.24058825e-01 5.94503641e-01 -9.29095596e-02 -1.19005477e+00 1.09693432e+00 9.11087468e-02 1.23136640e+00 -5.57328939e-01 2.47783348e-01 3.82267803e-01 -5.69078550e-02 -7.44471133e-01 6.03641689e-01 -9.54473242e-02 -2.09490120e-01 2.21781611e-01 1.29954413e-01 -3.71064276e-01 3.16153079e-01 -2.80620128e-01 8.48410070e-01 -1.71677724e-01 1.62054896e-01 -5.45801483e-02 9.52518106e-01 -2.43360728e-01 4.88720655e-01 3.59578133e-01 -2.31839210e-01 5.23179948e-01 5.36656976e-01 -5.48251629e-01 -1.05238771e+00 -9.44811761e-01 -5.40648758e-01 4.13470805e-01 4.33776230e-01 -3.63577992e-01 -1.01072870e-01 -3.14004213e-01 2.80735850e-01 -2.50318706e-01 -3.83242190e-01 1.25545859e-01 -5.58885396e-01 -1.05796051e+00 5.27720928e-01 1.62888557e-01 5.66870451e-01 -5.54829240e-01 -9.46639538e-01 1.02391995e-01 -1.78340729e-02 -1.29921877e+00 -5.53930998e-02 -3.94482426e-02 -9.22293484e-01 -1.53098738e+00 -6.95759475e-01 -3.76775146e-01 6.72940135e-01 4.09484178e-01 6.94713295e-01 -5.36526851e-02 -7.37317801e-01 8.17197621e-01 -4.58509535e-01 6.83357427e-03 -7.94711709e-02 -3.61240059e-01 5.43868206e-02 5.67241073e-01 -1.61152303e-01 -6.28409863e-01 -4.53446597e-01 4.87262189e-01 -1.23995936e+00 -1.95484594e-01 7.20298648e-01 7.96512783e-01 8.45248699e-01 1.09838666e-02 -1.24928206e-02 -2.53273815e-01 6.62481487e-01 -7.74570405e-02 -7.66566515e-01 3.40659857e-01 -3.02508980e-01 3.74450505e-01 6.72017395e-01 -4.83764708e-01 -5.72609961e-01 2.78539926e-01 8.39001313e-02 -6.31865799e-01 -4.23604734e-02 4.53212023e-01 6.99577555e-02 -8.94343436e-01 3.25578213e-01 3.76347870e-01 -1.78770889e-02 -4.63279575e-01 3.19075078e-01 6.43592715e-01 4.93313938e-01 -8.28346670e-01 9.76402044e-01 7.72818744e-01 8.38555217e-01 -1.45725024e+00 -1.03061356e-01 -8.02833200e-01 -8.84526074e-01 -4.89824504e-01 7.67555118e-01 -3.11308205e-01 -1.06331015e+00 7.91284800e-01 -1.26694739e+00 2.92669952e-01 1.42287418e-01 7.36225903e-01 -4.76033270e-01 6.46251440e-01 -3.64257991e-01 -6.62737906e-01 -1.16143182e-01 -1.56219792e+00 1.22565103e+00 2.72780269e-01 4.19431299e-01 -1.07449043e+00 1.90400273e-01 -3.56166027e-02 4.36498344e-01 7.39043415e-01 7.66777694e-01 -4.49566431e-02 -5.68834603e-01 -3.79917145e-01 -1.63585842e-01 4.17551160e-01 2.58161038e-01 4.53641504e-01 -5.70569634e-01 -1.09695770e-01 1.19862389e-02 -1.37577966e-01 5.67091823e-01 2.82286167e-01 1.28903937e+00 3.67109962e-02 -1.50152981e-01 1.12263048e+00 1.62547445e+00 1.74029440e-01 4.32129234e-01 2.62584567e-01 8.72970521e-01 3.37537795e-01 4.86301482e-01 7.25312591e-01 2.44145840e-01 6.69363379e-01 3.73558164e-01 -9.15747657e-02 4.73068446e-01 2.98256010e-01 7.55808800e-02 1.07139528e+00 -7.86098063e-01 3.70249778e-01 -7.85759687e-01 2.40756452e-01 -1.59269071e+00 -9.05570567e-01 -2.63493210e-01 2.17561078e+00 4.70417082e-01 -3.22247893e-01 7.60005414e-02 2.93543547e-01 7.09007442e-01 2.16827214e-01 -4.50925231e-01 -1.91679403e-01 -4.56575155e-01 5.14250040e-01 8.57739210e-01 3.06967556e-01 -1.41529953e+00 5.93022704e-01 6.27249718e+00 5.24498999e-01 -1.66153431e+00 -1.52030915e-01 1.43224448e-01 3.68530780e-01 -3.45578164e-01 1.46734893e-01 -5.85681140e-01 1.70164764e-01 4.79564220e-01 -1.81528285e-01 3.52228284e-01 6.85158074e-01 -1.08402044e-01 -1.39284089e-01 -4.97094184e-01 1.47760856e+00 8.00180286e-02 -1.09338725e+00 1.35323346e-01 9.82722640e-02 6.43745005e-01 -2.91346699e-01 2.18755230e-02 -2.74017423e-01 -4.96893644e-01 -6.55664265e-01 3.99763525e-01 9.28019166e-01 7.15638399e-01 -6.88260436e-01 4.58797336e-01 -9.25433040e-02 -1.48032188e+00 3.98375615e-02 -4.10460979e-01 7.40543008e-02 1.26876399e-01 6.27868235e-01 -3.14641416e-01 9.23699737e-01 3.96689534e-01 9.06654060e-01 -7.35652506e-01 1.08123517e+00 1.48426548e-01 2.55768821e-02 -3.27663809e-01 -1.72648221e-01 9.10289809e-02 -5.69619477e-01 5.77445090e-01 9.80183542e-01 5.24311662e-01 -2.29493584e-02 1.99596152e-01 6.07727051e-01 4.10886943e-01 3.77652436e-01 -6.28488421e-01 -1.84096806e-02 5.65763339e-02 1.76566792e+00 -9.67547178e-01 -5.62683232e-02 -6.23100519e-01 8.70726645e-01 -3.09909552e-01 5.15214503e-01 -7.81683803e-01 -7.23189592e-01 7.60909200e-01 -1.67128831e-01 1.60834074e-01 -7.62039661e-01 -5.70167042e-02 -1.39791024e+00 1.70198590e-01 -6.15407646e-01 5.01405410e-02 -4.65191185e-01 -1.10496318e+00 5.10750592e-01 3.95483345e-01 -1.57301927e+00 -2.27869347e-01 -1.17058480e+00 -3.00182730e-01 8.17482948e-01 -1.60817063e+00 -1.25681710e+00 -4.98524219e-01 1.09916925e+00 -2.16163531e-01 -1.36543974e-01 7.98655093e-01 2.57415265e-01 -2.78343111e-01 2.81448007e-01 3.34070593e-01 2.40201473e-01 5.77655673e-01 -9.70451653e-01 -2.91690171e-01 8.15776467e-01 5.90319633e-02 7.03579962e-01 5.14524817e-01 -3.71986777e-01 -2.17473102e+00 -5.52695215e-01 1.89283967e-01 -1.10154726e-01 6.20515227e-01 -5.75344078e-02 -6.25720739e-01 4.50817496e-01 -3.38643819e-01 5.84147274e-01 5.95859349e-01 -2.67842293e-01 -1.92236766e-01 -4.51512486e-01 -1.02205944e+00 4.04702872e-02 4.91080433e-01 -6.35033727e-01 -2.24047855e-01 2.68819749e-01 1.91305846e-01 -4.47420478e-01 -1.38219023e+00 6.00491166e-01 8.92868340e-01 -1.19402587e+00 1.24016178e+00 -2.75958657e-01 1.71249032e-01 -5.12743592e-01 -3.02231044e-01 -1.20098424e+00 -1.39417648e-01 -7.10069656e-01 1.58919513e-01 8.45143616e-01 -1.41619772e-01 -7.85702407e-01 2.96499133e-01 3.15474451e-01 -1.99688114e-02 -7.46409893e-01 -1.05395019e+00 -8.41464281e-01 -8.71174932e-02 -5.28033555e-01 5.67867100e-01 9.61976409e-01 -1.20292649e-01 -7.56183267e-02 -4.03447479e-01 1.90852940e-01 8.31950724e-01 5.90323329e-01 9.48712826e-01 -1.38390386e+00 -1.53094843e-01 -4.00676876e-01 -1.12788975e+00 -6.73287690e-01 1.27641305e-01 -7.70680726e-01 -4.35237706e-01 -1.14629579e+00 2.37117764e-02 -4.92546558e-01 -2.94476897e-01 3.70747209e-01 8.40622932e-02 3.88398796e-01 1.91637039e-01 3.95724714e-01 -2.32983321e-01 4.39761072e-01 1.63689101e+00 5.12039550e-02 -2.64096469e-01 -1.08284235e-01 -1.31212831e-01 6.92194700e-01 4.99782592e-01 3.42688970e-02 -1.23117387e-01 1.54325450e-02 2.18422994e-01 1.80190712e-01 6.32060170e-01 -1.22705579e+00 -4.14899662e-02 -4.60356861e-01 4.34098780e-01 -5.83584785e-01 2.92640001e-01 -8.55029881e-01 2.62028962e-01 2.67549813e-01 2.22045794e-01 3.79563272e-01 1.02320857e-01 1.23124003e-01 -5.06034374e-01 9.96668711e-02 7.82353699e-01 -1.20737933e-01 -1.06961036e+00 4.14267898e-01 1.07148765e-02 -1.17864892e-01 1.09923363e+00 -1.27990142e-01 -2.77531803e-01 -2.11745247e-01 -4.66759592e-01 -3.69269401e-01 4.38029110e-01 3.28259259e-01 9.84642446e-01 -1.58081770e+00 -3.78534168e-01 7.58601844e-01 1.18476771e-01 -2.31085911e-01 2.41607308e-01 1.25477982e+00 -8.71734917e-01 4.96818066e-01 -7.87789702e-01 -8.13080251e-01 -1.18902671e+00 3.31715375e-01 3.27975571e-01 -1.11447684e-01 -2.93711483e-01 2.94969231e-01 -2.90160626e-01 -2.11248487e-01 -3.57909262e-01 -6.94415152e-01 -2.26854607e-01 1.66806072e-01 2.62384802e-01 5.27305603e-01 1.09809123e-01 -1.28386724e+00 -4.87893283e-01 1.47899139e+00 5.66093683e-01 -1.56859815e-01 1.40239930e+00 4.80072647e-02 -5.04917979e-01 4.53148603e-01 1.78292024e+00 1.17664129e-01 -8.23448181e-01 9.64785591e-02 -3.05596769e-01 -7.86360502e-01 -4.97194268e-02 -5.64638786e-02 -1.27591586e+00 9.24021244e-01 6.12449825e-01 2.95389950e-01 1.40108001e+00 -5.69859333e-02 7.92673051e-01 2.82960922e-01 4.10126507e-01 -9.59492385e-01 5.75500205e-02 4.72737819e-01 8.23106110e-01 -1.16473877e+00 -1.72498403e-03 -3.54924202e-01 -3.21838588e-01 1.70268953e+00 3.29847962e-01 -1.55292809e-01 8.76116276e-01 3.94311436e-02 2.73398752e-03 -2.02130899e-01 9.26696658e-02 -4.38603252e-01 6.21772587e-01 4.86230344e-01 4.12765056e-01 1.23091795e-01 -5.69484055e-01 3.02124918e-01 -1.41534032e-02 -1.56439424e-01 5.15396297e-01 9.86685216e-01 -3.01181525e-01 -1.32734859e+00 -8.14798534e-01 4.02360618e-01 -3.81052077e-01 3.34501177e-01 7.00800866e-02 1.01784587e+00 1.66391522e-01 6.35810435e-01 -8.25696215e-02 -8.00987661e-01 5.02626777e-01 -9.10064951e-02 7.67968535e-01 1.08644940e-01 -5.40691726e-02 2.09128410e-01 -4.71014529e-01 -7.18806863e-01 -9.82654274e-01 -6.17104888e-01 -1.15033412e+00 -1.11823380e-01 -2.19481066e-01 1.11190584e-02 9.83959079e-01 7.78413653e-01 1.74495935e-01 2.26561666e-01 9.67400372e-01 -1.05008781e+00 -4.22153443e-01 -6.24189258e-01 -1.01691628e+00 7.86622465e-01 5.95992208e-01 -1.00490546e+00 -3.66760045e-01 -3.45606431e-02]
[10.18285083770752, -1.2646727561950684]
6d105c1d-530d-440b-9604-fc15f047499d
did-the-models-understand-documents
2306.11386
null
https://arxiv.org/abs/2306.11386v1
https://arxiv.org/pdf/2306.11386v1.pdf
Did the Models Understand Documents? Benchmarking Models for Language Understanding in Document-Level Relation Extraction
Document-level relation extraction (DocRE) attracts more research interest recently. While models achieve consistent performance gains in DocRE, their underlying decision rules are still understudied: Do they make the right predictions according to rationales? In this paper, we take the first step toward answering this question and then introduce a new perspective on comprehensively evaluating a model. Specifically, we first conduct annotations to provide the rationales considered by humans in DocRE. Then, we conduct investigations and reveal the fact that: In contrast to humans, the representative state-of-the-art (SOTA) models in DocRE exhibit different decision rules. Through our proposed RE-specific attacks, we next demonstrate that the significant discrepancy in decision rules between models and humans severely damages the robustness of models and renders them inapplicable to real-world RE scenarios. After that, we introduce mean average precision (MAP) to evaluate the understanding and reasoning capabilities of models. According to the extensive experimental results, we finally appeal to future work to consider evaluating both performance and the understanding ability of models for the development of their applications. We make our annotations and code publicly available.
['Xiangdong Zhou', 'Bingsheng Chen', 'Haotian Chen']
2023-06-20
null
null
null
null
['benchmarking', 'document-level-relation-extraction', 'relation-extraction', 'benchmarking']
['miscellaneous', 'natural-language-processing', 'natural-language-processing', 'robots']
[ 1.78026244e-01 5.69788277e-01 -3.04200649e-01 -2.86464840e-01 -5.81162274e-01 -7.95164645e-01 7.88729131e-01 2.48170286e-01 -1.06185064e-01 5.76624632e-01 5.90751655e-02 -5.32146811e-01 -2.40826532e-01 -7.02488780e-01 -4.75753725e-01 3.25705372e-02 1.72536999e-01 4.75505263e-01 3.26994151e-01 -3.13500613e-01 3.13998729e-01 3.65222096e-01 -1.41554761e+00 4.97744501e-01 8.56665373e-01 8.45666111e-01 -5.59147298e-01 4.69657868e-01 3.32264930e-01 1.19085228e+00 -6.93471253e-01 -1.18036222e+00 1.21069670e-01 2.25485656e-02 -1.10686791e+00 -2.22267598e-01 4.22913700e-01 -2.36740932e-01 -1.60469353e-01 1.14169395e+00 2.41811752e-01 -3.17233741e-01 9.32568550e-01 -1.37975824e+00 -6.68564975e-01 1.12492251e+00 -1.92839652e-01 2.73191661e-01 5.12717366e-01 3.47041176e-03 1.34816313e+00 -8.76377881e-01 6.63658917e-01 9.26074922e-01 6.05092406e-01 6.46682441e-01 -8.39838147e-01 -6.48102045e-01 4.36619401e-01 4.35089439e-01 -1.48258519e+00 -6.98558986e-01 5.43696165e-01 -5.37915051e-01 1.16191614e+00 6.30652130e-01 2.11175352e-01 1.35095608e+00 1.53894663e-01 6.51312947e-01 1.05634189e+00 -5.77740371e-01 1.78344265e-01 4.08018440e-01 6.92871392e-01 6.84921265e-01 8.24842811e-01 3.28635462e-02 -5.76622486e-01 -2.58049190e-01 1.63073868e-01 -5.35234511e-01 -2.75746465e-01 -1.66213959e-01 -7.36277044e-01 5.26724160e-01 6.23811558e-02 3.08743864e-01 -3.18235815e-01 -1.45042315e-01 2.63910919e-01 1.96041279e-02 1.86195523e-01 7.12385178e-01 -6.65377021e-01 -2.43411362e-02 -8.14122915e-01 3.85551006e-01 1.04176342e+00 1.01831794e+00 1.89833328e-01 -3.37575287e-01 -2.35860363e-01 4.12912279e-01 3.67311448e-01 1.48534447e-01 3.07474971e-01 -6.54436529e-01 5.44995010e-01 7.14385211e-01 4.09917772e-01 -1.41448343e+00 -7.98893273e-02 -9.53006566e-01 -6.68404341e-01 -2.14603841e-01 2.52155811e-01 -6.74698725e-02 -3.83596122e-01 1.39374197e+00 -7.51768425e-02 2.89831823e-03 1.61004543e-01 5.70919693e-01 6.96425498e-01 -7.90679071e-04 3.33665043e-01 -1.90404564e-01 1.56213212e+00 -7.48641312e-01 -7.67052233e-01 -4.05856192e-01 7.25326002e-01 -5.58000803e-01 9.23146307e-01 6.48579121e-01 -8.90221775e-01 -3.14350337e-01 -1.33013034e+00 8.18109587e-02 -3.81248683e-01 3.84082824e-01 6.73504353e-01 9.91889238e-01 -5.63781738e-01 4.34956431e-01 -6.56790257e-01 -4.20297027e-01 3.00597876e-01 1.78129151e-01 -1.95655853e-01 6.09529614e-02 -1.52270329e+00 1.22161436e+00 3.65600497e-01 1.48071051e-01 -6.72840774e-01 -5.01695037e-01 -5.99798799e-01 2.66858667e-01 8.70096803e-01 -6.12993658e-01 1.41282797e+00 -2.49221981e-01 -1.20119882e+00 8.83350313e-01 -2.13206097e-01 -8.36341441e-01 8.25207233e-01 -6.27284646e-01 -7.49720395e-01 -1.02896608e-01 -1.36613458e-01 6.08749948e-02 3.42041463e-01 -1.47664297e+00 -6.22430444e-01 -1.60058215e-01 4.55809444e-01 -2.66932428e-01 -3.60440254e-01 2.46713147e-01 -3.34465802e-01 -5.09534180e-01 -6.42673224e-02 -7.75412858e-01 -9.95879248e-02 -6.66320264e-01 -9.14190710e-01 -2.24682540e-01 3.54416162e-01 -5.82815588e-01 2.20361280e+00 -1.80639255e+00 -2.16064975e-01 2.71101952e-01 5.32847226e-01 5.59877336e-01 2.14015797e-01 6.82574093e-01 -1.93121761e-01 7.81892240e-01 -1.08143277e-01 -8.72821063e-02 2.33650580e-01 -5.32360636e-02 -6.55980647e-01 7.71078700e-03 1.34233698e-01 1.11171532e+00 -7.15581834e-01 -4.56816554e-01 -1.31641760e-01 1.54725581e-01 -5.19463718e-01 1.34244546e-01 -2.47889757e-01 -5.53816445e-02 -6.60915852e-01 6.39415860e-01 4.77925628e-01 -4.98061448e-01 6.59181893e-01 -4.47009951e-01 2.13142663e-01 6.27333403e-01 -1.08654547e+00 7.49882579e-01 -2.32396513e-01 3.05477262e-01 -5.58404565e-01 -6.75015152e-01 8.39794338e-01 2.69543529e-01 -9.78658348e-02 -3.70228052e-01 5.15672192e-02 2.05293655e-01 1.66783497e-01 -5.42901874e-01 7.61684954e-01 1.12255163e-01 -2.47238904e-01 3.90002072e-01 -3.64744782e-01 2.66962722e-02 2.77920514e-01 3.72864485e-01 1.15927279e+00 -3.08677889e-02 9.70554471e-01 -5.08020297e-02 7.32983828e-01 1.60188794e-01 6.12826645e-01 8.93750191e-01 -1.10210143e-01 3.55142772e-01 6.79359019e-01 -4.20202136e-01 -5.63630402e-01 -7.52105176e-01 2.03749239e-02 6.46634042e-01 1.26103953e-01 -1.11152112e+00 -9.80781496e-01 -1.32570207e+00 -1.98287398e-01 1.31205630e+00 -5.51059425e-01 -1.75017744e-01 -4.30440933e-01 -8.63499045e-01 1.01912904e+00 6.66120291e-01 6.10356212e-01 -6.49389386e-01 -6.87849343e-01 1.32147551e-01 -4.34828997e-01 -1.53218055e+00 1.57764740e-02 -1.28393188e-01 -5.91644943e-01 -1.63602960e+00 1.06851511e-01 -3.15390117e-02 4.60510671e-01 -1.17405467e-02 1.38921976e+00 4.57852870e-01 1.72859848e-01 2.58867234e-01 -5.59048355e-01 -3.80619526e-01 -6.04817629e-01 3.76455843e-01 1.69771761e-02 -3.73028576e-01 9.68695760e-01 -5.19854009e-01 -3.12545985e-01 5.91960013e-01 -7.60962486e-01 4.74569276e-02 5.87874293e-01 3.27983558e-01 1.44782603e-01 1.85497612e-01 4.52308536e-01 -1.52231777e+00 9.46136713e-01 -2.70488560e-01 -4.83972281e-01 8.71512890e-01 -1.22188413e+00 8.11932087e-02 5.03855705e-01 -3.60736325e-02 -9.37754452e-01 -2.90621251e-01 1.57479886e-02 2.62474686e-01 -2.28435293e-01 6.43306136e-01 -3.75806004e-01 3.76722932e-01 8.38504255e-01 -1.54537380e-01 -8.01361680e-01 -5.14045179e-01 3.15928340e-01 8.60408545e-01 5.73395014e-01 -8.57165933e-01 9.91587996e-01 5.05508855e-02 -3.73288155e-01 -2.74425358e-01 -1.25264382e+00 -8.31389278e-02 -4.61294651e-01 1.19911388e-01 6.05752170e-01 -8.79621148e-01 -7.72107363e-01 2.00034514e-01 -1.34061265e+00 3.92792048e-03 5.73018678e-02 8.53672177e-02 -2.93829888e-01 4.96063620e-01 -5.85441411e-01 -8.64744782e-01 -4.58777010e-01 -9.60633934e-01 6.78996503e-01 -6.61073029e-02 -7.00015843e-01 -7.23410428e-01 -6.52147308e-02 7.33111680e-01 1.26194149e-01 1.41148670e-02 1.28824365e+00 -1.20372319e+00 -3.79258156e-01 -3.24302971e-01 -2.76325464e-01 2.22092822e-01 -7.72538111e-02 2.34797433e-01 -1.10142529e+00 1.78047404e-01 -1.84398755e-01 2.40752324e-02 4.74795252e-01 -3.08529258e-01 1.05764401e+00 -4.18314397e-01 -4.97954220e-01 -6.59256522e-03 1.11354339e+00 1.06118664e-01 7.74873614e-01 3.66547227e-01 3.72042835e-01 6.83470249e-01 8.49300563e-01 2.79224515e-01 6.67770684e-01 8.78969848e-01 3.83942217e-01 2.99018025e-01 6.77999631e-02 -6.79845512e-01 1.32642254e-01 6.54672980e-01 -4.34145212e-01 -6.36637866e-01 -1.10963714e+00 3.71922523e-01 -1.93197453e+00 -7.79950619e-01 -2.15898141e-01 1.99609780e+00 7.29237378e-01 5.10814250e-01 -1.77366450e-01 4.10584539e-01 6.83783174e-01 1.26350941e-02 -1.32008776e-01 -3.45173240e-01 5.21318428e-02 6.96783736e-02 2.50257641e-01 4.36126918e-01 -1.04358351e+00 1.15886867e+00 6.89097023e+00 8.30098093e-01 -7.25066602e-01 -1.84567180e-02 5.88090956e-01 4.07696843e-01 -2.94530600e-01 2.67019480e-01 -1.03448367e+00 1.19954556e-01 8.49109888e-01 -2.28768975e-01 1.41276404e-01 9.19144809e-01 1.28782690e-02 1.16246894e-01 -1.44775701e+00 6.43403530e-01 -1.57373771e-02 -1.15863609e+00 2.37006381e-01 1.07570060e-01 4.23654258e-01 -4.91052181e-01 -1.67511016e-01 4.64825243e-01 6.98989332e-01 -1.21881652e+00 8.13703120e-01 3.70971233e-01 5.00752091e-01 -4.84171629e-01 1.00489724e+00 5.19086897e-01 -9.37116921e-01 -1.71151295e-01 -1.95033357e-01 -2.55320191e-01 -8.47421139e-02 7.62006283e-01 -1.02916515e+00 9.07683253e-01 3.37975472e-01 5.37632048e-01 -9.79061723e-01 5.57389498e-01 -9.15667176e-01 6.79972410e-01 4.83747711e-03 3.05544436e-02 -3.43308866e-01 4.73733693e-01 3.26887518e-01 1.18249786e+00 1.83810964e-01 3.03818196e-01 -9.98458564e-02 8.52199256e-01 -1.56126231e-01 -7.85242543e-02 -3.25526625e-01 -2.38648981e-01 7.92336226e-01 1.06365049e+00 -4.88892287e-01 -4.77140307e-01 -1.72220275e-01 6.58699751e-01 5.15099883e-01 3.33893657e-01 -7.12982476e-01 6.08083680e-02 4.03741390e-01 3.08050036e-01 1.33857606e-02 -7.19063133e-02 -4.44651097e-01 -1.53131628e+00 2.13049576e-01 -1.22960699e+00 4.44538742e-01 -7.42077708e-01 -1.13000488e+00 8.80562723e-01 2.43476748e-01 -1.11404538e+00 -3.57643574e-01 -4.98197496e-01 -3.48965585e-01 4.35171932e-01 -1.53801191e+00 -1.09429598e+00 -1.44542888e-01 -1.04261152e-01 3.98302972e-01 -5.56079336e-02 9.23117340e-01 2.32233912e-01 -8.34636390e-01 8.90987158e-01 -6.85607433e-01 2.90924609e-01 5.93984127e-01 -1.05581439e+00 7.16610134e-01 1.17544007e+00 3.32738429e-01 1.29710245e+00 1.09440005e+00 -7.95755148e-01 -9.36640680e-01 -7.72885442e-01 1.43990564e+00 -8.90073657e-01 6.87339067e-01 -2.11194262e-01 -7.42420971e-01 9.26310599e-01 3.75172272e-02 -1.35060921e-01 6.51450396e-01 3.74734342e-01 -6.89639509e-01 1.64869323e-01 -1.05345082e+00 6.82244718e-01 1.33433127e+00 -4.80278373e-01 -9.93821084e-01 4.30447124e-02 4.07533526e-01 -2.57105321e-01 -8.12501073e-01 7.50948250e-01 7.23464549e-01 -1.31416678e+00 8.00116241e-01 -1.00674474e+00 6.86961651e-01 -3.39888453e-01 -3.20577800e-01 -1.04230452e+00 -1.23464517e-01 -4.47942942e-01 -3.81251007e-01 1.42266262e+00 7.17697322e-01 -6.38494551e-01 6.42812669e-01 1.07334936e+00 2.87398756e-01 -8.39037001e-01 -3.95550489e-01 -5.69885790e-01 -6.14751950e-02 -6.95141256e-01 8.17169428e-01 9.45365012e-01 1.71176642e-01 3.36533606e-01 -3.07589024e-01 6.00293696e-01 3.88682336e-01 1.30762577e-01 7.87882626e-01 -1.37668240e+00 -4.20303464e-01 -1.13828406e-02 -1.32023916e-01 -1.05825675e+00 1.17540777e-01 -6.60161078e-01 -1.92081407e-01 -1.30277836e+00 3.48452061e-01 -4.58816916e-01 -3.25881928e-01 5.43282270e-01 -3.09776574e-01 1.56295434e-01 2.26220414e-01 2.71415979e-01 -6.42030239e-01 1.19466126e-01 7.36842513e-01 1.67930927e-02 1.85919121e-01 1.34189591e-01 -1.18608034e+00 9.57421184e-01 4.86916691e-01 -6.54951811e-01 -4.49222684e-01 -2.14202762e-01 7.49519289e-01 -9.87596437e-02 4.80809063e-01 -7.91500509e-01 3.11954528e-01 -1.81361988e-01 -4.00686003e-02 -3.30979109e-01 -2.17802554e-01 -7.98626781e-01 1.17146894e-01 3.52431953e-01 -5.34069955e-01 2.05847751e-02 -1.97977990e-01 3.32213938e-01 -2.44551107e-01 -4.96493667e-01 3.27361792e-01 -2.31239814e-02 -5.26516199e-01 -3.42719741e-02 -3.83025855e-01 1.03446588e-01 7.73280442e-01 -7.46877193e-02 -5.62200308e-01 -3.80878478e-01 -7.43815124e-01 4.50791372e-03 3.80287439e-01 3.55495363e-01 2.96424657e-01 -7.82679915e-01 -6.84583604e-01 -1.67573661e-01 3.85699034e-01 -3.30072254e-01 -2.24659988e-03 5.78957975e-01 -2.82584310e-01 7.12759733e-01 1.93802670e-01 -1.02152200e-02 -1.39150476e+00 5.44085503e-01 4.04803127e-01 -8.15463245e-01 -2.76821107e-01 5.16755223e-01 1.00542262e-01 -2.76969701e-01 8.08151588e-02 -3.56169343e-01 -6.04059458e-01 -1.01343691e-01 6.32579684e-01 3.85055751e-01 4.42559987e-01 -3.35930854e-01 -6.92700326e-01 3.58625829e-01 -4.13852662e-01 8.23978633e-02 1.15076566e+00 -5.46897165e-02 1.41172990e-01 -6.57009035e-02 3.24894339e-01 6.49385929e-01 -6.99235678e-01 -1.59430772e-01 5.53053260e-01 -3.44499439e-01 -1.66046709e-01 -1.40299690e+00 -7.36083210e-01 8.02831531e-01 8.67825970e-02 4.24758643e-01 8.41012955e-01 -1.38899773e-01 5.85899055e-01 5.14580071e-01 5.50868571e-01 -7.65255928e-01 -3.05788159e-01 5.30808926e-01 7.60567188e-01 -9.67756927e-01 3.72788727e-01 -1.23073804e+00 -7.63478458e-01 9.44307089e-01 6.80841327e-01 1.77713901e-01 3.08547765e-01 3.57457012e-01 1.95086434e-01 -2.66188085e-01 -1.06425214e+00 5.24534211e-02 4.27294493e-01 5.55536449e-01 6.07719123e-01 1.29277319e-01 -6.35962069e-01 1.39706409e+00 -4.62076128e-01 1.94315270e-01 7.40156472e-01 5.77862978e-01 -1.62216455e-01 -1.40652466e+00 -6.61568344e-02 2.00939506e-01 -7.53341377e-01 -2.54720449e-01 -9.05155480e-01 9.51597452e-01 -9.51250363e-03 1.36206424e+00 -7.02042222e-01 -7.33635902e-01 5.02512634e-01 2.09594429e-01 3.58886987e-01 -7.97041774e-01 -8.19993913e-01 -5.31512022e-01 8.87728751e-01 -4.32996452e-01 -2.69464642e-01 -3.80739331e-01 -9.41978395e-01 -1.57875940e-01 -4.13733125e-01 3.19192082e-01 3.20126891e-01 1.26940966e+00 3.70143086e-01 4.14114714e-01 1.44844711e-01 1.07410118e-01 -7.95572281e-01 -8.72752666e-01 -1.31406695e-01 1.28735170e-01 -2.42690623e-01 -6.49857223e-01 -4.43467826e-01 2.74049630e-03]
[9.362767219543457, 8.648487091064453]
4863607e-0146-49ec-a2e6-1793a20cb82a
combining-deep-neural-reranking-and
2302.01148
null
https://arxiv.org/abs/2302.01148v1
https://arxiv.org/pdf/2302.01148v1.pdf
Combining Deep Neural Reranking and Unsupervised Extraction for Multi-Query Focused Summarization
The CrisisFACTS Track aims to tackle challenges such as multi-stream fact-finding in the domain of event tracking; participants' systems extract important facts from several disaster-related events while incorporating the temporal order. We propose a combination of retrieval, reranking, and the well-known Integer Linear Programming (ILP) and Maximal Marginal Relevance (MMR) frameworks. In the former two modules, we explore various methods including an entity-based baseline, pre-trained and fine-tuned Question Answering systems, and ColBERT. We then use the latter module as an extractive summarization component by taking diversity and novelty criteria into account. The automatic scoring runs show strong results across the evaluation setups but also reveal shortcomings and challenges.
['Korbinian Riedhammer', 'Philipp Seeberger']
2023-02-02
null
null
null
null
['extractive-summarization']
['natural-language-processing']
[ 3.27236429e-02 1.89102754e-01 -4.42163497e-01 -1.24080427e-01 -1.69782770e+00 -6.87740803e-01 9.03679550e-01 1.03999841e+00 -8.85864675e-01 1.06782222e+00 1.24092162e+00 7.24881664e-02 -5.56471169e-01 -6.77659810e-01 -3.61689270e-01 -1.83810085e-01 -5.25388777e-01 6.58838987e-01 5.19742489e-01 -4.01645303e-01 7.79913068e-01 2.47529209e-01 -1.27568305e+00 7.18450963e-01 1.06349468e+00 8.26296151e-01 -3.10185134e-01 7.76736498e-01 -1.04361467e-01 1.29437256e+00 -9.70160186e-01 -5.32837212e-01 6.24753814e-03 -8.47123563e-02 -1.00288236e+00 -4.70372856e-01 4.04720366e-01 -4.14989144e-01 -3.16730678e-01 5.47797620e-01 8.48141253e-01 4.36768711e-01 5.99258244e-01 -1.10660815e+00 -3.79383415e-01 7.53705382e-01 -5.47208548e-01 1.07025039e+00 9.75950122e-01 -7.64428303e-02 1.28620148e+00 -8.76283467e-01 7.85536408e-01 1.15357852e+00 6.08178318e-01 -5.19247986e-02 -6.31227851e-01 -5.47180623e-02 5.15060961e-01 5.22135794e-01 -1.07864869e+00 -5.08774102e-01 4.87787038e-01 -2.53215879e-01 1.54118681e+00 6.95640922e-01 2.41063029e-01 7.90867448e-01 9.70109701e-02 9.64252651e-01 5.91709614e-01 -2.22253785e-01 3.42501104e-01 -1.82841673e-01 5.72712660e-01 2.53404558e-01 2.55111158e-01 -3.17849338e-01 -9.74837959e-01 -8.22736204e-01 -7.23566115e-02 -2.39509922e-02 -4.73691851e-01 6.49066329e-01 -1.47098696e+00 7.21308231e-01 2.23779663e-01 2.74042904e-01 -1.06258917e+00 -2.13322327e-01 7.88083553e-01 1.94826171e-01 8.07645321e-01 7.87057340e-01 -5.98411500e-01 -2.41620298e-02 -1.42228329e+00 7.23309278e-01 1.13725889e+00 5.25858402e-01 3.55560571e-01 -3.67505431e-01 -1.06320274e+00 6.81938529e-01 -1.62714690e-01 2.50651389e-01 3.99234802e-01 -8.73399556e-01 1.24819911e+00 6.23658001e-01 3.86294752e-01 -1.29118407e+00 -6.27926588e-01 -4.01119351e-01 -6.61365330e-01 -5.70078671e-01 -6.30194768e-02 -3.85592878e-01 -6.35115325e-01 1.44163036e+00 4.07394141e-01 3.02937299e-01 7.76055530e-02 7.42327452e-01 1.12905741e+00 9.51184392e-01 2.27888837e-01 -7.59956181e-01 1.43825436e+00 -1.05159628e+00 -8.36789072e-01 -1.31932795e-01 2.86552966e-01 -5.29657006e-01 6.30570769e-01 3.22813988e-01 -1.36314666e+00 4.43449505e-02 -8.31780851e-01 -1.24794185e-01 -5.09612083e-01 3.35056372e-02 3.48667026e-01 -7.73332119e-02 -1.01415312e+00 5.45830011e-01 -4.55306143e-01 -4.27087218e-01 9.66947824e-02 -8.14101920e-02 -1.93784803e-01 1.19934134e-01 -1.52024949e+00 1.11012793e+00 8.00653875e-01 -2.86639482e-01 -5.25831223e-01 -9.33890104e-01 -5.40818453e-01 3.07532191e-01 6.02358639e-01 -1.08624041e+00 1.22171605e+00 1.10550886e-02 -1.00030887e+00 6.28330588e-01 -2.94990629e-01 -6.49024189e-01 3.85656357e-01 -7.77708411e-01 -5.62495887e-01 7.99349606e-01 3.86583507e-01 2.18598410e-01 2.96254426e-01 -7.82015026e-01 -1.02899277e+00 -5.68744913e-02 2.20641553e-01 5.36314070e-01 -2.19483808e-01 4.64713395e-01 -3.29969347e-01 -7.88810909e-01 -1.41422957e-01 -2.25171134e-01 -4.05134171e-01 -7.67657518e-01 -7.57424176e-01 -4.59299386e-01 3.29663098e-01 -1.30739307e+00 2.14218020e+00 -1.51573527e+00 6.31789416e-02 6.48485422e-02 -3.44716571e-02 1.22023627e-01 -1.89441457e-01 1.14459491e+00 1.26908749e-01 2.55004227e-01 -2.69115061e-01 -2.40099907e-01 1.34779707e-01 -2.17328474e-01 -7.92031527e-01 -1.25474334e-01 3.99162650e-01 8.94175410e-01 -1.20613790e+00 -7.59426475e-01 -3.31519067e-01 -1.30526647e-01 -5.71298063e-01 7.74191245e-02 -3.77430707e-01 1.15375232e-03 -5.90917885e-01 6.39663815e-01 2.47685969e-01 -3.13894525e-02 -1.54682204e-01 -1.76735565e-01 -3.24315637e-01 7.25674212e-01 -1.20382833e+00 1.46629012e+00 7.17209950e-02 2.62921304e-01 -2.35599086e-01 -7.26755857e-01 3.40154111e-01 6.39657974e-01 6.39894843e-01 -6.85163558e-01 -4.03655916e-01 9.95660722e-02 -8.23124528e-01 -8.38679910e-01 1.37147141e+00 1.93511009e-01 -4.81405377e-01 6.80092335e-01 7.28122368e-02 3.48350227e-01 7.55685091e-01 9.23348248e-01 1.51638913e+00 -3.92521709e-01 7.40997851e-01 4.85919043e-02 3.20410490e-01 4.36122090e-01 4.58485544e-01 1.10668087e+00 2.98168510e-02 8.87274325e-01 6.06447101e-01 -4.02084291e-01 -6.31824553e-01 -8.22057128e-01 2.12426305e-01 1.30044818e+00 6.20166725e-03 -9.47578967e-01 -3.21531922e-01 -9.04739916e-01 2.64228974e-02 1.04862285e+00 -5.62023163e-01 8.71204026e-03 -7.95054197e-01 -1.16930699e+00 6.32521451e-01 3.36642146e-01 3.82398367e-01 -1.08211815e+00 -6.95973039e-01 5.03461361e-01 -1.07477808e+00 -9.38518882e-01 -4.46190894e-01 2.18788832e-02 -5.20003259e-01 -1.06819773e+00 -6.92411005e-01 -2.44789481e-01 1.27388626e-01 1.30428806e-01 1.44750750e+00 1.40560521e-02 5.56116439e-02 6.55924320e-01 -6.71974123e-01 -5.03836751e-01 1.89288229e-01 4.33606952e-01 -8.47558752e-02 -3.94730985e-01 4.18180376e-02 -6.16656840e-01 -7.98705876e-01 -1.17658861e-01 -1.09606636e+00 -4.74370450e-01 4.70482528e-01 6.19841695e-01 4.83056754e-01 -5.04578836e-02 1.16560268e+00 -8.61386776e-01 1.23897123e+00 -9.99364018e-01 8.88871476e-02 7.96938598e-01 -3.43724132e-01 -9.25127789e-03 2.86142647e-01 -1.89365670e-01 -1.13585508e+00 -6.00777447e-01 -1.31537601e-01 3.10780972e-01 5.62732108e-02 1.15681982e+00 3.66966836e-02 5.88303745e-01 9.13698733e-01 1.97602317e-01 -9.82680559e-01 -4.72960800e-01 5.28490007e-01 5.09863913e-01 8.44352543e-01 -4.19637531e-01 7.19771028e-01 2.56490529e-01 -6.39171183e-01 -5.72065651e-01 -1.35116792e+00 -1.04391336e+00 -3.19107473e-01 -1.36412039e-01 5.85016310e-01 -1.06174827e+00 -2.87763000e-01 8.95166248e-02 -1.67279172e+00 1.48819342e-01 -5.72993100e-01 3.38538319e-01 -1.79944545e-01 5.84094763e-01 -6.96683943e-01 -8.29002500e-01 -1.01754594e+00 -2.27803424e-01 1.28042948e+00 2.49410540e-01 -3.14308316e-01 -6.87670529e-01 6.57856703e-01 3.30173969e-01 3.76911640e-01 6.78502917e-01 7.47612357e-01 -1.29887772e+00 -4.81055439e-01 -3.71628046e-01 -1.46000713e-01 -2.91084856e-01 -2.77748734e-01 -2.31213689e-01 -6.72005594e-01 -1.02256224e-01 -2.61583209e-01 -2.71958262e-01 1.43272579e+00 3.36446762e-01 4.73268747e-01 -1.02889907e+00 -3.55837762e-01 -4.80927080e-02 1.23972452e+00 -1.85147643e-01 5.84035516e-01 7.36497998e-01 2.69322991e-01 6.52851105e-01 7.88365543e-01 8.54172766e-01 1.03405309e+00 4.59957361e-01 1.84645608e-01 2.27339983e-01 3.17338020e-01 -2.58869082e-01 3.50307792e-01 7.89913416e-01 -1.22348107e-01 -5.57754517e-01 -9.49635983e-01 8.83531272e-01 -2.11938977e+00 -1.63009012e+00 -3.48499976e-02 1.95286834e+00 9.68654990e-01 1.29489049e-01 3.48086923e-01 1.74697071e-01 5.04445553e-01 5.23338795e-01 -2.72326320e-01 -2.93689817e-01 -3.95605922e-01 4.45012040e-02 3.31840724e-01 4.33681250e-01 -1.28906560e+00 7.12956965e-01 6.63711882e+00 8.07380319e-01 -6.04227126e-01 2.36081123e-01 4.47837919e-01 -3.78677309e-01 -4.53293502e-01 2.47230455e-02 -7.49066949e-01 3.88079047e-01 9.81021464e-01 -6.06989384e-01 3.14105349e-03 3.64945441e-01 2.21605182e-01 -5.64313531e-01 -7.70398259e-01 5.16124249e-01 3.33126605e-01 -1.59163380e+00 1.86113775e-01 -3.42500299e-01 8.57136011e-01 2.37412721e-01 -5.19298911e-01 6.24447942e-01 2.82127410e-01 -4.41499949e-01 8.19937944e-01 9.27493513e-01 2.78544575e-01 -5.94142079e-01 7.66579032e-01 6.85153902e-01 -1.16580498e+00 -1.97835758e-01 -1.24369144e-01 1.65679883e-02 7.91162968e-01 9.99001920e-01 -7.92182148e-01 1.25721669e+00 7.43570209e-01 4.87002015e-01 -6.29974544e-01 1.57926643e+00 -3.42458218e-01 7.68354416e-01 -5.72004855e-01 1.71905816e-01 1.83601707e-01 3.94587666e-01 1.18972838e+00 1.67743444e+00 1.69726610e-01 5.09483635e-01 3.78736347e-01 1.65915877e-01 -9.86422747e-02 2.66451657e-01 -1.92687988e-01 2.90321290e-01 5.89347661e-01 1.23754120e+00 -6.21222794e-01 -5.67476213e-01 8.76235291e-02 7.02310562e-01 4.03438687e-01 3.61821741e-01 -6.76034153e-01 -4.80401874e-01 2.44230945e-02 -7.01020882e-02 1.57873914e-01 1.29412133e-02 -1.40458807e-01 -1.35426795e+00 3.10297191e-01 -7.09505200e-01 1.26301610e+00 -5.92725098e-01 -1.23901856e+00 7.77362287e-01 5.93989670e-01 -1.01486206e+00 -4.74698216e-01 2.46780396e-01 -1.07402968e+00 4.15824652e-01 -1.94285309e+00 -8.08677018e-01 -6.67220131e-02 3.33968401e-01 5.63222051e-01 2.92998124e-02 6.05036378e-01 5.99397302e-01 -7.51707256e-01 3.55086595e-01 -1.66203771e-02 -2.66905934e-01 8.31949055e-01 -1.38122475e+00 3.32616568e-01 1.01338506e+00 5.85426874e-02 4.68927562e-01 7.28321731e-01 -8.84646952e-01 -7.13819325e-01 -1.23486805e+00 1.69586849e+00 -4.99499828e-01 5.82048595e-01 1.05065487e-01 -1.06534934e+00 4.54875261e-01 6.41379058e-01 -4.77051228e-01 6.55231118e-01 3.37028772e-01 -2.77526349e-01 7.75426030e-02 -1.08265185e+00 3.56004000e-01 8.73431027e-01 -3.27322215e-01 -1.26982617e+00 6.99922383e-01 8.58673275e-01 -3.25629562e-01 -7.53660142e-01 3.93235743e-01 2.13647574e-01 -7.74685562e-01 1.18017089e+00 -8.45276415e-01 4.55798626e-01 -1.93372682e-01 2.78465319e-02 -1.28147781e+00 -2.26854309e-01 -8.07245970e-01 -4.54015553e-01 1.71800637e+00 6.78893745e-01 -5.29116154e-01 1.92796081e-01 5.81896603e-01 -1.30667508e-01 -6.48128688e-01 -9.44632828e-01 -4.47552472e-01 -2.79314399e-01 -1.81704432e-01 5.76842010e-01 9.83519673e-01 5.27882934e-01 5.62855542e-01 -4.86754239e-01 3.98509532e-01 2.78633118e-01 1.13190889e-01 4.14735585e-01 -1.17377007e+00 -1.34757683e-01 -3.45137835e-01 1.07689224e-01 -7.03424513e-01 -2.14451119e-01 -7.33636320e-01 -1.16246217e-03 -2.07277322e+00 5.02027929e-01 -2.07266547e-02 -6.80503666e-01 4.38839287e-01 -8.30800116e-01 -1.93298653e-01 2.45424822e-01 5.04771411e-01 -1.61409831e+00 5.67678273e-01 5.09670079e-01 -5.85547723e-02 -4.32689995e-01 4.35433015e-02 -9.17780399e-01 6.21386588e-01 7.95136213e-01 -8.67641389e-01 -9.84960198e-02 -4.73423481e-01 6.65432215e-01 4.02332038e-01 3.93642873e-01 -9.07692134e-01 8.75752270e-01 -2.96773672e-01 -3.02697439e-02 -1.26022220e+00 -8.55669938e-03 -1.99418887e-01 1.57828312e-02 4.85729501e-02 -7.20800042e-01 4.65386420e-01 2.02757835e-01 7.28892922e-01 -5.13395309e-01 -1.63866371e-01 1.18787671e-02 -1.73632175e-01 -6.48720205e-01 1.44726902e-01 -2.71861047e-01 7.05190182e-01 6.84050024e-01 8.88444707e-02 -8.98269653e-01 -5.47187328e-01 -4.10556644e-01 8.83137941e-01 -3.38711619e-01 2.93679446e-01 6.17503405e-01 -1.15726388e+00 -1.34263909e+00 -8.24364722e-01 2.83110499e-01 -1.06922708e-01 3.76185268e-01 1.12784791e+00 -2.11710379e-01 5.16299009e-01 1.29542693e-01 -2.27403138e-02 -8.44218135e-01 2.93532312e-01 -1.20044991e-01 -1.37546980e+00 -4.25765127e-01 6.27605677e-01 -3.34997147e-01 -3.23474228e-01 3.69322777e-01 -1.58814147e-01 -8.04816663e-01 7.41333604e-01 9.94812012e-01 7.91136980e-01 3.01059961e-01 -3.49416912e-01 -4.77152318e-01 3.79846022e-02 -2.09197924e-01 -3.38575900e-01 1.56206572e+00 -2.05115229e-01 -1.08189061e-01 1.21077351e-01 6.98915184e-01 1.40310720e-01 -7.04541922e-01 -5.20996273e-01 7.50286400e-01 2.35297680e-02 -4.64751050e-02 -1.19562542e+00 -4.90539104e-01 2.72001624e-01 -9.63777527e-02 5.82730770e-01 1.33106613e+00 -2.22947076e-02 9.66775596e-01 5.31273067e-01 1.90111995e-01 -1.23135626e+00 5.55478372e-02 8.48681688e-01 1.40375006e+00 -9.96472061e-01 3.88316274e-01 -1.32407174e-01 -9.06766951e-01 7.65116751e-01 2.83980697e-01 -1.59762204e-02 3.97535443e-01 2.55740974e-02 -2.07834229e-01 -4.37846601e-01 -1.23034549e+00 -4.81996566e-01 6.67176366e-01 7.76788443e-02 2.51616091e-02 -2.82850921e-01 -8.47371519e-01 1.03751373e+00 -1.16001487e-01 -1.05056547e-01 2.13243902e-01 1.08357370e+00 -7.35505283e-01 -6.73448265e-01 -4.03648436e-01 6.68647766e-01 -8.49773049e-01 -3.24034005e-01 -5.14454365e-01 5.08388579e-01 -1.05247766e-01 1.21563673e+00 -2.99276918e-01 -2.71934509e-01 9.12437499e-01 1.29994258e-01 -1.57718554e-01 -6.15232944e-01 -1.44330692e+00 -2.28607997e-01 5.80593109e-01 -5.87558389e-01 -4.83857930e-01 -8.54562819e-01 -9.61274743e-01 3.65004018e-02 -3.95197779e-01 5.66576004e-01 2.46971384e-01 1.01629162e+00 6.16394699e-01 4.26100850e-01 6.10664964e-01 -7.69403636e-01 -5.08420348e-01 -1.02107894e+00 -1.29433632e-01 2.45978490e-01 4.07219350e-01 -1.44162416e-01 -3.59694988e-01 -5.15886098e-02]
[12.52050495147705, 9.487093925476074]
f1a340ed-d769-4047-a208-ada753fcad39
ego-pose-estimation-and-forecasting-as-real
1906.03173
null
https://arxiv.org/abs/1906.03173v2
https://arxiv.org/pdf/1906.03173v2.pdf
Ego-Pose Estimation and Forecasting as Real-Time PD Control
We propose the use of a proportional-derivative (PD) control based policy learned via reinforcement learning (RL) to estimate and forecast 3D human pose from egocentric videos. The method learns directly from unsegmented egocentric videos and motion capture data consisting of various complex human motions (e.g., crouching, hopping, bending, and motion transitions). We propose a video-conditioned recurrent control technique to forecast physically-valid and stable future motions of arbitrary length. We also introduce a value function based fail-safe mechanism which enables our method to run as a single pass algorithm over the video data. Experiments with both controlled and in-the-wild data show that our approach outperforms previous art in both quantitative metrics and visual quality of the motions, and is also robust enough to transfer directly to real-world scenarios. Additionally, our time analysis shows that the combined use of our pose estimation and forecasting can run at 30 FPS, making it suitable for real-time applications.
['Ye Yuan', 'Kris Kitani']
2019-06-07
ego-pose-estimation-and-forecasting-as-real-1
http://openaccess.thecvf.com/content_ICCV_2019/html/Yuan_Ego-Pose_Estimation_and_Forecasting_As_Real-Time_PD_Control_ICCV_2019_paper.html
http://openaccess.thecvf.com/content_ICCV_2019/papers/Yuan_Ego-Pose_Estimation_and_Forecasting_As_Real-Time_PD_Control_ICCV_2019_paper.pdf
iccv-2019-10
['human-pose-forecasting', 'egocentric-pose-estimation']
['computer-vision', 'computer-vision']
[-1.40687957e-01 1.66698024e-02 1.76188711e-03 1.36563517e-02 -6.60403728e-01 -5.96252441e-01 6.42163396e-01 -5.11789739e-01 -3.77025366e-01 6.95078611e-01 3.68798435e-01 4.97613512e-02 4.56675291e-02 -3.90882641e-01 -9.22184050e-01 -5.74864149e-01 -5.34402847e-01 4.80468303e-01 4.64288682e-01 -2.89630294e-01 8.97356868e-02 5.69467068e-01 -1.60552335e+00 1.95227866e-03 2.80100673e-01 6.99313343e-01 3.02507252e-01 1.26634848e+00 8.18825543e-01 1.08741999e+00 -1.55567974e-01 7.01393858e-02 4.44655061e-01 -3.18847775e-01 -6.74927533e-01 5.06079018e-01 3.51235747e-01 -8.50109279e-01 -7.05493689e-01 3.81124586e-01 7.71479666e-01 6.64379895e-01 3.64915252e-01 -1.08027864e+00 1.44531980e-01 -1.17653064e-01 -3.62248540e-01 1.99630201e-01 9.65672016e-01 6.04215205e-01 6.11788869e-01 -6.92761540e-01 1.13373446e+00 1.25695002e+00 6.24519587e-01 7.81713843e-01 -1.04866052e+00 -8.71741325e-02 9.18098092e-02 1.60504535e-01 -1.01763904e+00 -5.84457099e-01 6.29114449e-01 -6.05860233e-01 1.14043057e+00 1.25471815e-01 1.04941761e+00 1.39982128e+00 5.07571638e-01 1.00996244e+00 5.04646599e-01 -3.33774656e-01 4.52797860e-01 -6.52241647e-01 -8.55205536e-01 8.82932305e-01 -6.98106661e-02 2.56786853e-01 -4.15545344e-01 2.20564418e-02 1.32703984e+00 -1.25037981e-02 -3.54736060e-01 -8.88829112e-01 -1.62244904e+00 5.72709143e-01 2.23771706e-01 -2.04951629e-01 -6.03526711e-01 6.96676493e-01 3.27789158e-01 2.51010507e-02 3.19283575e-01 2.74660140e-01 -4.78360444e-01 -7.25308776e-01 -7.54612327e-01 8.49653721e-01 6.59065485e-01 1.00468922e+00 3.20235252e-01 2.50772923e-01 -1.70945585e-01 3.10462236e-01 6.65168315e-02 6.03311300e-01 3.39932173e-01 -1.81833553e+00 3.67263019e-01 1.03350870e-01 7.63971865e-01 -1.03054035e+00 -6.08608484e-01 1.24751776e-01 -2.28663027e-01 3.67288202e-01 5.08231938e-01 -4.20605361e-01 -7.60113835e-01 1.69723010e+00 5.51324427e-01 3.71752024e-01 -1.78815007e-01 1.29001653e+00 1.05107203e-01 5.59890926e-01 1.89107296e-03 -2.33420566e-01 8.54844511e-01 -9.44947183e-01 -4.15160894e-01 -1.88236982e-01 6.68037713e-01 -4.65100139e-01 9.19854641e-01 5.24766445e-01 -1.33567870e+00 -5.89953482e-01 -7.31104195e-01 6.94358870e-02 2.33657107e-01 7.96346068e-02 4.30156887e-01 1.85817450e-01 -9.82342541e-01 1.01685071e+00 -1.42468607e+00 -5.31057894e-01 -1.23866580e-01 3.79900068e-01 -5.47365487e-01 1.00187436e-01 -8.63174915e-01 7.76628733e-01 2.10543439e-01 -1.39242224e-02 -1.20152652e+00 -3.30693334e-01 -1.03620374e+00 -3.00140500e-01 5.93105555e-01 -1.01425779e+00 1.53747869e+00 -9.06882346e-01 -1.89859879e+00 6.19311869e-01 -6.20658956e-02 -4.46639717e-01 9.34950888e-01 -8.37278605e-01 5.86610883e-02 6.84834719e-01 -1.20686382e-01 8.63525569e-01 8.79507422e-01 -1.04622364e+00 -4.91692632e-01 -9.23300087e-02 1.97283834e-01 6.94000542e-01 1.32354409e-01 -3.16998810e-01 -5.85950255e-01 -7.05408454e-01 -8.49585980e-02 -1.41246676e+00 -3.50288093e-01 1.97015822e-01 -1.84624553e-01 1.48605742e-02 7.62998402e-01 -7.19322920e-01 9.54893172e-01 -1.76368558e+00 7.24362493e-01 -1.34594187e-01 -1.14666246e-01 -6.18241681e-03 3.08219474e-02 5.63737571e-01 1.31476417e-01 -4.42568660e-01 1.77494273e-01 -3.04886401e-01 -1.57718152e-01 2.98941344e-01 -2.31371731e-01 6.96091592e-01 9.48048756e-02 1.04687226e+00 -1.24568605e+00 -4.14031416e-01 4.86732751e-01 4.75498557e-01 -8.56000543e-01 5.57633281e-01 -3.04691911e-01 8.74385536e-01 -5.32566309e-01 5.64174056e-01 -4.43152562e-02 -1.04545996e-01 2.29056761e-01 7.18422905e-02 4.94378107e-03 -7.47204944e-02 -1.05988753e+00 2.09441233e+00 -3.28230202e-01 5.34103751e-01 -4.45148908e-02 -6.57669902e-01 4.82410967e-01 3.66564482e-01 7.41438210e-01 -3.36691737e-01 2.57724792e-01 -5.29710799e-02 -4.11227793e-01 -9.86687064e-01 5.77008963e-01 8.58292133e-02 -3.69957760e-02 3.14523518e-01 -3.87266986e-02 -1.76682249e-01 1.75067604e-01 1.01107195e-01 1.25557041e+00 9.91523027e-01 1.78167820e-01 -5.70978671e-02 3.80888283e-01 1.50349692e-01 5.15131891e-01 4.98029858e-01 -3.77714276e-01 7.64306962e-01 3.40696305e-01 -5.63347995e-01 -1.43187296e+00 -1.09783626e+00 6.44843280e-01 9.52718914e-01 2.50270873e-01 -2.75284886e-01 -8.78073752e-01 -3.38574976e-01 -6.71211928e-02 4.42757875e-01 -4.66096044e-01 -1.86399728e-01 -1.06479526e+00 -1.67757217e-02 2.34057810e-02 7.82673955e-01 2.94517130e-01 -1.16792989e+00 -1.39604223e+00 4.32056129e-01 -3.31218719e-01 -1.24072194e+00 -6.62379086e-01 -4.11257178e-01 -9.87388790e-01 -1.06367552e+00 -9.57796574e-01 -5.08880794e-01 4.55556959e-01 2.15677813e-01 9.37246084e-01 -2.22996220e-01 -3.44218612e-01 1.02364004e+00 -4.04945523e-01 3.00228238e-01 -1.62741542e-01 -3.79029095e-01 4.80196148e-01 8.37142207e-03 -2.07035407e-01 -5.50675631e-01 -1.00683534e+00 3.60805154e-01 -4.91685301e-01 7.52491206e-02 2.29829982e-01 7.75647283e-01 5.47976315e-01 -4.21438664e-01 1.80644944e-01 -1.97282404e-01 4.16323245e-01 -2.98691690e-01 -4.62922662e-01 4.91143428e-02 -1.01130798e-01 -8.23069513e-02 4.50878710e-01 -7.70532370e-01 -9.81618166e-01 5.61748087e-01 1.95031717e-01 -9.23383355e-01 -6.02723062e-02 4.86688428e-02 1.39044985e-01 9.58109871e-02 6.55134022e-01 1.38083145e-01 1.94178388e-01 -8.58864188e-02 4.68642741e-01 1.71364263e-01 9.43329990e-01 -6.14419341e-01 6.13646746e-01 7.81217933e-01 1.75277978e-01 -8.62921059e-01 -2.43169293e-01 -3.40541750e-01 -7.46892989e-01 -6.78694665e-01 1.03034329e+00 -1.13498902e+00 -1.19206059e+00 4.63601083e-01 -9.70999360e-01 -7.99172282e-01 -2.07062095e-01 7.24763811e-01 -1.52800822e+00 6.17585063e-01 -8.70389700e-01 -1.02952325e+00 -1.50296420e-01 -1.05773807e+00 1.47503841e+00 1.16426580e-01 -4.53971416e-01 -8.74185026e-01 3.29413444e-01 1.20717332e-01 9.14293453e-02 7.05727756e-01 1.69730097e-01 7.32672289e-02 -7.13189542e-01 -3.23206335e-01 3.60183060e-01 -1.00759409e-01 -1.33443207e-01 6.18155189e-02 -5.71979225e-01 -6.64678633e-01 -2.76740044e-01 -5.49009383e-01 5.00624418e-01 6.77501082e-01 9.18906271e-01 -3.87047112e-01 -2.41993606e-01 4.82292414e-01 1.07925177e+00 2.28936132e-02 7.57454991e-01 3.27560931e-01 5.44674933e-01 6.11470699e-01 9.62460756e-01 7.69616902e-01 3.89569312e-01 9.81272757e-01 3.74361277e-01 2.80923933e-01 3.89086120e-02 -5.05948663e-01 6.15713298e-01 4.41527903e-01 -5.85043192e-01 -2.22327307e-01 -7.52669573e-01 5.07185161e-01 -2.15757537e+00 -1.31979322e+00 3.84204715e-01 2.26262355e+00 2.97807604e-01 1.78528443e-01 5.08963645e-01 -1.18332550e-01 5.18276691e-01 -7.67032653e-02 -7.46669948e-01 -2.19398320e-01 3.56622785e-01 -1.45677045e-01 4.02416617e-01 5.29526174e-01 -1.15215993e+00 9.66223896e-01 6.75531292e+00 3.41497272e-01 -1.14092803e+00 -2.66114205e-01 3.04487497e-01 -4.21660066e-01 6.13093674e-02 -3.07212025e-02 -2.83734232e-01 3.28959078e-01 8.98834944e-01 6.44089654e-02 5.51871717e-01 9.81693625e-01 8.02775502e-01 -3.17639023e-01 -1.01265717e+00 1.04293835e+00 -7.84345120e-02 -1.19385254e+00 -2.06518933e-01 -4.99525629e-02 6.00373983e-01 8.33668688e-04 -1.06772937e-01 1.96144104e-01 2.69239068e-01 -7.21365690e-01 9.18515325e-01 8.70609224e-01 7.23779142e-01 -7.69142926e-01 2.48648822e-01 5.98631501e-01 -1.11524785e+00 -2.33589575e-01 -8.16500410e-02 -2.70593524e-01 5.66905737e-01 -4.50260937e-02 -6.62579894e-01 3.70219737e-01 6.20017827e-01 8.77152324e-01 -2.15037733e-01 8.82373452e-01 -2.05448836e-01 2.76245713e-01 -3.53353202e-01 -1.09106228e-01 1.42580122e-01 2.32292991e-02 8.35155189e-01 9.78290260e-01 4.07988906e-01 2.83049822e-01 4.20276761e-01 2.85487741e-01 4.66676146e-01 -2.36116469e-01 -7.73062229e-01 1.01807907e-01 5.85845001e-02 1.00857282e+00 -7.42835283e-01 -2.83457845e-01 -1.50235603e-02 1.37809002e+00 2.62578636e-01 5.51633775e-01 -8.52590740e-01 -1.71087116e-01 6.32237494e-01 2.10067213e-01 6.47718072e-01 -8.78317893e-01 3.37346733e-01 -1.58190489e+00 1.91823348e-01 -5.48370659e-01 2.05317780e-01 -1.05379128e+00 -7.19091356e-01 3.66046816e-01 2.95870453e-01 -1.68302333e+00 -1.18014002e+00 -4.18384939e-01 -5.13971388e-01 2.80192971e-01 -9.71615374e-01 -9.06166971e-01 -3.46208572e-01 6.06827974e-01 8.00235450e-01 -4.39144559e-02 5.78288674e-01 -1.21226616e-01 -2.52168268e-01 2.92734858e-02 -1.11239798e-01 -6.71026483e-02 5.65959632e-01 -1.12302136e+00 5.91576874e-01 6.88778996e-01 -1.07165918e-01 4.24514323e-01 1.14709926e+00 -7.01406717e-01 -1.79739332e+00 -8.01728189e-01 3.35428894e-01 -6.95339262e-01 4.67333496e-01 -2.62799442e-01 -4.67533946e-01 7.38137841e-01 -1.12656869e-01 1.53223976e-01 2.10988522e-03 -5.15986621e-01 8.97369012e-02 8.62127319e-02 -1.06534648e+00 8.49333584e-01 1.17467499e+00 -1.75892711e-01 -5.20464063e-01 2.82404304e-01 5.56883216e-01 -8.60505223e-01 -5.91805160e-01 3.41405720e-01 9.59786296e-01 -9.45434272e-01 1.01677763e+00 -7.11293340e-01 4.95883435e-01 -1.56117946e-01 -1.19888857e-01 -1.17758763e+00 -2.66175807e-01 -1.16398621e+00 -5.66799760e-01 4.42761987e-01 -2.06864163e-01 -1.30151689e-01 9.85279083e-01 6.01970494e-01 1.26949146e-01 -7.19673455e-01 -9.31330562e-01 -8.98091555e-01 -2.45705992e-01 -3.02854180e-01 2.09730342e-02 4.14609909e-01 1.61061987e-01 6.00043312e-02 -8.71075571e-01 2.47751977e-02 7.34319746e-01 -5.79927154e-02 1.13915050e+00 -6.32029772e-01 -5.14995694e-01 1.59173607e-04 -7.43347406e-01 -1.36888945e+00 5.60976230e-02 -2.78677583e-01 2.35330373e-01 -1.22347867e+00 -6.46328032e-02 1.79383889e-01 2.17381865e-01 1.86820984e-01 -4.02884884e-03 6.72939122e-02 3.20165396e-01 3.08011293e-01 -6.80966735e-01 7.16463447e-01 1.39720190e+00 3.66317838e-01 -3.89773935e-01 5.98389767e-02 1.87092856e-01 7.14852512e-01 5.67113280e-01 -1.96150571e-01 -4.76474315e-01 -2.72260666e-01 4.38397601e-02 6.09392226e-01 5.83665490e-01 -1.30798483e+00 1.16146356e-01 -3.10665339e-01 5.17411172e-01 -4.64215785e-01 7.35274434e-01 -5.36281049e-01 1.45009071e-01 8.39696944e-01 -2.33101428e-01 3.19038093e-01 5.69277965e-02 1.00090694e+00 2.85789251e-01 3.81512076e-01 4.98485118e-01 -3.27941597e-01 -9.84556556e-01 2.37368777e-01 -4.43817586e-01 1.19029664e-01 1.29624593e+00 -4.18576270e-01 1.59673765e-01 -9.32376862e-01 -9.51392233e-01 3.71298939e-01 6.47745311e-01 5.06525278e-01 8.16875458e-01 -1.28707945e+00 -6.57441854e-01 2.44209170e-02 -7.82326162e-02 -8.81165341e-02 3.91528904e-01 7.40721285e-01 -9.93750155e-01 3.72831196e-01 -3.62851113e-01 -9.08278227e-01 -9.70486045e-01 5.69061995e-01 2.71032721e-01 -7.78672770e-02 -9.78678584e-01 3.31651241e-01 3.80026773e-02 -1.95975080e-01 1.09017774e-01 -1.85062930e-01 7.83020109e-02 -4.13130492e-01 3.94748092e-01 5.97096980e-01 -3.40559155e-01 -6.45394921e-01 -2.24645153e-01 6.41703486e-01 3.43717039e-01 -5.52741826e-01 1.35284150e+00 -2.97996938e-01 5.09690940e-01 5.05847573e-01 9.28439796e-01 -2.54924923e-01 -2.28079462e+00 2.26192743e-01 -2.04305306e-01 -5.71137488e-01 -2.97027767e-01 -4.06953037e-01 -7.49177039e-01 6.47917628e-01 4.40148443e-01 -2.92200178e-01 8.42674971e-01 -1.75104365e-01 9.18402851e-01 5.58607459e-01 7.19596803e-01 -1.28470922e+00 5.11279464e-01 3.54577363e-01 9.81658220e-01 -9.38487470e-01 8.95938128e-02 -1.65506452e-01 -8.92823160e-01 1.25325477e+00 5.31341136e-01 -5.17753005e-01 3.71296287e-01 1.11581208e-02 -2.98448652e-02 7.86649287e-02 -1.11120296e+00 -5.08908667e-02 2.65915334e-01 6.36325777e-01 8.69363248e-02 -6.38885647e-02 -1.06138267e-01 -1.16622597e-01 1.57936104e-02 2.60911703e-01 7.95989752e-01 1.33378053e+00 -5.31029582e-01 -7.23445535e-01 -3.09878439e-01 -1.01457559e-01 -2.25154728e-01 6.28596365e-01 -5.37126176e-02 1.02747190e+00 -3.61157537e-01 7.61517107e-01 5.26947603e-02 -4.34456706e-01 3.97951365e-01 -5.01206405e-02 8.34392071e-01 -3.50487590e-01 -3.15565288e-01 3.36954653e-01 9.34796557e-02 -1.23961401e+00 -4.43680912e-01 -8.49987507e-01 -1.31589997e+00 -2.50891566e-01 5.16035669e-02 -3.15501928e-01 3.71861875e-01 8.55278194e-01 4.60939109e-01 1.50871336e-01 6.04666471e-01 -1.60327792e+00 -6.22193933e-01 -6.95230007e-01 -4.35283273e-01 6.67547464e-01 5.88771462e-01 -8.98469210e-01 -1.64705232e-01 2.99240679e-01]
[7.158624649047852, -0.5237789154052734]
85db75ce-d935-470d-908d-40ba62ba8b14
discourse-analysis-for-evaluating-coherence
2201.06207
null
https://arxiv.org/abs/2201.06207v1
https://arxiv.org/pdf/2201.06207v1.pdf
Discourse Analysis for Evaluating Coherence in Video Paragraph Captions
Video paragraph captioning is the task of automatically generating a coherent paragraph description of the actions in a video. Previous linguistic studies have demonstrated that coherence of a natural language text is reflected by its discourse structure and relations. However, existing video captioning methods evaluate the coherence of generated paragraphs by comparing them merely against human paragraph annotations and fail to reason about the underlying discourse structure. At UCLA, we are currently exploring a novel discourse based framework to evaluate the coherence of video paragraphs. Central to our approach is the discourse representation of videos, which helps in modeling coherence of paragraphs conditioned on coherence of videos. We also introduce DisNet, a novel dataset containing the proposed visual discourse annotations of 3000 videos and their paragraphs. Our experiment results have shown that the proposed framework evaluates coherence of video paragraphs significantly better than all the baseline methods. We believe that many other multi-discipline Artificial Intelligence problems such as Visual Dialog and Visual Storytelling would also greatly benefit from the proposed visual discourse framework and the DisNet dataset.
['Song-Chun Zhu', 'Arjun R Akula']
2022-01-17
null
null
null
null
['visual-storytelling']
['natural-language-processing']
[ 8.47777948e-02 6.37539029e-01 -3.73523742e-01 -3.65765363e-01 -4.98520494e-01 -6.11185789e-01 1.30083537e+00 1.99875563e-01 1.21708333e-01 9.81384456e-01 1.29535747e+00 -6.98265061e-02 4.14966106e-01 -3.57546955e-01 -8.29266191e-01 -3.91287863e-01 -8.67860243e-02 1.91181675e-01 1.47691533e-01 -7.08492696e-02 5.21376967e-01 -2.75500417e-01 -1.56948900e+00 9.94525552e-01 4.51157272e-01 5.93258262e-01 2.83983618e-01 8.90723646e-01 -4.41153944e-01 1.83538175e+00 -9.48813319e-01 -2.34188572e-01 -3.96733791e-01 -8.88960302e-01 -1.11698520e+00 4.84370738e-01 6.69236898e-01 -5.81498981e-01 -3.85565579e-01 8.41679871e-01 1.19276389e-01 1.96009383e-01 8.17360759e-01 -1.56163859e+00 -7.61034846e-01 1.14645505e+00 -4.28857893e-01 1.59317136e-01 1.06926906e+00 1.84036136e-01 1.06919515e+00 -6.81129217e-01 1.28955126e+00 1.47651315e+00 1.89750060e-01 3.36667269e-01 -9.02787328e-01 -2.61808515e-01 3.79450887e-01 4.76141036e-01 -1.00216937e+00 -5.64274728e-01 9.63288486e-01 -8.31981242e-01 9.39152837e-01 4.47498053e-01 7.29348063e-01 1.44878662e+00 -2.70498216e-01 9.74655330e-01 6.94495022e-01 -4.37867254e-01 5.22630326e-02 1.45211980e-01 1.66367754e-01 6.34346306e-01 9.50909033e-03 -5.28772414e-01 -6.76735163e-01 1.84257671e-01 6.61612809e-01 -4.72504705e-01 -4.95495379e-01 -2.19370216e-01 -1.87372375e+00 7.60241985e-01 1.54928967e-01 4.66324687e-01 -3.24338466e-01 1.99730530e-01 7.85325766e-01 -8.12257901e-02 4.27949131e-01 4.57524508e-01 5.98971546e-01 -3.44508410e-01 -1.10865402e+00 4.91337150e-01 8.59211683e-01 1.24933982e+00 3.82837951e-01 -1.50387332e-01 -8.94998252e-01 4.29804891e-01 3.09666812e-01 2.24820510e-01 3.39981794e-01 -1.52733850e+00 8.39292228e-01 6.09308958e-01 3.39616746e-01 -1.52493155e+00 -1.45703882e-01 4.16044414e-01 -5.91744959e-01 -2.81634405e-02 2.24767044e-01 -6.30327016e-02 -4.13879722e-01 1.64884031e+00 1.43556505e-01 2.54853368e-01 2.58031934e-01 9.63604033e-01 1.71780491e+00 1.00467610e+00 2.32160091e-01 -6.39186919e-01 1.41078353e+00 -1.17700005e+00 -1.19141638e+00 1.44956350e-01 4.07981336e-01 -9.38893318e-01 1.04966974e+00 -8.44390765e-02 -1.33207715e+00 -5.62656343e-01 -1.14872384e+00 -3.58026057e-01 9.84788239e-02 1.76721457e-02 1.34889901e-01 2.52090748e-02 -9.27632332e-01 -1.10162990e-02 -4.44754809e-01 -4.06275749e-01 4.00106966e-01 -3.72069746e-01 -4.34770495e-01 1.99411720e-01 -1.00143278e+00 7.31709242e-01 6.49369180e-01 -1.27764076e-01 -8.71110797e-01 -4.42663163e-01 -1.08505034e+00 -3.09663899e-02 4.28888440e-01 -7.72106826e-01 1.43582821e+00 -1.21145976e+00 -1.20103741e+00 9.89243329e-01 -3.18818271e-01 -5.77913642e-01 6.47316158e-01 -2.71463007e-01 -6.18926473e-02 7.55548477e-01 2.87187904e-01 1.12950015e+00 5.73369563e-01 -1.46718156e+00 -5.00482678e-01 2.48924240e-01 6.38277173e-01 5.46460032e-01 -2.11555853e-01 -1.94364592e-01 -5.46099544e-01 -6.71580076e-01 -6.16356254e-01 -7.13531494e-01 1.69205815e-01 -2.82258630e-01 -8.23476911e-01 -3.78608406e-01 1.02670169e+00 -6.94959521e-01 1.40907156e+00 -1.88825142e+00 3.07246476e-01 -4.58161533e-01 3.27700764e-01 -6.01485698e-03 -1.19447008e-01 7.67703950e-01 1.04763500e-01 2.57477522e-01 8.88176709e-02 -2.45991752e-01 5.31195141e-02 -2.67925598e-02 -5.29444277e-01 3.56480807e-01 5.70144542e-02 9.61743116e-01 -9.38430190e-01 -1.14820862e+00 3.98380548e-01 5.28041899e-01 -4.05249834e-01 4.36794758e-01 -8.86109054e-01 6.31735146e-01 -4.16907370e-01 2.71184832e-01 1.99519008e-01 -6.17259920e-01 4.56342876e-01 -5.26802421e-01 -2.60952204e-01 2.10528642e-01 -7.46169746e-01 1.90505350e+00 8.13409612e-02 1.50145733e+00 -3.36726844e-01 -9.02824342e-01 5.30430198e-01 8.26689482e-01 4.85934079e-01 -6.85108066e-01 5.82314022e-02 -5.32032788e-01 -3.89098525e-01 -1.18794501e+00 9.29411888e-01 1.84457541e-01 -2.29837343e-01 6.22342110e-01 -1.14978768e-01 -4.02264595e-02 7.43643880e-01 7.42793143e-01 7.43522286e-01 5.20394504e-01 5.06035984e-01 -5.16098216e-02 5.66889226e-01 5.78018248e-01 -7.46990740e-02 6.72697902e-01 -2.35021651e-01 7.74737239e-01 9.94361043e-01 -3.04290324e-01 -1.43349183e+00 -7.53732443e-01 2.55840689e-01 8.11045408e-01 3.92425150e-01 -6.98788166e-01 -9.25023317e-01 -3.88156086e-01 -3.91032487e-01 8.12346816e-01 -6.85264051e-01 5.23080707e-01 -8.07457030e-01 -4.25998047e-02 4.51848030e-01 2.47313172e-01 6.38608277e-01 -1.25043356e+00 -8.38161230e-01 3.33464332e-02 -1.00200343e+00 -1.72125459e+00 -4.18085307e-01 -8.23454022e-01 -4.38031197e-01 -1.34871495e+00 -7.82075346e-01 -9.27269220e-01 6.79316878e-01 4.29478109e-01 1.37978399e+00 2.23448694e-01 3.03221084e-02 7.26847589e-01 -6.01964831e-01 -1.29674911e-01 -8.29525471e-01 -3.74411583e-01 -3.44600648e-01 -3.01721603e-01 5.83341494e-02 -1.51878133e-01 -6.52392387e-01 6.09025359e-02 -9.69059110e-01 9.45249438e-01 2.88362708e-02 5.45125902e-01 2.71674186e-01 -3.95354092e-01 4.08518195e-01 -1.05155396e+00 9.24652100e-01 -6.22527838e-01 -8.20859745e-02 4.97792602e-01 8.75370502e-02 -8.51870030e-02 2.63159096e-01 -3.43147099e-01 -1.30365896e+00 -2.42152333e-01 3.17953199e-01 -4.12091583e-01 -1.41265720e-01 4.88225967e-01 2.32633799e-01 7.99634576e-01 4.84101236e-01 -1.13394707e-01 -1.50271878e-01 2.04718411e-01 8.32519531e-01 5.16146004e-01 9.23518717e-01 -5.49368620e-01 3.44267964e-01 4.94251430e-01 -1.11690380e-01 -1.06995034e+00 -8.16001713e-01 -4.42586988e-01 -3.60448450e-01 -8.75519097e-01 1.30778670e+00 -1.23449755e+00 -9.54574645e-01 -1.63003936e-01 -1.72072697e+00 -9.29662511e-02 -5.46760373e-02 3.52476716e-01 -9.81989503e-01 7.66618013e-01 -4.40656692e-01 -6.55089557e-01 -2.86492735e-01 -1.02419412e+00 1.04744816e+00 1.45622760e-01 -8.53871703e-01 -1.01220036e+00 2.24395066e-01 6.70137405e-01 -3.24366912e-02 9.99958754e-01 7.32138455e-01 -3.96475643e-01 -5.44285119e-01 2.86423773e-01 -2.90686876e-01 -3.56208421e-02 1.10812515e-01 4.38542396e-01 -8.53126347e-01 1.29736885e-02 -2.29992773e-02 -4.74715471e-01 5.39760113e-01 5.21070063e-01 7.18628168e-01 -8.92110407e-01 -3.10888350e-01 -1.41561672e-01 1.30444610e+00 1.92883745e-01 7.38601625e-01 4.19783860e-01 6.85675144e-01 7.51250744e-01 8.59072447e-01 4.99769449e-01 5.56336224e-01 6.08940780e-01 4.73982871e-01 3.68598104e-02 -5.30385435e-01 -4.31415230e-01 4.43712860e-01 9.22268271e-01 -2.80917406e-01 -8.10368657e-01 -9.13174033e-01 7.17541337e-01 -2.34533691e+00 -1.78170848e+00 -3.21268052e-01 1.45650458e+00 7.58960783e-01 -2.00938229e-02 3.00122261e-01 -1.36571363e-01 9.63137031e-01 4.92025256e-01 -1.08026922e-01 -3.45115215e-01 -2.17641324e-01 -6.55154407e-01 -2.52879024e-01 5.85013807e-01 -1.05367947e+00 7.33366609e-01 6.59977579e+00 4.47469413e-01 -8.01256180e-01 -8.09438154e-02 6.64163291e-01 -1.47356898e-01 -3.78888786e-01 -1.54335812e-01 -4.46608186e-01 5.10981083e-01 7.81669378e-01 -5.73514223e-01 2.99763568e-02 6.49531543e-01 6.77149773e-01 -4.29695994e-01 -1.36729395e+00 8.77026260e-01 5.13304889e-01 -2.09157205e+00 3.79049748e-01 -4.75491792e-01 1.00900090e+00 -5.87134182e-01 -1.41536698e-01 1.20780379e-01 6.07481110e-04 -1.12984419e+00 1.07257771e+00 6.12577260e-01 6.49703979e-01 -4.20760810e-01 6.34356081e-01 2.82542020e-01 -1.05011439e+00 3.35881412e-01 -1.63994685e-01 -2.29687020e-01 5.33287406e-01 1.22825131e-01 -1.33415174e+00 5.44882655e-01 4.62867975e-01 1.06408405e+00 -5.23565471e-01 7.08926678e-01 -2.90197581e-01 3.87207419e-01 3.13947767e-01 -3.21636915e-01 3.13126355e-01 5.66371940e-02 7.99029946e-01 1.53316903e+00 1.60101384e-01 1.87003389e-01 9.84770507e-02 8.14636827e-01 -1.90547496e-01 4.55307066e-02 -9.99055445e-01 -5.96316397e-01 4.73988533e-01 9.42300558e-01 -7.98980892e-01 -7.05277264e-01 -5.66509902e-01 5.76754212e-01 1.79882273e-01 4.45394218e-01 -1.11413527e+00 2.22080108e-02 2.25834578e-01 8.20494071e-02 2.83586621e-01 -2.04448909e-01 -1.53046295e-01 -1.21527088e+00 7.07388371e-02 -8.92364025e-01 1.81124434e-01 -1.43693554e+00 -8.75631511e-01 8.25070381e-01 4.83041584e-01 -1.42718959e+00 -6.55904949e-01 -7.84523934e-02 -7.05143750e-01 2.06397563e-01 -1.22073174e+00 -1.10859919e+00 -7.01623797e-01 4.40381348e-01 1.16598296e+00 -7.71559477e-02 6.01458907e-01 -7.65797123e-02 -2.39034787e-01 -1.25408992e-01 -3.39252502e-01 1.24259666e-01 7.62761176e-01 -9.52970505e-01 6.05554134e-03 7.63144493e-01 2.47384131e-01 3.38081747e-01 1.45210099e+00 -7.21465230e-01 -1.01251006e+00 -8.16085994e-01 8.54834020e-01 -4.83591110e-01 7.19317198e-01 -1.59325406e-01 -7.41440594e-01 9.27877426e-01 1.38198495e+00 -7.57753432e-01 7.20609784e-01 -3.66879225e-01 -1.42473876e-01 5.38710713e-01 -7.35807657e-01 7.66109288e-01 9.61548448e-01 -4.64622051e-01 -1.19325554e+00 5.38154364e-01 1.06183076e+00 -4.75599229e-01 -7.44752586e-01 1.35663643e-01 5.11456847e-01 -1.15133464e+00 9.27311838e-01 -6.01541102e-01 1.45876265e+00 -4.11817193e-01 -1.30620196e-01 -8.36991131e-01 7.99952149e-02 -5.87706268e-01 -1.93759635e-01 1.60364318e+00 1.34337112e-01 3.97701740e-01 6.74887717e-01 4.70177680e-01 -2.05423206e-01 -5.60298488e-02 -5.68757415e-01 -3.49227309e-01 -2.96571404e-01 -2.02459246e-01 2.95604676e-01 1.12435460e+00 6.24853134e-01 7.46450961e-01 -6.95974231e-01 -2.19133943e-01 4.20566827e-01 2.57005304e-01 1.04963267e+00 -6.91607714e-01 -5.10531366e-02 -5.03019214e-01 -3.55060607e-01 -9.48767304e-01 4.06225532e-01 -6.56428099e-01 1.72564521e-01 -2.03079700e+00 6.48911953e-01 4.36526775e-01 5.06450832e-01 5.33694625e-02 -6.30858988e-02 1.80110559e-01 4.87422675e-01 4.89630371e-01 -1.27231276e+00 4.21296209e-01 1.63117790e+00 -4.04065490e-01 -8.27921852e-02 -7.45492399e-01 -5.15030503e-01 7.42898345e-01 4.01237339e-01 -1.72832370e-01 -6.22111320e-01 -4.70086575e-01 2.56632447e-01 5.49819767e-01 4.21443135e-01 -8.08369994e-01 3.82283419e-01 -3.57936621e-01 1.15733929e-01 -8.97713900e-01 3.37296963e-01 -4.10101801e-01 4.56117690e-01 2.14822546e-01 -8.09662521e-01 1.53781652e-01 1.24773525e-01 5.15894592e-01 -6.53916538e-01 -7.96096325e-02 3.00981700e-01 -3.01178128e-01 -8.06654692e-01 -3.28737170e-01 -6.71703935e-01 2.32193410e-01 1.37527645e+00 -2.64134854e-01 -1.07161164e+00 -8.60479653e-01 -8.47113967e-01 4.16595072e-01 5.31745315e-01 6.55399144e-01 6.94890678e-01 -1.52821028e+00 -8.95483851e-01 -5.18782794e-01 7.08508790e-02 -1.21342100e-01 1.99884653e-01 6.06933236e-01 -9.32276368e-01 7.79341996e-01 -5.36210179e-01 -8.76792669e-01 -1.61500049e+00 6.32211506e-01 -3.10771048e-01 -5.93423657e-02 -8.30369234e-01 4.21243638e-01 5.29211760e-01 4.89652693e-01 5.17463028e-01 -2.65626937e-01 -7.81003118e-01 2.88510919e-01 7.97596872e-01 1.98369473e-01 -7.23871350e-01 -1.04148352e+00 -1.22507334e-01 3.83351058e-01 2.03527316e-01 -2.01897457e-01 1.29577065e+00 -6.08419597e-01 -4.93265651e-02 7.02734172e-01 1.06004977e+00 -8.26471597e-02 -1.11161089e+00 1.69021666e-01 6.38782829e-02 -2.27918118e-01 -4.52199459e-01 -4.75459337e-01 -4.40789640e-01 7.31584668e-01 -3.16264592e-02 5.83852708e-01 8.43915284e-01 2.56355733e-01 4.56494689e-01 2.82411695e-01 -1.01981284e-02 -8.51248324e-01 5.86022496e-01 2.45584399e-01 1.42953217e+00 -1.40855753e+00 3.46258551e-01 -4.30683464e-01 -1.14021540e+00 1.36132228e+00 5.24127364e-01 4.96763140e-02 -2.39230739e-03 -3.24815251e-02 8.70238617e-03 -4.19997185e-01 -1.09342551e+00 -2.87718792e-02 3.63712043e-01 5.26714087e-01 7.86293387e-01 -4.80300970e-02 -5.35858274e-01 2.91552693e-01 -1.56253070e-01 1.10419720e-01 1.03692281e+00 8.19005370e-01 -6.05052352e-01 -6.26084030e-01 -4.55697834e-01 -1.40523568e-01 -2.99230516e-01 2.21265510e-01 -5.61769843e-01 1.07458389e+00 -2.65317678e-01 1.12542272e+00 3.95262063e-01 4.36682906e-03 5.08489832e-02 -9.32435319e-02 4.81482416e-01 -5.90314627e-01 -4.13966566e-01 -5.75192831e-03 7.04450071e-01 -5.05606890e-01 -1.39792311e+00 -5.16256690e-01 -1.38115990e+00 -4.53521281e-01 3.02220911e-01 2.28653640e-01 4.48857218e-01 9.20962870e-01 1.73167080e-01 6.37822509e-01 2.01350540e-01 -7.77850211e-01 3.81298810e-01 -9.24643040e-01 7.17398226e-02 8.11433792e-01 3.52379888e-01 -4.86750543e-01 -1.71205014e-01 8.54025483e-01]
[10.851067543029785, 0.8148797750473022]
951b42b4-4ed5-48d5-8b9d-808dffbf7a7b
handling-and-mining-linguistic-variation-in
null
null
https://aclanthology.org/W15-5402
https://aclanthology.org/W15-5402.pdf
Handling and Mining Linguistic Variation in UGC
null
['Leon Derczynski']
2015-09-01
null
null
null
ws-2015-9
['rumour-detection']
['natural-language-processing']
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
[-7.461014270782471, 3.6235029697418213]
aee7032e-33e6-47df-baed-876dc863ddd2
rsa-inr-riemannian-shape-autoencoding-via-4d
2305.12854
null
https://arxiv.org/abs/2305.12854v1
https://arxiv.org/pdf/2305.12854v1.pdf
RSA-INR: Riemannian Shape Autoencoding via 4D Implicit Neural Representations
Shape encoding and shape analysis are valuable tools for comparing shapes and for dimensionality reduction. A specific framework for shape analysis is the Large Deformation Diffeomorphic Metric Mapping (LDDMM) framework, which is capable of shape matching and dimensionality reduction. Researchers have recently introduced neural networks into this framework. However, these works can not match more than two objects simultaneously or have suboptimal performance in shape variability modeling. The latter limitation occurs as the works do not use state-of-the-art shape encoding methods. Moreover, the literature does not discuss the connection between the LDDMM Riemannian distance and the Riemannian geometry for deep learning literature. Our work aims to bridge this gap by demonstrating how LDDMM can integrate Riemannian geometry into deep learning. Furthermore, we discuss how deep learning solves and generalizes shape matching and dimensionality reduction formulations of LDDMM. We achieve both goals by designing a novel implicit encoder for shapes. This model extends a neural network-based algorithm for LDDMM-based pairwise registration, results in a nonlinear manifold PCA, and adds a Riemannian geometry aspect to deep learning models for shape variability modeling. Additionally, we demonstrate that the Riemannian geometry component improves the reconstruction procedure of the implicit encoder in terms of reconstruction quality and stability to noise. We hope our discussion paves the way to more research into how Riemannian geometry, shape/image analysis, and deep learning can be combined.
['Christoph Brune', 'Nicola Strisciuglio', 'Sven Dummer']
2023-05-22
null
null
null
null
['dimensionality-reduction']
['methodology']
[-3.48798841e-01 5.82068972e-02 8.91082957e-02 -4.39048022e-01 -4.62569207e-01 -6.44272447e-01 6.88790798e-01 -2.65936911e-01 -1.25383779e-01 5.28653618e-03 2.22063258e-01 -1.36650398e-01 -4.70926344e-01 -1.00292921e+00 -6.09577835e-01 -7.11752534e-01 -3.37249562e-02 5.05438328e-01 -3.33503842e-01 -9.36539322e-02 3.57374847e-02 1.04927289e+00 -1.25235641e+00 -1.55313596e-01 3.82774174e-01 7.42921233e-01 1.73461786e-03 3.16133350e-01 -9.83771756e-02 2.41267160e-02 1.14356816e-01 -5.54986298e-01 5.36252856e-01 -1.82675317e-01 -8.87707114e-01 1.34196915e-02 7.36400962e-01 -2.56035358e-01 -7.50114381e-01 9.37150419e-01 6.37010217e-01 6.02588430e-02 8.53860617e-01 -1.21323276e+00 -1.36477804e+00 1.68452874e-01 -1.52809665e-01 -4.92583513e-02 -2.35131115e-01 -5.96325174e-02 8.93231630e-01 -1.33544922e+00 5.59311211e-01 1.37657917e+00 8.51004779e-01 5.63965619e-01 -1.09542716e+00 -1.12013310e-01 -3.34661245e-01 7.46451691e-02 -1.46355546e+00 -2.54156679e-01 1.03856337e+00 -5.78500211e-01 7.69091308e-01 2.50259399e-01 5.48812807e-01 7.58968651e-01 1.56336218e-01 6.22561336e-01 7.30970502e-01 -9.13505554e-02 -5.32282144e-03 -2.57295579e-01 -1.00309767e-01 7.63176501e-01 2.65043527e-01 3.29880685e-01 3.89069435e-03 -6.23579230e-03 1.24508381e+00 1.72687054e-01 2.32975297e-02 -7.29086339e-01 -1.24058545e+00 9.43381071e-01 5.21588027e-01 6.24518573e-01 2.95845475e-02 2.37526134e-01 9.19456780e-02 4.54346597e-01 5.14580965e-01 4.34126198e-01 -2.73269653e-01 -1.24376535e-01 -6.95243418e-01 1.72474355e-01 6.43829465e-01 9.96735156e-01 7.47043729e-01 1.52065665e-01 -2.62068100e-02 9.14405584e-01 6.64682806e-01 6.89657092e-01 2.22085625e-01 -1.27059484e+00 2.31826484e-01 7.58439660e-01 -4.40180480e-01 -1.50288129e+00 -7.63724625e-01 -2.51205981e-01 -1.08806288e+00 2.84153372e-01 3.85249436e-01 1.90014586e-01 -3.58145207e-01 1.94199479e+00 2.10810140e-01 2.20040098e-01 -2.18553767e-01 1.10037589e+00 9.00089502e-01 4.32075746e-02 -3.49384457e-01 3.43836606e-01 1.31986284e+00 -6.55999005e-01 -5.68099439e-01 3.49324375e-01 9.44701493e-01 -7.15474665e-01 9.94303703e-01 -1.55217633e-01 -1.36412203e+00 -4.44503605e-01 -1.05148542e+00 -4.90048528e-01 -4.29711014e-01 2.95809239e-01 7.18838990e-01 7.63983846e-01 -1.34215212e+00 1.00667357e+00 -1.09637630e+00 -3.63783479e-01 6.15995526e-01 5.00345528e-01 -5.45848429e-01 3.13590094e-02 -8.64213824e-01 1.02219272e+00 -2.88570762e-01 2.69619703e-01 -6.08774662e-01 -8.32035661e-01 -1.12826324e+00 -1.06532186e-01 -2.53059596e-01 -9.80048895e-01 5.96211255e-01 -2.79951006e-01 -1.53436124e+00 1.16631484e+00 1.01450067e-02 -1.73777252e-01 3.38496625e-01 1.00884922e-01 -1.81738019e-01 -2.71386076e-02 -2.50063956e-01 7.25835025e-01 7.11806953e-01 -9.58606482e-01 2.65516043e-01 -8.02350402e-01 9.25952718e-02 9.56707746e-02 -4.06637102e-01 -2.32743278e-01 -2.71071702e-01 -7.72050738e-01 4.67372417e-01 -1.08286440e+00 -9.79277566e-02 3.57002527e-01 3.79752442e-02 -1.57910869e-01 9.04112637e-01 -6.29256606e-01 8.01744998e-01 -2.11696935e+00 4.65812057e-01 1.58228740e-01 2.74037063e-01 2.41920635e-01 -7.56613553e-01 3.70584488e-01 -1.93971649e-01 3.12387437e-01 -5.14948130e-01 -7.81217694e-01 2.98101366e-01 3.81865174e-01 -1.47068173e-01 8.68743122e-01 4.55322653e-01 1.45584047e+00 -6.69731498e-01 -1.19719051e-01 2.81668097e-01 9.90588903e-01 -7.69376099e-01 4.20474112e-02 2.93559223e-01 7.51898408e-01 -1.26424372e-01 3.78992736e-01 1.09063411e+00 -5.82545176e-02 -8.84556919e-02 -5.13910592e-01 -6.17029704e-02 2.23070055e-01 -1.12684417e+00 2.08843279e+00 -4.09503311e-01 5.34685254e-01 6.81623742e-02 -1.16251683e+00 1.08784592e+00 1.69810310e-01 7.85328507e-01 -5.08222342e-01 3.22149009e-01 2.81178504e-01 6.43326044e-02 -4.45157558e-01 2.84435898e-01 -2.74670427e-03 3.85231674e-01 6.22627676e-01 1.00051470e-01 -4.16921198e-01 -1.59808904e-01 -9.56447944e-02 8.31341505e-01 3.20756972e-01 -1.31693989e-01 -4.19626862e-01 6.78566873e-01 -6.48421764e-01 2.82710880e-01 1.43358916e-01 -1.39073089e-01 9.41142619e-01 2.19796553e-01 -5.80972970e-01 -1.32417595e+00 -1.31577742e+00 -5.76670110e-01 6.80143297e-01 -1.05150864e-01 -2.59267896e-01 -8.91418576e-01 -4.14064527e-01 1.58838704e-01 2.33219460e-01 -5.73222280e-01 -4.49726820e-01 -7.71257281e-01 -8.47120583e-01 7.24514604e-01 5.44504583e-01 2.73062974e-01 -8.48596096e-01 7.86753744e-02 -6.80841506e-02 -2.25297585e-02 -1.02118814e+00 -8.11428249e-01 -4.69350100e-01 -1.25049996e+00 -1.15682423e+00 -5.55251718e-01 -1.02571297e+00 7.81606138e-01 3.20123166e-01 9.61498916e-01 2.35556751e-01 -3.67408246e-01 8.92554581e-01 -4.97590825e-02 -3.22754309e-02 -5.45641899e-01 4.69667204e-02 2.50284016e-01 2.03626335e-01 5.45118868e-01 -1.15557194e+00 -5.58912694e-01 4.43317950e-01 -1.05869985e+00 -2.29173794e-01 4.66111392e-01 6.07296765e-01 5.63807487e-01 -4.04330969e-01 4.87978339e-01 -2.36900583e-01 5.64151645e-01 -2.72258222e-01 -5.02559900e-01 2.66892500e-02 -6.64599359e-01 3.37048680e-01 3.12373519e-01 -1.30585045e-01 -4.53188688e-01 -5.11419848e-02 -2.91239440e-01 -7.31818259e-01 -7.29126781e-02 2.92289257e-01 -2.53919065e-01 -6.05292261e-01 4.64029908e-01 2.85400003e-01 6.94428802e-01 -8.17393064e-01 6.78016186e-01 4.06038195e-01 2.93751806e-01 -6.38378918e-01 9.91627634e-01 8.21215153e-01 4.87189054e-01 -7.52071559e-01 -2.60359138e-01 -2.18746006e-01 -1.26227272e+00 1.38101697e-01 8.72402728e-01 -6.82975113e-01 -8.82296085e-01 5.12264431e-01 -1.27216661e+00 -9.26565528e-02 -4.07570064e-01 4.98590410e-01 -9.81943071e-01 9.57917929e-01 -6.84944928e-01 -5.18882513e-01 -3.29077572e-01 -1.20963752e+00 1.22684455e+00 -2.45900899e-01 -3.65968496e-02 -1.43402326e+00 2.46008426e-01 2.44961962e-01 6.70303166e-01 2.09621876e-01 9.73523080e-01 -3.73097003e-01 -6.52427495e-01 -7.85020217e-02 -2.53522366e-01 4.33091640e-01 3.99321824e-01 -3.43115717e-01 -8.79576504e-01 -5.07439137e-01 4.59700972e-01 1.00088254e-01 6.96012616e-01 3.78392696e-01 1.29049325e+00 -1.58649728e-01 4.32377383e-02 1.12170744e+00 1.22661138e+00 -1.37201831e-01 7.98804462e-01 7.72020668e-02 1.15624142e+00 7.98100233e-01 3.99201810e-02 1.72795296e-01 7.51767993e-01 9.90922272e-01 6.14237666e-01 -1.10356614e-01 -3.96684885e-01 -7.00773485e-03 2.30076671e-01 1.37386143e+00 -3.35714519e-01 5.25774658e-01 -7.56369174e-01 5.06964564e-01 -1.73850513e+00 -7.24693358e-01 -2.38708168e-01 2.36259198e+00 5.28024077e-01 -6.64701521e-01 2.75027268e-02 1.13833817e-02 5.90213895e-01 3.37095484e-02 -4.11739707e-01 -3.79255801e-01 -4.44438219e-01 1.18312493e-01 2.30933025e-01 5.96483052e-01 -1.20949662e+00 8.05053055e-01 6.10272169e+00 5.23017824e-01 -9.50155854e-01 2.62655884e-01 2.80185789e-01 1.79593608e-01 -5.79602063e-01 -1.93807513e-01 -4.31453049e-01 1.48610786e-01 8.42898011e-01 -1.26763117e-02 8.07511985e-01 7.31472790e-01 2.68507689e-01 7.70123959e-01 -1.47109628e+00 1.42080593e+00 2.94895291e-01 -1.38158607e+00 2.47095957e-01 5.60509861e-01 5.75892746e-01 1.78093225e-01 2.64295280e-01 6.23784885e-02 -1.89481899e-01 -1.30315304e+00 3.03288341e-01 7.24655509e-01 7.06042230e-01 -6.29570365e-01 4.96205032e-01 5.25919087e-02 -1.43217087e+00 2.04049751e-01 -8.39343786e-01 1.08786307e-01 5.17974310e-02 3.89878184e-01 -4.29586977e-01 7.90631831e-01 3.20171416e-01 1.11613739e+00 -5.67053854e-01 9.11252379e-01 7.26642236e-02 -1.72959473e-02 -1.47111371e-01 3.89882386e-01 1.25656143e-01 -9.91688848e-01 8.51118863e-01 1.01031554e+00 4.33209181e-01 -1.21447429e-01 -9.13516209e-02 1.58922946e+00 -5.36242761e-02 2.38673791e-01 -1.02554107e+00 -6.46607280e-02 1.77229986e-01 1.42136705e+00 -4.43149120e-01 1.53461963e-01 -4.99308020e-01 8.43574405e-01 4.59699452e-01 2.01102942e-01 -6.13026798e-01 -2.20335294e-02 1.19661939e+00 1.11008406e-01 2.14581028e-01 -7.50482857e-01 -6.38295293e-01 -1.29887664e+00 1.38605237e-01 -6.50612593e-01 6.01620115e-02 -4.47148919e-01 -1.65545940e+00 3.64756733e-01 -1.98409647e-01 -1.33592439e+00 3.63255665e-02 -1.00042319e+00 -5.61215460e-01 8.44677091e-01 -1.46278882e+00 -1.31602657e+00 -1.44945115e-01 6.21126354e-01 1.67016998e-01 -2.79467076e-01 1.00484443e+00 6.17972612e-01 -3.13635051e-01 6.79725170e-01 1.11781538e-01 5.16353130e-01 4.65766430e-01 -1.28490138e+00 7.58510113e-01 5.90355515e-01 3.67303848e-01 8.98383439e-01 1.39473332e-02 -3.64282548e-01 -1.96572161e+00 -1.27942359e+00 7.84836888e-01 -9.72773910e-01 5.35012841e-01 -3.45444649e-01 -1.01900244e+00 6.04981720e-01 -1.45113707e-01 1.16415322e-01 7.51280963e-01 -7.67773315e-02 -4.51782405e-01 1.37638912e-01 -1.18299854e+00 7.24870443e-01 1.52758074e+00 -8.83609653e-01 -5.84414899e-01 2.85242140e-01 6.88327312e-01 -2.65307009e-01 -1.45260489e+00 4.72916484e-01 5.31853497e-01 -7.24605381e-01 1.17486441e+00 -8.47373366e-01 1.72459185e-01 -2.80939847e-01 -4.56539661e-01 -1.20676541e+00 -5.90028286e-01 -4.63612527e-01 -3.73478830e-01 1.14092696e+00 6.08261749e-02 -6.45225346e-01 7.17526734e-01 7.41775334e-01 -4.57804680e-01 -8.82968068e-01 -1.15257156e+00 -1.16362953e+00 8.16784143e-01 -6.86400950e-01 8.72625411e-01 1.15649259e+00 -2.33987391e-01 -9.75267440e-02 -2.45245159e-01 1.27815753e-01 6.81150138e-01 1.68695584e-01 7.78806210e-01 -1.33305824e+00 -2.48424243e-02 -8.53513300e-01 -8.95475745e-01 -1.04052114e+00 3.31514537e-01 -1.65280366e+00 -5.53310931e-01 -1.39933765e+00 1.48534864e-01 -3.49689901e-01 -3.18290554e-02 2.83669829e-01 3.10289502e-01 4.48235452e-01 3.44955713e-01 3.03981900e-01 -5.65948784e-02 9.57573533e-01 1.51709867e+00 -2.07457036e-01 -9.32578370e-02 -2.07964018e-01 -5.96448779e-01 6.78545117e-01 6.35565221e-01 -2.48915032e-01 -2.33543202e-01 -6.63371146e-01 1.86516434e-01 -4.93742347e-01 5.50218165e-01 -7.13731050e-01 1.06913999e-01 2.62333870e-01 1.59102187e-01 -3.87540549e-01 3.23626935e-01 -9.81710851e-01 1.85746208e-01 2.69412160e-01 -3.36837471e-02 1.21992201e-01 4.37874086e-02 3.81674409e-01 -5.76401986e-02 -1.20341264e-01 8.91492426e-01 6.69885203e-02 -1.86968148e-01 8.72172475e-01 -2.55376324e-02 9.47714746e-02 6.02666855e-01 -3.29021156e-01 -6.54467866e-02 -3.38935167e-01 -9.71253097e-01 -1.17550075e-01 5.39421976e-01 7.90677607e-01 6.86532497e-01 -2.09800529e+00 -7.87717402e-01 5.09391606e-01 -1.63789764e-01 -1.46585405e-01 1.37673944e-01 1.37534940e+00 -3.13522995e-01 4.21984166e-01 -1.87471896e-01 -8.39959323e-01 -9.19482648e-01 5.64788878e-01 7.56283522e-01 -1.79687701e-02 -6.65916920e-01 4.30130213e-01 3.92892957e-01 -1.16676962e+00 5.98253608e-02 -3.89536440e-01 -1.76527023e-01 -9.77287963e-02 3.63193631e-01 5.89781106e-01 2.88928092e-01 -1.02423835e+00 -4.33227181e-01 1.34719491e+00 5.03481507e-01 6.52633309e-02 1.53522909e+00 -2.91780829e-01 -5.05310416e-01 3.01286191e-01 1.73057997e+00 -1.89510062e-01 -1.11687148e+00 -2.55964160e-01 -4.28068973e-02 -3.77316087e-01 -1.13125995e-01 -9.06812102e-02 -1.43982732e+00 1.14559996e+00 6.37484014e-01 8.58915895e-02 8.31548452e-01 -1.84800718e-02 7.37956166e-01 3.69972140e-01 3.91747236e-01 -8.15775752e-01 1.33389354e-01 9.20444667e-01 1.39373219e+00 -1.23464644e+00 -2.47567624e-01 -3.63773108e-01 -1.91426829e-01 1.38842118e+00 2.97687799e-01 -4.84491229e-01 1.17367685e+00 1.40169472e-01 -5.40681407e-02 -4.63345855e-01 -1.57374442e-01 -1.68891907e-01 8.64869833e-01 7.01914370e-01 5.41221857e-01 -4.33242954e-02 -4.40605402e-01 5.32375395e-01 -4.60252494e-01 -3.89706999e-01 2.28535473e-01 4.03508574e-01 -7.51255751e-02 -1.47369504e+00 -2.86144733e-01 1.71310157e-01 -2.35255465e-01 -7.19063133e-02 -4.25806105e-01 6.32389307e-01 7.42365271e-02 5.24910450e-01 2.78037488e-01 -4.22264129e-01 2.57332385e-01 8.72857049e-02 8.20789754e-01 -4.16888148e-01 -2.53083736e-01 -7.32535794e-02 -4.42504317e-01 -7.10931122e-01 -4.91822064e-01 -6.97011530e-01 -1.18536663e+00 -4.57438409e-01 2.41533797e-02 -2.14633897e-01 7.84820795e-01 9.71315801e-01 7.93866336e-01 1.22552477e-01 7.23287225e-01 -1.06279600e+00 -5.69982231e-01 -7.98825681e-01 -6.46911263e-01 7.64634609e-01 8.03809240e-02 -7.86388874e-01 -3.63944441e-01 -1.45256072e-01]
[8.210700988769531, -3.2123796939849854]
1b486c60-3c14-489f-bff7-84ef57cadfc3
deep-discourse-analysis-for-generating
2103.07785
null
https://arxiv.org/abs/2103.07785v1
https://arxiv.org/pdf/2103.07785v1.pdf
Deep Discourse Analysis for Generating Personalized Feedback in Intelligent Tutor Systems
We explore creating automated, personalized feedback in an intelligent tutoring system (ITS). Our goal is to pinpoint correct and incorrect concepts in student answers in order to achieve better student learning gains. Although automatic methods for providing personalized feedback exist, they do not explicitly inform students about which concepts in their answers are correct or incorrect. Our approach involves decomposing students answers using neural discourse segmentation and classification techniques. This decomposition yields a relational graph over all discourse units covered by the reference solutions and student answers. We use this inferred relational graph structure and a neural classifier to match student answers with reference solutions and generate personalized feedback. Although the process is completely automated and data-driven, the personalized feedback generated is highly contextual, domain-aware and effectively targets each student's misconceptions and knowledge gaps. We test our method in a dialogue-based ITS and demonstrate that our approach results in high-quality feedback and significantly improved student learning gains.
['Jackie C. K. Cheung', 'François St-Hilaire', 'Iulian V. Serban', 'Ekaterina Kochmar', 'Robert Belfer', 'Matt Grenander']
2021-03-13
null
null
null
null
['misconceptions', 'discourse-segmentation']
['miscellaneous', 'natural-language-processing']
[ 4.89832789e-01 9.70968008e-01 -3.89030576e-01 -3.82587820e-01 -8.47925007e-01 -8.84064853e-01 3.36025834e-01 9.38755751e-01 7.90085122e-02 9.51042533e-01 7.00044572e-01 -9.06425834e-01 -3.43304664e-01 -1.00111318e+00 -4.52581525e-01 1.07962534e-01 3.82992178e-01 6.41521335e-01 5.39410293e-01 -7.82272518e-01 6.03272676e-01 1.94393456e-01 -1.74330878e+00 6.21925235e-01 1.65218341e+00 2.58939981e-01 -8.25271532e-02 1.31296408e+00 -6.39646173e-01 1.54514599e+00 -1.22416699e+00 -4.50668305e-01 -4.77621883e-01 -1.05802453e+00 -1.64745927e+00 -4.36610542e-03 7.61212409e-01 -3.12607020e-01 -6.64728582e-02 8.99545491e-01 1.10938124e-01 3.69535983e-01 6.67971671e-01 -8.91739845e-01 -9.44707870e-01 9.86429393e-01 1.47586465e-01 6.58724532e-02 1.13398993e+00 -1.76321864e-01 9.38207388e-01 -1.95110530e-01 7.32301772e-01 1.04656112e+00 5.67383349e-01 1.03015006e+00 -1.33699250e+00 -9.81115922e-02 1.21236734e-01 3.32609624e-01 -3.90333295e-01 -1.51332274e-01 6.55242383e-01 -7.44189322e-01 7.11489737e-01 7.37498760e-01 8.76245797e-01 6.21878862e-01 -3.94778028e-02 1.05408883e+00 6.81602359e-01 -8.01407933e-01 6.26096353e-02 5.23278594e-01 8.93641174e-01 7.21318245e-01 -9.03832838e-02 -3.66071582e-01 -5.55777669e-01 1.44332303e-02 2.81151026e-01 -1.99556530e-01 -6.77795708e-01 -1.40304148e-01 -6.99630499e-01 8.81237209e-01 2.96795100e-01 3.21041495e-01 -6.55080080e-02 -2.79664069e-01 2.44392887e-01 7.90672004e-01 1.06951237e-01 1.09447682e+00 -5.30609548e-01 -2.55279273e-01 -6.85447514e-01 5.96460640e-01 1.46625018e+00 8.81552160e-01 6.04168653e-01 -2.47686833e-01 -5.15406787e-01 8.56589317e-01 3.12216640e-01 2.86294892e-02 7.36520588e-01 -1.37871432e+00 5.02194047e-01 1.31949961e+00 6.49505258e-02 -1.08178723e+00 -3.91983241e-02 -1.89059377e-01 3.50066051e-02 7.78735578e-02 7.65731573e-01 -5.11332929e-01 -5.95605314e-01 1.49963057e+00 3.29559088e-01 -1.75568298e-01 5.07318914e-01 5.03149748e-01 1.41843343e+00 3.71204346e-01 9.26097110e-02 -1.24682978e-01 1.25586641e+00 -9.54586446e-01 -9.13251281e-01 -9.93905663e-02 1.38593137e+00 -7.20458806e-01 8.75379503e-01 4.11556393e-01 -1.27988338e+00 -3.37162435e-01 -8.85362923e-01 -3.01841587e-01 -1.45020455e-01 -3.09407324e-01 6.16341606e-02 7.00871885e-01 -9.90751088e-01 7.31229365e-01 -2.57611036e-01 9.52231232e-03 1.46057889e-01 1.93098322e-01 1.21849805e-01 2.41691153e-02 -1.09726167e+00 1.05218053e+00 2.85245508e-01 -4.45850432e-01 -3.09681237e-01 -1.07552791e+00 -1.09225821e+00 3.09011400e-01 4.51714694e-01 -5.49217939e-01 2.14694524e+00 -9.42626476e-01 -1.88362920e+00 6.14631534e-01 -2.26568654e-01 -4.40412343e-01 5.03096342e-01 -2.70510316e-01 2.19129890e-01 2.31734082e-01 -4.14931849e-02 6.25739872e-01 -6.70129359e-02 -1.10929441e+00 -7.92295039e-01 -2.67095268e-01 3.02550226e-01 6.65920436e-01 -2.63950825e-01 -2.74489731e-01 4.40297090e-02 -1.74686447e-01 3.89490068e-01 -4.61779952e-01 -6.98482618e-02 -7.65084267e-01 -3.53459924e-01 -8.12065482e-01 7.27492809e-01 -7.84151018e-01 1.54851151e+00 -1.21806788e+00 8.76066238e-02 1.57968521e-01 6.88656092e-01 3.65005583e-01 1.43042713e-01 5.23964286e-01 6.03864267e-02 1.74185753e-01 7.26699010e-02 2.97756255e-01 1.92815065e-01 1.21352270e-01 -4.82784241e-01 -3.46003026e-01 -3.76573466e-02 8.36695910e-01 -1.39040184e+00 -4.02564645e-01 2.07926154e-01 -3.47511354e-03 -7.52869308e-01 7.78232813e-01 -9.34625268e-01 3.32366198e-01 -5.95882833e-01 2.94320285e-01 -3.29982936e-02 -1.71712525e-02 2.57189095e-01 5.14105618e-01 9.16116014e-02 9.61866856e-01 -7.98143327e-01 1.31584048e+00 -3.04170221e-01 8.73984575e-01 -1.10375501e-01 -8.88256729e-01 1.33078122e+00 3.88367265e-01 1.55737355e-01 -7.93495595e-01 -9.09455642e-02 1.59294620e-01 7.26709440e-02 -1.00794399e+00 1.03465986e+00 1.50473431e-01 2.09212229e-02 8.29642832e-01 1.11511365e-01 -3.97406429e-01 3.74836922e-01 6.94172144e-01 1.22783518e+00 3.13079864e-01 9.75382254e-02 -1.49625525e-01 5.03018677e-01 4.46285099e-01 1.17939442e-01 7.45213926e-01 -1.28857955e-01 3.39270622e-01 8.15618098e-01 -8.46144259e-02 -4.08252060e-01 -6.58266723e-01 3.51767510e-01 1.35249460e+00 -2.14389667e-01 -5.13366282e-01 -1.10090744e+00 -9.20106769e-01 -2.98128556e-03 1.22184646e+00 -4.08703238e-01 -2.44040549e-01 -6.61001444e-01 1.50207728e-01 3.42991650e-01 3.71027559e-01 1.56947553e-01 -1.15656233e+00 -6.98443532e-01 4.67964649e-01 -4.51139987e-01 -4.76212710e-01 -3.32549572e-01 9.71207023e-02 -8.95984054e-01 -1.47475529e+00 -2.87511796e-01 -9.27845299e-01 8.40902805e-01 -1.03749672e-03 1.38548934e+00 6.24689281e-01 1.32345617e-01 8.56750011e-01 -2.26901785e-01 -3.17435503e-01 -9.35799241e-01 6.57628253e-02 -6.24357164e-01 -7.00672269e-01 7.24233747e-01 -2.01630250e-01 -4.93726671e-01 3.84267457e-02 -4.79522467e-01 9.47411284e-02 -8.86519924e-02 6.72380447e-01 6.18483275e-02 -3.76196980e-01 9.43393111e-01 -1.53848660e+00 1.47687554e+00 -4.88023460e-01 -3.59331846e-01 5.71043849e-01 -8.20256054e-01 3.05646539e-01 4.90725011e-01 -1.61690995e-01 -1.27659333e+00 -2.89846987e-01 -2.42278531e-01 1.34352535e-01 -5.59886336e-01 8.11612010e-01 1.87644452e-01 1.10167041e-01 1.18751967e+00 -3.00287921e-02 -7.07199275e-02 -2.23571539e-01 4.19792891e-01 6.82695806e-01 7.43953645e-01 -9.82762158e-01 1.42888933e-01 -7.86691427e-01 -3.88614386e-01 -6.49892271e-01 -1.02430081e+00 -3.76430154e-01 -6.12419426e-01 -7.00315595e-01 6.71206295e-01 -4.31653947e-01 -1.23301029e+00 -1.19843096e-01 -1.26051879e+00 -6.50317490e-01 -4.86884058e-01 6.22112378e-02 -3.88889343e-01 2.03076944e-01 -7.23736286e-01 -9.57583785e-01 -1.50009602e-01 -1.30838799e+00 1.97152361e-01 8.34269524e-01 -1.04193187e+00 -1.38655436e+00 2.38366634e-01 1.06817138e+00 1.13003284e-01 1.65337145e-01 1.21087527e+00 -1.40780914e+00 -4.71194178e-01 -6.76452890e-02 1.45842865e-01 2.54883082e-03 -8.27102065e-02 1.39561638e-01 -8.40867281e-01 3.99278462e-01 -1.16740830e-01 -6.09723032e-01 4.56404358e-01 2.75644995e-02 8.07254791e-01 -9.84418511e-01 -3.05512473e-02 -2.15280965e-01 8.96205008e-01 2.78989524e-01 3.11143100e-01 2.92117357e-01 6.33147359e-01 1.42558634e+00 4.23038751e-01 -1.13547161e-01 7.55363047e-01 3.07414651e-01 1.40503153e-01 7.38393366e-01 -1.62561372e-01 -4.49506491e-01 3.57275754e-01 9.18593645e-01 3.40438932e-01 -2.59362698e-01 -1.17384946e+00 9.42267358e-01 -1.91899812e+00 -8.79994571e-01 -7.18955457e-01 1.95565689e+00 1.49761677e+00 2.55537815e-02 6.36128187e-02 1.54188350e-01 3.76663208e-01 -3.13054562e-01 -3.54426384e-01 -9.53731477e-01 4.39427078e-01 4.56913412e-01 -8.24844241e-02 1.22323239e+00 -2.24271253e-01 7.74485767e-01 6.31127739e+00 2.25107506e-01 -5.81118405e-01 -3.92351627e-01 4.64360058e-01 2.61504441e-01 -8.64641368e-01 -2.73087949e-01 -8.42647493e-01 -4.35626134e-02 1.27334952e+00 -5.37743211e-01 2.45951191e-01 6.61068320e-01 -6.42360076e-02 -2.66042709e-01 -1.28780138e+00 4.99755219e-02 5.09145893e-02 -1.70198202e+00 -6.81796521e-02 -3.09466720e-01 9.60151494e-01 -6.57565415e-01 -3.82391542e-01 5.09729326e-01 1.08726513e+00 -1.09703898e+00 3.88839632e-01 6.69942796e-01 1.03112213e-01 -9.27035868e-01 5.20049751e-01 5.24889052e-01 -4.39544201e-01 -6.94508627e-02 1.47437289e-01 -3.07945520e-01 -6.21298909e-01 -7.04795271e-02 -1.61501515e+00 2.90208638e-01 2.84951508e-01 4.09599870e-01 -6.77879155e-01 8.92168164e-01 -8.21257472e-01 7.62373447e-01 1.82528585e-01 -5.69094956e-01 1.84546858e-02 -1.31625101e-01 3.89605641e-01 1.05502236e+00 8.32520425e-02 6.54300809e-01 3.83285671e-01 8.81006777e-01 -1.74085990e-01 1.13383159e-01 -4.68468636e-01 -9.04414281e-02 7.49169946e-01 9.25020158e-01 -4.01606888e-01 -4.88808453e-01 1.39574977e-02 4.74659681e-01 3.78401816e-01 5.01586556e-01 1.22000799e-01 -6.53401375e-01 3.82134885e-01 1.19490102e-01 -1.79638177e-01 2.79439718e-01 -7.29532540e-01 -7.49229789e-01 -2.43022159e-01 -1.15451264e+00 5.80543518e-01 -7.86080658e-01 -8.29312742e-01 2.04641029e-01 -1.60637230e-01 -7.09979832e-01 -8.98053527e-01 -5.03904283e-01 -9.86747324e-01 1.10669076e+00 -1.13797355e+00 -2.68501043e-01 -3.85410219e-01 2.61023611e-01 5.98016918e-01 -8.64680558e-02 1.05991137e+00 -2.00789154e-01 -4.74861920e-01 6.45611525e-01 -2.06170902e-01 1.37912199e-01 6.65699780e-01 -1.96568918e+00 -1.54308558e-01 5.69584548e-01 -2.01262534e-01 6.81787074e-01 1.20684242e+00 -9.09227133e-01 -1.15975988e+00 -6.81182444e-01 1.23570848e+00 -6.93083823e-01 7.09058702e-01 4.33353722e-01 -1.55440974e+00 6.90493226e-01 8.49628985e-01 -8.17243934e-01 1.18495798e+00 2.19370469e-01 -2.10295677e-01 4.21975523e-01 -1.05844223e+00 7.88563251e-01 5.28841317e-01 -4.34896201e-01 -1.43818426e+00 4.23594773e-01 8.56683493e-01 -9.76182222e-01 -1.16205573e+00 -1.02957804e-02 5.89564145e-02 -1.07257569e+00 5.66092610e-01 -8.97047579e-01 8.98921907e-01 6.52943598e-03 4.32542861e-01 -1.50584555e+00 3.23036164e-01 -7.59659708e-01 -3.54418814e-01 1.25738811e+00 6.52156949e-01 -3.57481897e-01 1.22205949e+00 1.35771430e+00 -6.92815244e-01 -8.78665566e-01 -2.49772444e-01 9.34678242e-02 1.99077994e-01 -2.08207428e-01 5.90491176e-01 1.16364014e+00 9.12031651e-01 7.34584272e-01 3.80398452e-01 -2.16930494e-01 4.32969421e-01 4.13755476e-01 7.07514524e-01 -1.46902096e+00 -8.71533975e-02 -8.74896169e-01 1.03221998e-01 -1.19421148e+00 2.09577575e-01 -9.08736527e-01 2.57130563e-01 -1.76096916e+00 -3.46153796e-01 -3.17660302e-01 3.82717401e-01 5.84480345e-01 -4.86586392e-01 -4.57675755e-01 -1.54993743e-01 -1.25219047e-01 -6.84088171e-01 2.13424161e-01 1.37100363e+00 -4.47176285e-02 -7.31652081e-01 2.10190907e-01 -1.08596349e+00 6.43226504e-01 7.44478285e-01 -2.27997899e-01 -6.85292900e-01 -2.31647193e-01 1.87700406e-01 6.10174775e-01 -1.74834742e-03 -5.71582735e-01 7.88804650e-01 -4.30021465e-01 2.50096649e-01 -4.91064310e-01 -3.19684684e-01 -3.73880386e-01 -5.08422136e-01 4.29037720e-01 -1.14593649e+00 -4.82125729e-02 2.55980074e-01 1.81255803e-01 -3.83617908e-01 -8.53386462e-01 3.51582289e-01 -1.33216336e-01 -3.17058593e-01 -4.40861851e-01 -5.29766977e-01 2.77450532e-01 8.71064544e-01 -3.17425102e-01 -7.98024178e-01 -6.22401297e-01 -8.46420705e-01 8.18559408e-01 6.44649863e-02 3.16221625e-01 8.18473697e-01 -1.04156160e+00 -7.04219759e-01 -1.70210283e-02 -2.90269498e-02 3.99477243e-01 5.35372458e-02 3.07204843e-01 -6.61667347e-01 7.07414150e-01 -9.75623429e-02 -5.14774084e-01 -1.75225663e+00 -5.33606298e-02 4.96367484e-01 -3.76057625e-01 -4.13083345e-01 9.59503829e-01 -4.84984875e-01 -9.82455134e-01 6.00676596e-01 -4.34685767e-01 -1.02197289e+00 4.23566312e-01 7.07867146e-01 3.86997700e-01 1.75929800e-01 -3.96766849e-02 4.81602013e-01 -1.07335366e-01 -1.64600760e-01 -1.80545934e-02 1.19036818e+00 -1.21749274e-01 6.25444204e-02 3.65451723e-01 7.47649193e-01 2.32310012e-01 -8.54016662e-01 -3.37675899e-01 6.19540572e-01 -1.99228019e-01 -3.47449422e-01 -1.16476560e+00 -4.85662222e-01 7.70727754e-01 -1.71741843e-01 8.18198740e-01 7.83375740e-01 -1.49176538e-01 5.71736515e-01 7.52727985e-01 -3.77843618e-01 -8.97835374e-01 2.69049346e-01 9.02975261e-01 8.19985211e-01 -1.02054119e+00 -2.08153516e-01 -3.70152146e-01 -5.10639608e-01 1.52296090e+00 1.36427248e+00 1.43531635e-01 -1.14654608e-01 -9.89156216e-02 3.51141393e-01 -2.28103712e-01 -1.10745382e+00 8.39213431e-02 6.20416939e-01 5.27599216e-01 1.18208754e+00 -1.76142566e-02 -1.98093966e-01 4.66753393e-01 -7.38509297e-01 -2.60953039e-01 1.05027008e+00 1.16441512e+00 -1.06839049e+00 -1.22969317e+00 -5.89982033e-01 5.37643611e-01 -2.65984923e-01 1.19819961e-01 -1.11466539e+00 3.47805917e-01 -4.65720475e-01 1.34858620e+00 -2.26781249e-01 -4.81715411e-01 5.71882248e-01 6.83373034e-01 5.81622958e-01 -1.23688138e+00 -1.34294116e+00 -4.52639788e-01 4.53188390e-01 -1.64328963e-01 -1.56103358e-01 -4.62188482e-01 -1.66674316e+00 -3.11718076e-01 -2.90857673e-01 8.93365383e-01 3.09777647e-01 1.16929495e+00 2.25442916e-01 1.11896718e+00 1.01899326e-01 -9.91942808e-02 -6.92839622e-01 -9.26667094e-01 3.15674573e-01 2.53658950e-01 5.65341771e-01 -9.97617990e-02 -3.86913382e-02 7.89096132e-02]
[11.843454360961914, 8.032219886779785]
efa2f8bb-0d03-4e6d-966f-74f00d48f123
document-level-relation-extraction-as
2106.03618
null
https://arxiv.org/abs/2106.03618v2
https://arxiv.org/pdf/2106.03618v2.pdf
Document-level Relation Extraction as Semantic Segmentation
Document-level relation extraction aims to extract relations among multiple entity pairs from a document. Previously proposed graph-based or transformer-based models utilize the entities independently, regardless of global information among relational triples. This paper approaches the problem by predicting an entity-level relation matrix to capture local and global information, parallel to the semantic segmentation task in computer vision. Herein, we propose a Document U-shaped Network for document-level relation extraction. Specifically, we leverage an encoder module to capture the context information of entities and a U-shaped segmentation module over the image-style feature map to capture global interdependency among triples. Experimental results show that our approach can obtain state-of-the-art performance on three benchmark datasets DocRED, CDR, and GDA.
['Huajun Chen', 'Luo Si', 'Fei Huang', 'Mosha Chen', 'Chuanqi Tan', 'Shumin Deng', 'Xin Xie', 'Xiang Chen', 'Ningyu Zhang']
2021-06-07
null
null
null
null
['document-level-relation-extraction']
['natural-language-processing']
[ 8.27067196e-02 4.25574332e-01 -6.65453494e-01 -4.14305151e-01 -7.16206312e-01 -6.43373728e-01 7.78980970e-01 6.28481150e-01 -5.45064993e-02 3.99013937e-01 1.05477281e-01 -2.48706847e-01 -3.73836875e-01 -1.30477810e+00 -8.31597030e-01 -6.76937625e-02 -1.34658918e-01 5.43719947e-01 2.85714388e-01 -7.35700652e-02 5.67899942e-02 3.27551544e-01 -1.04512644e+00 6.39440596e-01 9.32892442e-01 1.34554517e+00 -1.16482705e-01 5.07593155e-01 -4.90714490e-01 1.40644670e+00 -1.51315093e-01 -9.64246988e-01 1.36732191e-01 -3.67596835e-01 -1.32462645e+00 2.96571016e-01 6.05673432e-01 -4.23503593e-02 -4.75759685e-01 1.20451546e+00 -1.90492451e-01 -3.41388762e-01 6.20663226e-01 -1.04324543e+00 -8.31905425e-01 9.55278933e-01 -8.79935801e-01 2.94781718e-02 5.27772367e-01 -4.63905275e-01 1.77992618e+00 -9.85682845e-01 1.23668015e+00 1.09830809e+00 6.13965631e-01 -2.12175235e-01 -1.00432920e+00 -1.52484864e-01 2.42145419e-01 5.23296893e-01 -1.51664019e+00 -9.50326845e-02 7.67640948e-01 -4.09370661e-01 1.17991972e+00 1.98622003e-01 6.04942560e-01 5.42391181e-01 1.59711406e-01 1.09254718e+00 7.72789717e-01 -3.70744616e-01 -2.53649563e-01 -1.08441226e-01 5.98520517e-01 1.12123108e+00 4.25144374e-01 -5.78654706e-01 -6.58563018e-01 2.40621850e-01 6.05784237e-01 -2.33161315e-01 -1.58341244e-01 -6.80846393e-01 -1.03289616e+00 5.58020771e-01 8.79107654e-01 3.56981009e-01 -4.18675572e-01 -1.16761789e-01 4.21782911e-01 1.08948052e-01 4.42138076e-01 4.34246242e-01 -5.39956689e-01 2.42701054e-01 -8.11546981e-01 -2.02406663e-02 1.03028858e+00 1.47112858e+00 9.69263792e-01 -6.60481811e-01 -2.49958232e-01 6.10106409e-01 3.37228924e-01 7.14786425e-02 5.56235164e-02 -6.43818080e-01 1.03242242e+00 1.47629178e+00 -3.74561161e-01 -1.55634868e+00 -3.78813535e-01 -4.99614209e-01 -7.94148386e-01 -6.20536864e-01 6.56388029e-02 7.03453198e-02 -1.00430918e+00 1.07400918e+00 5.27714074e-01 4.28677769e-03 3.47839743e-02 6.32035553e-01 1.12964833e+00 4.02847826e-01 -1.64982453e-01 -1.26127362e-01 1.53602040e+00 -1.11745131e+00 -8.91969740e-01 -2.73193508e-01 8.36965084e-01 -4.88873154e-01 4.31239516e-01 -2.40200758e-02 -9.80532229e-01 -2.92410016e-01 -1.13621068e+00 -4.91592318e-01 -8.07445288e-01 2.48001397e-01 8.09392154e-01 3.51797521e-01 -8.13729882e-01 4.36946392e-01 -7.59098351e-01 -3.94993961e-01 6.30169868e-01 2.99650073e-01 -6.17647707e-01 -1.66485474e-01 -1.20060802e+00 8.48820627e-01 7.25526392e-01 2.17211366e-01 -2.00665832e-01 -5.42005956e-01 -1.25809598e+00 2.74097323e-01 9.77582097e-01 -8.73648763e-01 7.85182834e-01 -2.90533036e-01 -8.57370973e-01 9.69896495e-01 -1.70601755e-01 -5.96961021e-01 2.39979345e-02 -1.70628011e-01 -3.26455802e-01 4.99423802e-01 3.31694812e-01 3.65739852e-01 3.12150925e-01 -1.46977580e+00 -9.51342165e-01 -6.34897590e-01 2.90355265e-01 2.87354648e-01 -3.29946160e-01 -1.05601735e-01 -1.12887776e+00 -4.57655668e-01 4.24732685e-01 -7.21714079e-01 7.10202381e-02 -5.07583141e-01 -1.09188378e+00 -3.12748104e-01 9.72817183e-01 -7.34663367e-01 1.39158762e+00 -1.59542656e+00 3.43319744e-01 5.67458928e-01 6.40086532e-01 -8.30095857e-02 8.32023472e-03 4.91034269e-01 -5.77484816e-02 2.72158414e-01 -2.97013551e-01 -9.96379703e-02 -2.58408915e-02 2.67726123e-01 -1.18235409e-01 6.17820434e-02 4.51143414e-01 1.39387190e+00 -9.78449881e-01 -9.59750473e-01 -2.12214559e-01 2.79303491e-01 -4.37995464e-01 -1.78813897e-02 -2.60555267e-01 -6.46900758e-02 -5.78798950e-01 8.28926504e-01 7.27961779e-01 -5.68993986e-01 8.67636681e-01 -7.37468719e-01 2.89978027e-01 5.01494944e-01 -1.09141457e+00 1.62558198e+00 -3.57455075e-01 4.10536736e-01 -6.91799521e-02 -1.08290303e+00 9.08337116e-01 6.61673304e-03 7.90408075e-01 -7.89121270e-01 1.52922481e-01 6.03408143e-02 -2.89388418e-01 -3.82916510e-01 7.18598783e-01 3.98067832e-01 -1.34317011e-01 -1.85019581e-03 4.05787498e-01 -2.57054716e-02 6.63036346e-01 7.94440985e-01 1.32985234e+00 1.07432984e-01 4.66918647e-01 -7.91779459e-02 6.25375390e-01 8.69652778e-02 5.96823692e-01 5.57935059e-01 3.36159974e-01 3.82249266e-01 1.05207694e+00 -2.10182682e-01 -5.57143629e-01 -8.93030763e-01 3.14158536e-02 5.91221273e-01 3.09568077e-01 -1.23984849e+00 -6.22565210e-01 -1.28092611e+00 2.26377323e-02 2.51928955e-01 -6.89788461e-01 1.13155194e-01 -7.60521948e-01 -6.63054347e-01 3.36934268e-01 6.87191606e-01 6.94229424e-01 -5.83398640e-01 5.41359819e-02 3.06227691e-02 -4.07526493e-01 -1.89231098e+00 -3.80687982e-01 3.31660658e-01 -6.68232262e-01 -1.48180342e+00 2.80215964e-03 -1.08721554e+00 7.08833635e-01 6.93869963e-02 1.41189837e+00 -5.97019568e-02 -1.14559703e-01 4.25972521e-01 -3.06958467e-01 1.50085866e-01 1.40902311e-01 4.60188329e-01 -5.24695337e-01 1.48381904e-01 4.13004905e-01 -3.16050649e-01 -3.17574680e-01 1.21010765e-01 -6.93195343e-01 8.87804292e-03 6.12156689e-01 6.72460735e-01 9.75241601e-01 2.58063704e-01 2.75954455e-01 -1.50838017e+00 6.53104603e-01 -3.87805969e-01 -5.37459731e-01 7.17453480e-01 -7.80586898e-01 1.68410361e-01 4.85169739e-01 3.53820086e-01 -1.13547051e+00 2.24810004e-01 3.84555548e-01 -1.48756340e-01 1.45955041e-01 1.11333168e+00 -6.41434431e-01 1.88672021e-01 2.35750258e-01 1.03561683e-02 -7.19725013e-01 -2.31474161e-01 8.69335175e-01 5.02454817e-01 8.38270128e-01 -6.81521535e-01 9.21614408e-01 4.72364575e-01 3.54759902e-01 -4.14967299e-01 -1.41073453e+00 -7.62779415e-01 -1.22222984e+00 7.81865492e-02 1.12725568e+00 -1.06610858e+00 -5.67297637e-01 1.69158623e-01 -1.23099554e+00 1.13823235e-01 -3.51005256e-01 6.31773770e-02 -3.10082465e-01 1.77273765e-01 -8.60046208e-01 -2.13878810e-01 -2.60996997e-01 -9.44984972e-01 1.24579942e+00 2.21935913e-01 4.81988071e-03 -1.03511906e+00 1.78938825e-02 7.39018321e-01 -3.96553636e-01 2.85026908e-01 1.06613326e+00 -5.05711436e-01 -1.16970754e+00 -2.70461351e-01 -7.93997288e-01 1.20020874e-01 4.09302622e-01 -7.22347721e-02 -4.75937873e-01 2.27525532e-01 -4.43817079e-01 -1.90152586e-01 1.12727129e+00 -1.38446037e-02 1.08879459e+00 -5.70347548e-01 -7.31402099e-01 6.18862927e-01 1.50344646e+00 4.78836559e-02 5.30948162e-01 3.02149236e-01 1.57369590e+00 5.83489478e-01 6.77772343e-01 1.54904723e-01 1.15456069e+00 6.39386773e-01 2.97323495e-01 9.01821256e-02 -3.58901471e-01 -4.79518265e-01 -1.26333132e-01 1.05864549e+00 7.05651492e-02 -3.51384759e-01 -1.13286567e+00 7.92615116e-01 -2.09648252e+00 -7.63855457e-01 -5.96970022e-01 1.54227209e+00 1.07366908e+00 2.26663575e-01 -1.45597205e-01 -4.13973406e-02 6.75896704e-01 2.97572017e-01 -1.10660121e-01 -1.94121644e-01 -3.00143778e-01 1.16525561e-01 6.23219013e-01 4.89220291e-01 -1.34532428e+00 1.23567283e+00 5.47666025e+00 6.93641245e-01 -6.48857296e-01 -5.92649020e-02 5.08522272e-01 3.05762082e-01 -3.66130710e-01 2.54832566e-01 -1.00954270e+00 -3.11806593e-02 7.04827666e-01 -1.10150389e-01 4.68324907e-02 5.41477799e-01 -4.81707960e-01 -1.23854361e-01 -1.26984787e+00 7.43654788e-01 1.19056262e-01 -1.68338609e+00 1.44325301e-01 1.66200459e-01 9.37633097e-01 -7.74377286e-02 -9.66870487e-02 1.18486784e-01 5.94281912e-01 -7.87214041e-01 4.51909900e-01 6.04104102e-01 6.30805135e-01 -8.62665176e-01 6.60138607e-01 2.95267254e-02 -1.70794916e+00 1.91762373e-01 -1.16539381e-01 3.36243659e-01 -1.66062657e-02 8.52545857e-01 -1.02784026e+00 1.22215331e+00 6.32602692e-01 1.20608282e+00 -8.23323429e-01 6.50363743e-01 -4.79896307e-01 4.42429781e-01 -2.18385518e-01 2.48688087e-01 4.02838498e-01 -3.29384983e-01 3.71257722e-01 1.51870966e+00 -1.52189791e-01 -2.79557779e-02 1.68321222e-01 6.54531777e-01 -7.86865354e-01 1.25777483e-01 -6.30914688e-01 -8.19749236e-02 2.66105741e-01 1.69353259e+00 -9.37321126e-01 -3.73858780e-01 -6.83965921e-01 9.88853812e-01 7.04262137e-01 2.60947764e-01 -5.22137165e-01 -5.79882085e-01 2.15667874e-01 -2.91646402e-02 7.46149540e-01 -3.74361396e-01 -6.15591228e-01 -1.31092095e+00 3.74908656e-01 -6.68797731e-01 7.34570324e-01 -5.52220583e-01 -1.11804688e+00 4.92202669e-01 -3.54239568e-02 -9.47758853e-01 -1.33562818e-01 -6.45288587e-01 -1.90902501e-01 4.89378452e-01 -1.62355077e+00 -1.68292439e+00 -1.82928979e-01 5.37473977e-01 1.98234558e-01 -3.75859961e-02 5.66779017e-01 2.80367613e-01 -8.15201402e-01 5.24880409e-01 -1.13538630e-01 8.34469140e-01 3.80102992e-01 -1.67077124e+00 4.41467345e-01 9.07951891e-01 7.95897067e-01 8.97201061e-01 4.84643765e-02 -1.00324750e+00 -1.61301565e+00 -1.32424843e+00 1.23222029e+00 -6.96508467e-01 1.01116145e+00 -4.06240612e-01 -8.26803684e-01 1.11196756e+00 3.66631478e-01 4.82703775e-01 5.29971898e-01 5.06859541e-01 -7.59039760e-01 -2.14065135e-01 -8.76111746e-01 4.57907587e-01 1.34223747e+00 -8.93443346e-01 -4.41218615e-01 2.86143988e-01 6.62680566e-01 -6.52419209e-01 -1.40903020e+00 4.13176715e-01 4.31550980e-01 -6.53969049e-01 1.05580544e+00 -5.88461161e-01 9.52109337e-01 -3.36162955e-01 -1.44017816e-01 -1.08588827e+00 -2.76995242e-01 -3.29764187e-01 -8.34361315e-01 1.64273691e+00 6.86035573e-01 -1.73888415e-01 9.10248995e-01 3.85109186e-01 1.05765313e-01 -9.99174237e-01 -6.33316755e-01 -7.75949121e-01 -1.42203689e-01 -2.07875162e-01 6.57581508e-01 1.19948709e+00 1.74306467e-01 9.00486708e-01 -4.42646816e-02 4.53897595e-01 6.33908749e-01 6.02450550e-01 5.44766963e-01 -1.14855123e+00 -7.74275512e-02 -1.96202204e-01 -5.88110924e-01 -1.07133830e+00 4.31947917e-01 -1.33208168e+00 -2.50707030e-01 -2.04460692e+00 4.74737734e-01 -3.16293925e-01 -3.08474153e-01 5.62846899e-01 -2.03656897e-01 1.71789870e-01 2.66353816e-01 5.08036092e-02 -1.15754986e+00 5.01031220e-01 1.39105141e+00 -6.53849542e-01 -5.09592593e-02 -4.92062628e-01 -9.28248048e-01 7.48372316e-01 2.78411955e-01 -5.75618565e-01 -4.74602878e-01 -5.77837110e-01 6.37180626e-01 9.96853411e-02 2.94848323e-01 -6.56969726e-01 6.23103976e-01 -1.67819038e-01 4.38493341e-01 -8.55201364e-01 8.05941895e-02 -9.31748509e-01 -4.10945117e-01 -7.50481114e-02 -2.62648314e-01 -1.36919007e-01 -2.30477065e-01 6.82179689e-01 -6.93464637e-01 2.96711605e-02 2.09499016e-01 -9.24401507e-02 -7.53170788e-01 4.91641581e-01 2.35272363e-01 2.91565061e-01 8.15478086e-01 9.56966579e-02 -6.32777870e-01 -9.41853747e-02 -7.21056521e-01 5.51791906e-01 4.64330688e-02 5.51336169e-01 5.75799644e-01 -1.43177056e+00 -4.65414375e-01 -1.91424444e-01 4.05066282e-01 3.33084702e-01 -4.37714122e-02 7.32666314e-01 -3.53146106e-01 7.22259462e-01 1.82414800e-01 -5.16466379e-01 -1.31611574e+00 4.98797059e-01 2.96924412e-01 -9.41221178e-01 -4.44364101e-01 9.56255436e-01 1.16433300e-01 -5.18090487e-01 -2.42441311e-01 -6.08531237e-01 -4.27280784e-01 4.87194270e-01 5.28992992e-03 4.67080995e-02 3.29057217e-01 -8.47034156e-01 -6.07221246e-01 6.03721797e-01 -4.80717421e-01 2.22838491e-01 1.18899357e+00 -1.18649781e-01 -5.15711427e-01 1.15977146e-01 1.34495759e+00 8.55911374e-02 -8.25756073e-01 -6.44060314e-01 6.30106688e-01 -3.08499038e-01 1.87103435e-01 -8.24491560e-01 -1.30445385e+00 5.39108753e-01 -1.90132514e-01 4.85743314e-01 1.06572282e+00 3.85751694e-01 8.63922536e-01 5.78035831e-01 1.64934292e-01 -1.14013076e+00 5.74789606e-02 7.32557058e-01 6.85347140e-01 -1.26594555e+00 3.59390140e-01 -1.08732474e+00 -6.20347083e-01 9.82464373e-01 7.41456091e-01 7.13199377e-02 7.77809680e-01 3.70013326e-01 -3.82368177e-01 -7.23703623e-01 -6.66501641e-01 -6.46540284e-01 8.30801308e-01 4.89320904e-01 4.73871946e-01 1.70442760e-01 -2.45131701e-01 7.81268001e-01 -1.04041599e-01 -1.31494656e-01 3.53638887e-01 9.89026845e-01 -1.29808709e-01 -1.32164252e+00 3.03637479e-02 6.34326398e-01 -4.83593851e-01 -3.64534795e-01 -9.19982493e-01 7.13675916e-01 2.19788387e-01 8.82215261e-01 9.72611904e-02 -4.79280442e-01 4.86572474e-01 -3.13472003e-02 6.54745340e-01 -7.64408052e-01 -4.63446528e-01 -1.54241294e-01 5.18822432e-01 -5.68794012e-01 -5.99539578e-01 -7.13015079e-01 -1.52973044e+00 8.59176666e-02 -3.84626180e-01 -8.88244659e-02 3.74678075e-01 1.05559790e+00 5.55467188e-01 7.43358433e-01 4.68816847e-01 3.47469561e-02 1.86113551e-01 -6.52205884e-01 -6.12269461e-01 3.43595415e-01 1.11124590e-01 -4.11982447e-01 2.97041774e-01 2.02998325e-01]
[9.133729934692383, 8.28341007232666]
74585c89-01ae-427f-9ed0-7cdef70e0a70
forensic-license-plate-recognition-with
2207.14686
null
https://arxiv.org/abs/2207.14686v2
https://arxiv.org/pdf/2207.14686v2.pdf
Forensic License Plate Recognition with Compression-Informed Transformers
Forensic license plate recognition (FLPR) remains an open challenge in legal contexts such as criminal investigations, where unreadable license plates (LPs) need to be deciphered from highly compressed and/or low resolution footage, e.g., from surveillance cameras. In this work, we propose a side-informed Transformer architecture that embeds knowledge on the input compression level to improve recognition under strong compression. We show the effectiveness of Transformers for license plate recognition (LPR) on a low-quality real-world dataset. We also provide a synthetic dataset that includes strongly degraded, illegible LP images and analyze the impact of knowledge embedding on it. The network outperforms existing FLPR methods and standard state-of-the art image recognition models while requiring less parameters. For the severest degraded images, we can improve recognition by up to 8.9 percent points.
['Christian Riess', 'Jürgen Seiler', 'Andreas Spruck', 'Anatol Maier', 'Denise Moussa']
2022-07-29
null
null
null
null
['license-plate-recognition']
['computer-vision']
[ 7.15674877e-01 -3.76585960e-01 -2.27585226e-01 -9.40804854e-02 -1.45545030e+00 -9.61256146e-01 3.90687495e-01 -7.52205729e-01 -2.70216405e-01 5.14880598e-01 1.33320034e-01 -5.97487390e-01 2.28705090e-02 -4.96217698e-01 -9.63837504e-01 -5.78286171e-01 4.00294006e-01 2.39913732e-01 2.73161173e-01 2.10956335e-01 5.89126766e-01 7.73859322e-01 -1.00290966e+00 6.98035419e-01 5.41746616e-01 9.89026845e-01 -1.25224367e-01 7.11852014e-01 3.45897645e-01 9.33371723e-01 -5.03501832e-01 -1.24380374e+00 7.26331413e-01 2.28678644e-01 -3.88389081e-01 1.91662759e-01 9.02778387e-01 -8.29716444e-01 -1.04609287e+00 1.17413223e+00 3.49279046e-01 -2.29251161e-01 7.55739987e-01 -8.72754395e-01 -9.33136940e-01 1.58616766e-01 -7.50169814e-01 5.61763346e-01 2.41248116e-01 5.32968044e-01 6.25004768e-01 -1.15833294e+00 7.13065743e-01 9.97240067e-01 1.00221300e+00 5.05255044e-01 -8.97118926e-01 -7.64452279e-01 -6.46093488e-01 7.75978863e-01 -1.45726788e+00 -9.87418354e-01 5.81822157e-01 -2.59590656e-01 1.03240001e+00 2.38008261e-01 -5.30124307e-02 1.14360464e+00 1.06102847e-01 8.16402793e-01 1.14485097e+00 -1.31406948e-01 -8.96782801e-02 -1.08043768e-01 1.59222186e-01 4.65130568e-01 6.56161249e-01 3.43965925e-02 -6.70002997e-01 5.47383167e-02 7.67852426e-01 3.88488211e-02 -4.37807024e-01 3.45942616e-01 -7.53221810e-01 5.35500526e-01 -8.53079557e-02 4.58104759e-02 -1.57454714e-01 -5.46431877e-02 1.93932071e-01 9.84734222e-02 6.35088161e-02 3.66132319e-01 1.00975305e-01 -4.94045258e-01 -1.38118327e+00 -2.54288614e-01 6.68208599e-01 7.83703983e-01 4.91603076e-01 3.15024912e-01 -1.21839285e-01 9.56280470e-01 2.04502746e-01 9.11431968e-01 1.52734011e-01 -1.00919795e+00 1.15894973e+00 1.88372701e-01 -1.35606686e-02 -1.10453629e+00 4.21424896e-01 -2.63610482e-01 -6.71354711e-01 2.36203521e-01 4.28455055e-01 1.68058783e-01 -9.23489273e-01 8.00786555e-01 -5.24249732e-01 6.54004812e-01 2.16273904e-01 8.77280772e-01 4.83588874e-01 7.61475027e-01 -4.82987225e-01 9.93135199e-02 1.31013346e+00 -8.89945388e-01 -5.33405602e-01 -5.77825069e-01 1.55199423e-01 -7.28705168e-01 8.11617076e-01 5.73049128e-01 -1.03479743e+00 -2.53013223e-01 -1.29035985e+00 -1.29108831e-01 -6.99598389e-03 4.38437730e-01 -3.30339223e-02 1.11162376e+00 -8.29093158e-01 3.99030358e-01 -6.45038486e-01 2.58550674e-01 9.57465887e-01 4.26311314e-01 -7.93567955e-01 -9.44533288e-01 -8.99815261e-01 8.71297777e-01 -1.17335498e-01 2.52561688e-01 -8.20086420e-01 -6.92045450e-01 -8.00650954e-01 1.67156130e-01 3.27132821e-01 1.38741225e-01 6.91539824e-01 -4.53539342e-01 -1.29821968e+00 9.48284209e-01 -6.35822415e-02 -6.35916591e-01 6.27640247e-01 -2.51893908e-01 -7.96599507e-01 7.52571464e-01 -1.75015599e-01 1.63677469e-01 1.20090592e+00 -1.05043805e+00 -1.52235433e-01 -4.29095834e-01 -2.54921764e-01 -9.86725464e-02 -3.25053900e-01 2.99114317e-01 -7.62606144e-01 -6.08221531e-01 -3.08108628e-01 -9.03663099e-01 2.63714284e-01 -8.86865929e-02 -5.18556535e-01 4.96983171e-01 1.08869314e+00 -1.42443454e+00 7.75229275e-01 -2.72747898e+00 -5.39574265e-01 2.35619619e-01 1.70209169e-01 8.75974000e-01 -1.84844628e-01 1.54711440e-01 9.89998728e-02 3.20696831e-01 -3.10571700e-01 -3.77714723e-01 1.27172649e-01 -2.26640813e-02 -7.02439249e-01 8.46877813e-01 4.20224339e-01 1.03858626e+00 -2.21415058e-01 -4.41054434e-01 1.38731360e-01 6.19842708e-01 -3.46939325e-01 -8.27958807e-02 5.06177604e-01 -3.57129276e-02 -1.43697098e-01 1.12712789e+00 1.11690342e+00 -2.79256791e-01 4.97822911e-02 -3.55839550e-01 1.85327142e-01 1.71647314e-02 -8.39316726e-01 1.03669512e+00 4.63364907e-02 1.32676458e+00 6.70161843e-02 -6.83098197e-01 1.05348015e+00 8.36402178e-02 -1.10977001e-01 -1.01513529e+00 -1.16158836e-02 3.57131392e-01 -2.62225240e-01 -4.80579466e-01 6.68058276e-01 -1.45680532e-01 -1.16382204e-02 3.12161386e-01 -2.69606054e-01 4.37449992e-01 6.83799759e-02 -1.11115791e-01 1.50996828e+00 -4.55000252e-01 -1.66901678e-01 3.98585588e-01 5.69772959e-01 -2.02705637e-01 5.13696194e-01 6.79049671e-01 -3.24246138e-01 9.32045400e-01 3.63387644e-01 -2.55772352e-01 -1.38513470e+00 -1.12543035e+00 -1.86702579e-01 4.09922659e-01 6.47608712e-02 1.37064874e-01 -6.39356315e-01 -4.73092765e-01 1.62876323e-02 5.68047464e-01 -9.51145664e-02 -1.92165896e-01 -9.46846068e-01 -6.50330245e-01 1.47942770e+00 6.53395295e-01 9.57808733e-01 -5.94562232e-01 -7.87961930e-02 -1.26921371e-01 -2.15096071e-01 -1.91775799e+00 -5.93215108e-01 -5.57088673e-01 -6.55008018e-01 -1.15588927e+00 -5.89690745e-01 -5.97712934e-01 5.33261478e-01 3.90763134e-01 6.71081603e-01 -9.00647864e-02 -5.64767383e-02 3.49494427e-01 -1.81273833e-01 1.48458764e-01 -4.88964081e-01 -3.11078221e-01 4.80309082e-03 4.66498733e-01 4.78894085e-01 -4.27605659e-01 -3.43091697e-01 4.77888852e-01 -1.15834308e+00 -3.22501540e-01 1.04715431e+00 6.32761598e-01 3.71540457e-01 1.02773234e-01 2.81968653e-01 -5.55182695e-01 5.83632052e-01 -2.48977870e-01 -6.60236001e-01 4.89145428e-01 -2.80945957e-01 -1.33853406e-01 6.62303388e-01 -4.14369047e-01 -1.15585136e+00 -1.66586205e-01 -2.50256836e-01 -1.00467920e+00 -7.75203183e-02 2.37513259e-01 -2.34845862e-01 -6.15590155e-01 4.31758702e-01 8.08655024e-01 -1.97955426e-02 -4.35105860e-01 6.62312284e-02 1.11215532e+00 1.14177775e+00 -3.34551215e-01 1.07219017e+00 4.56454188e-01 -9.66692716e-02 -9.22979712e-01 -7.57178366e-02 -3.25935930e-01 -3.07077259e-01 -3.93221915e-01 4.43641633e-01 -8.58221829e-01 -8.38536024e-01 7.61582792e-01 -1.06506896e+00 -2.47565284e-02 8.01531225e-02 5.21160603e-01 -4.14696097e-01 9.34768856e-01 -1.02600157e+00 -7.77602792e-01 -2.25269943e-01 -1.13605118e+00 1.16175830e+00 3.51092778e-02 4.98174161e-01 -4.36480969e-01 -1.01433761e-01 1.31072855e+00 4.04973626e-01 4.36549187e-02 5.14238179e-01 -7.38908947e-01 -1.01386678e+00 -6.40302539e-01 -6.04234278e-01 7.01027513e-01 -3.57167095e-01 -1.29830822e-01 -1.14235890e+00 -3.37561578e-01 3.69302928e-02 -4.22279954e-01 1.21943045e+00 -2.24626567e-02 9.10444438e-01 -7.19529390e-01 -7.81778917e-02 8.86720836e-01 1.46719599e+00 2.28853509e-01 1.60782146e+00 3.83327454e-01 6.38902485e-01 1.34208828e-01 2.15076581e-01 3.31505537e-01 4.46304381e-02 6.06597185e-01 1.04408756e-01 3.11341971e-01 -3.12147290e-01 -2.67474174e-01 8.94197881e-01 7.75138438e-01 -2.62284726e-01 -4.60543513e-01 -1.10176778e+00 2.38383427e-01 -1.30791485e+00 -1.30897391e+00 5.41055240e-02 1.97257614e+00 6.29784286e-01 1.58795774e-01 -2.88196921e-01 3.06353062e-01 1.00248790e+00 2.91423768e-01 -6.40323400e-01 -1.98296756e-01 -4.81688768e-01 -1.44569268e-02 9.17256892e-01 4.19522315e-01 -9.86614883e-01 8.27035427e-01 6.51221371e+00 1.17736006e+00 -1.18170917e+00 5.01890816e-02 8.78903091e-01 -3.58694466e-03 -1.41651472e-02 -3.73561174e-01 -8.55888605e-01 6.60623074e-01 1.05731678e+00 1.42888837e-02 5.20766735e-01 6.44204497e-01 -2.77013540e-01 2.36892313e-01 -8.80697548e-01 1.44364095e+00 6.87173247e-01 -1.69097340e+00 8.67996365e-03 4.82292771e-01 4.02211338e-01 -3.48337889e-02 5.77355444e-01 1.43416137e-01 -9.90311801e-03 -1.27628255e+00 4.67657506e-01 6.95005119e-01 1.52868760e+00 -5.64398587e-01 9.77776289e-01 1.77812323e-01 -8.58023822e-01 -2.68434912e-01 -5.53110600e-01 4.18516695e-01 1.41903862e-01 4.21670049e-01 -9.12493825e-01 2.66139716e-01 3.30058098e-01 8.53823960e-01 -6.85451627e-01 8.93410683e-01 -1.00301720e-01 1.11240208e+00 -2.07780838e-01 4.52865094e-01 2.03139096e-01 -2.77700201e-02 4.97603774e-01 1.48277867e+00 6.30459845e-01 2.25852966e-01 -5.91343045e-01 8.90593112e-01 -4.95277673e-01 -5.20749450e-01 -7.02823162e-01 -1.32215291e-01 5.52408814e-01 8.46806586e-01 -3.35547060e-01 -1.23573914e-01 -4.97812688e-01 1.07548785e+00 -2.58093961e-02 3.77584904e-01 -9.34740901e-01 -1.64011210e-01 5.23442686e-01 1.07265204e-01 8.00921082e-01 -2.16544896e-01 -1.92605734e-01 -1.48137009e+00 3.49882275e-01 -1.05027819e+00 1.54553413e-01 -7.23471820e-01 -1.41655660e+00 3.19057167e-01 -4.61556643e-01 -1.59298909e+00 -1.14772603e-01 -1.00806987e+00 -3.72204036e-01 5.61315656e-01 -1.79801035e+00 -1.12728357e+00 1.78262040e-01 7.29924738e-01 4.32457238e-01 -6.57443881e-01 3.15812051e-01 6.35662138e-01 -6.51148617e-01 1.00441563e+00 5.25697470e-01 8.49557400e-01 5.44017792e-01 -5.05261958e-01 5.10445356e-01 1.24732387e+00 5.52072190e-02 3.29770684e-01 2.15590894e-01 -6.57409072e-01 -1.98182249e+00 -1.20587504e+00 5.67358136e-01 -4.37974513e-01 4.54535156e-01 -3.39937776e-01 -1.04184127e+00 6.60947621e-01 -6.40253276e-02 3.24986547e-01 9.43635762e-01 -6.56633556e-01 -1.05118263e+00 -3.46902937e-01 -1.43798137e+00 3.81939113e-01 7.08382487e-01 -9.27848220e-01 -7.40126014e-01 8.45180475e-05 1.94357023e-01 -1.23384133e-01 -8.04532826e-01 5.06457053e-02 7.25454092e-01 -7.93042839e-01 1.35355318e+00 -2.04924166e-01 6.68091357e-01 -3.03446591e-01 -5.02505541e-01 -6.61850095e-01 -3.69353056e-01 -3.05776417e-01 -2.17133731e-01 1.03503692e+00 3.41915458e-01 -5.72385371e-01 9.15696025e-01 7.39673972e-01 -1.08954847e-01 -2.60617942e-01 -1.34866965e+00 -1.16479683e+00 -5.75008430e-02 -6.42368019e-01 4.45099443e-01 7.06260145e-01 -4.22147699e-02 -1.37850493e-01 -9.04235065e-01 4.65117633e-01 9.65590656e-01 -1.28708899e-01 4.62548971e-01 -5.60450315e-01 -3.01535577e-01 -3.16569597e-01 -1.00966799e+00 -9.62920785e-01 2.66443491e-01 -8.32401395e-01 -1.55024290e-01 -1.01983881e+00 4.14109170e-01 5.16302437e-02 -1.53023541e-01 4.83781934e-01 1.11075297e-01 9.41196799e-01 6.94183111e-01 6.23569012e-01 -7.04444945e-01 1.25840768e-01 8.60741079e-01 -6.02112710e-01 5.62561989e-01 -4.18089747e-01 -6.55692160e-01 4.13587928e-01 8.03149819e-01 -4.13659900e-01 1.14070056e-02 -5.86657465e-01 2.00388134e-01 3.80722731e-01 4.50011849e-01 -1.18035233e+00 5.97813129e-01 8.95218402e-02 6.36111915e-01 -3.64910245e-01 6.52217329e-01 -7.71440685e-01 1.17948972e-01 2.53007025e-01 -8.13568980e-02 -1.78358153e-01 3.36401135e-01 5.50292611e-01 -4.23007280e-01 -2.11522385e-01 7.45426118e-01 1.38824016e-01 -7.53132701e-01 1.49095207e-01 -4.16464955e-01 7.40667805e-02 9.17278826e-01 -7.38955379e-01 -1.03712261e+00 -3.40223581e-01 -1.19417123e-01 -3.17116857e-01 5.93133688e-01 1.32593215e-01 1.28114438e+00 -1.29099607e+00 -1.03284192e+00 3.61149430e-01 -1.10507078e-01 -6.79467142e-01 4.72930610e-01 6.09120250e-01 -9.16209519e-01 5.91340125e-01 -4.81575400e-01 -3.09140444e-01 -1.29420340e+00 4.28677320e-01 1.99552625e-01 -4.12656903e-01 -6.09586179e-01 3.41178149e-01 -2.29865551e-01 -3.34424563e-02 -1.37873575e-01 1.56286389e-01 -1.09886579e-01 -2.64477819e-01 1.02148581e+00 7.74664879e-01 2.15010390e-01 -9.42559898e-01 -5.03604174e-01 8.84646654e-01 -1.85835361e-01 -1.95582911e-01 1.56612146e+00 2.87617683e-01 2.34010801e-01 -3.11192065e-01 1.36739457e+00 2.34287709e-01 -1.49869192e+00 -2.02340201e-01 -1.39397457e-01 -1.10935795e+00 -9.49967206e-02 -6.38394833e-01 -1.32081318e+00 9.71631110e-01 3.14254582e-01 -3.43990624e-01 1.02498889e+00 -1.57681003e-01 1.02681410e+00 7.84521401e-01 4.12875742e-01 -1.01705933e+00 5.84269091e-02 4.32167321e-01 8.29204798e-01 -9.82402265e-01 5.24741076e-02 -4.27833825e-01 -7.06606686e-01 1.35392618e+00 7.00555146e-02 -1.90787539e-01 1.66895866e-01 6.69430733e-01 -2.67369092e-01 4.51000407e-02 -5.37977040e-01 4.71165866e-01 2.62293786e-01 7.39631414e-01 -3.34376365e-01 -2.06943657e-02 3.05395991e-01 6.06817722e-01 2.39401963e-02 9.61507186e-02 7.93712676e-01 7.15931058e-01 -1.73604876e-01 -7.90557861e-01 -9.01333153e-01 7.11284876e-01 -7.00306177e-01 -2.90406615e-01 -4.43370104e-01 4.98220563e-01 -1.12902567e-01 9.44106698e-01 3.43510471e-02 -6.40109539e-01 1.84076101e-01 -5.56164514e-03 2.67018437e-01 1.22766839e-02 -1.83190614e-01 -2.04977095e-01 1.38893411e-01 -4.88733560e-01 -1.63869798e-01 -9.87361193e-01 -8.54281247e-01 -8.71986330e-01 -1.21818684e-01 -4.54653323e-01 4.43087637e-01 7.56874621e-01 4.85517681e-01 2.67431233e-02 5.78110158e-01 -6.24062419e-01 -5.43601453e-01 -5.32060802e-01 -7.36086071e-01 3.43445599e-01 5.28871357e-01 -2.19248354e-01 -5.88868380e-01 4.26256508e-01]
[9.892252922058105, -4.835448265075684]
eb1b6c7c-a085-4e34-be80-677e52dd18b7
product-information-extraction-using-chatgpt
2306.14921
null
https://arxiv.org/abs/2306.14921v1
https://arxiv.org/pdf/2306.14921v1.pdf
Product Information Extraction using ChatGPT
Structured product data in the form of attribute/value pairs is the foundation of many e-commerce applications such as faceted product search, product comparison, and product recommendation. Product offers often only contain textual descriptions of the product attributes in the form of titles or free text. Hence, extracting attribute/value pairs from textual product descriptions is an essential enabler for e-commerce applications. In order to excel, state-of-the-art product information extraction methods require large quantities of task-specific training data. The methods also struggle with generalizing to out-of-distribution attributes and attribute values that were not a part of the training data. Due to being pre-trained on huge amounts of text as well as due to emergent effects resulting from the model size, Large Language Models like ChatGPT have the potential to address both of these shortcomings. This paper explores the potential of ChatGPT for extracting attribute/value pairs from product descriptions. We experiment with different zero-shot and few-shot prompt designs. Our results show that ChatGPT achieves a performance similar to a pre-trained language model but requires much smaller amounts of training data and computation for fine-tuning.
['Christian Bizer', 'Reng Chiz Der', 'Roee Shraga', 'Alexander Brinkmann']
2023-06-23
null
null
null
null
['product-recommendation']
['miscellaneous']
[-3.84195969e-02 6.63192719e-02 -6.84396803e-01 -6.47830486e-01 -9.57602799e-01 -7.69106328e-01 5.25419354e-01 6.91864192e-01 -1.86812028e-01 4.42533374e-01 2.52340108e-01 -2.34355837e-01 -1.56884998e-01 -8.35130095e-01 -3.30513060e-01 -1.93077341e-01 -1.32004758e-02 9.75793123e-01 -5.75526012e-03 -5.67834735e-01 2.52984762e-01 -4.11615558e-02 -1.65803587e+00 4.92182910e-01 7.18318582e-01 1.10344386e+00 1.62003756e-01 3.27532977e-01 -1.12769365e+00 5.85387647e-01 -5.24759352e-01 -7.02285230e-01 3.60120952e-01 -2.06205338e-01 -7.62035012e-01 2.47272134e-01 1.44714206e-01 -2.63066351e-01 9.16872397e-02 1.05030203e+00 4.39337671e-01 2.95530349e-01 6.02877319e-01 -1.35252035e+00 -7.14981735e-01 8.03693771e-01 -3.26033711e-01 -9.21287090e-02 6.03005469e-01 8.16512704e-02 1.40583408e+00 -8.07221353e-01 6.74861491e-01 9.78454292e-01 5.03749013e-01 2.54223913e-01 -1.40824437e+00 -4.76924926e-01 5.66816740e-02 2.70274524e-02 -1.33559155e+00 -4.39839751e-01 5.92938066e-01 -4.34231043e-01 1.33227301e+00 1.21249750e-01 6.02066994e-01 9.32690144e-01 2.37041265e-02 7.83884108e-01 6.99699700e-01 -2.84842432e-01 2.61051983e-01 7.17411220e-01 2.03583777e-01 1.76732451e-01 1.05490707e-01 -2.23716989e-01 -5.84943175e-01 -4.14794415e-01 4.36414093e-01 8.49655177e-03 3.00926715e-01 -3.02004188e-01 -9.10883009e-01 1.19249570e+00 -2.04884261e-01 9.95863825e-02 -5.52263558e-01 -2.49096304e-01 5.82264006e-01 5.21378636e-01 7.04689205e-01 7.37399876e-01 -8.60778749e-01 -5.78011930e-01 -8.90740633e-01 4.72996175e-01 1.39461946e+00 1.63774526e+00 8.41481030e-01 -2.80556053e-01 -2.36984640e-01 9.49822009e-01 2.17194691e-01 2.17110530e-01 5.72901726e-01 -5.97512782e-01 5.27793586e-01 6.87968075e-01 4.14015591e-01 -7.92663932e-01 -1.67782307e-01 -3.01952004e-01 -1.84694335e-01 -2.97722191e-01 4.49373931e-01 -1.88603893e-01 -8.56524825e-01 1.33321249e+00 3.53027344e-01 -6.30398273e-01 3.87309007e-02 7.78341591e-01 8.98215592e-01 7.95999527e-01 4.45791513e-01 -1.74574554e-01 1.71490026e+00 -6.29811108e-01 -9.14051712e-01 -3.98368567e-01 1.16625166e+00 -1.18036389e+00 1.39836824e+00 2.03211620e-01 -8.81147385e-01 -3.36344033e-01 -7.64093280e-01 -3.29262048e-01 -6.39427006e-01 -2.59761900e-01 8.87097836e-01 5.55485547e-01 -3.00506532e-01 5.61641276e-01 -3.80082786e-01 -3.68899554e-01 3.69667709e-01 2.20882609e-01 -1.88497275e-01 -2.05744922e-01 -1.36037219e+00 8.69032145e-01 2.82479495e-01 -3.93189639e-01 -3.10555518e-01 -9.97789025e-01 -1.07291448e+00 3.87243241e-01 7.81413615e-01 -5.11446238e-01 1.65309870e+00 -9.58797812e-01 -1.33625245e+00 4.45925862e-01 -1.13910586e-01 -1.25631914e-01 -3.11053405e-03 -2.57918298e-01 -5.30542672e-01 -2.30650604e-01 2.99047083e-01 5.30214012e-01 5.50805509e-01 -5.66397667e-01 -7.68591464e-01 -2.89079398e-01 -1.46654725e-01 2.71050155e-01 -3.18995327e-01 2.41504058e-01 -3.16722006e-01 -3.25944364e-01 -1.77086845e-01 -6.77439988e-01 -2.51269966e-01 -3.29116821e-01 -3.40114743e-01 -6.51383102e-01 6.09807670e-01 -4.28406209e-01 1.23698843e+00 -2.26909971e+00 -4.29455549e-01 1.59066051e-01 1.54584184e-01 8.06269273e-02 -2.33400807e-01 8.21316898e-01 2.22264454e-01 3.46221119e-01 4.71512616e-01 -1.28767177e-01 3.90735865e-01 1.44409359e-01 -1.22871362e-01 3.10793370e-02 1.57957450e-01 1.01405931e+00 -9.57753956e-01 -7.35039711e-01 -5.36473431e-02 3.09042335e-01 -2.64722049e-01 2.02496469e-01 -7.73200035e-01 3.85282212e-03 -8.29375923e-01 5.60673535e-01 2.50326902e-01 -3.92489403e-01 6.03482276e-02 -2.28281319e-01 -1.17134057e-01 8.33542526e-01 -9.19562817e-01 1.66081274e+00 -5.00052929e-01 2.71055102e-01 -1.03151709e-01 -6.09649003e-01 9.66404080e-01 4.57636654e-01 7.62741208e-01 -8.66077065e-01 2.57185847e-01 1.98078498e-01 1.41565144e-01 -5.71519077e-01 8.44666064e-01 -3.94404829e-01 -3.09400201e-01 6.22823894e-01 2.21087188e-01 -3.64284217e-01 5.96524775e-01 3.74086976e-01 8.91696215e-01 -1.75245535e-02 4.34637845e-01 -1.77250743e-01 2.46044877e-03 5.09915471e-01 4.88415629e-01 5.40014684e-01 1.51568368e-01 2.47709796e-01 6.97804272e-01 -5.50531983e-01 -1.43850863e+00 -5.51376045e-01 -5.22963814e-02 1.26845181e+00 -1.10950030e-01 -9.92935836e-01 -4.35283124e-01 -6.40970170e-01 2.22436681e-01 1.14742589e+00 -2.12292641e-01 -6.17490336e-02 3.90478894e-02 -4.68766570e-01 6.39285594e-02 3.56991351e-01 1.97601035e-01 -8.24482143e-01 -3.49875748e-01 6.08487666e-01 -9.96995866e-02 -1.23651683e+00 -7.19881475e-01 2.69255877e-01 -6.83808982e-01 -5.70374966e-01 -4.38756287e-01 -6.43418074e-01 3.05585802e-01 1.39019102e-01 1.46492612e+00 -3.37259114e-01 -1.21645190e-01 8.39929432e-02 -6.12293601e-01 -7.52047360e-01 -4.79411334e-01 4.40164834e-01 -2.59975702e-01 -1.05665028e-01 1.25330782e+00 -3.36863667e-01 -3.37551713e-01 4.47543740e-01 -7.08541334e-01 8.56165737e-02 7.64603019e-01 6.62931800e-01 4.85001832e-01 2.02015311e-01 7.34077811e-01 -1.26140797e+00 9.54623640e-01 -6.56519890e-01 -6.24185145e-01 1.15689501e-01 -9.47877526e-01 2.80561775e-01 4.71051604e-01 -8.12636793e-01 -1.07588017e+00 1.20550029e-01 -2.06425771e-01 -1.38159879e-02 -9.54528227e-02 1.04531479e+00 -6.28863797e-02 5.36396682e-01 5.75905442e-01 -2.95791905e-02 2.68335044e-01 -6.97975099e-01 5.41547120e-01 9.30940688e-01 -8.07761177e-02 -3.45347136e-01 3.86598736e-01 -2.08016515e-01 -3.08547974e-01 -8.59122276e-01 -1.04929972e+00 -8.75261128e-01 -4.59830225e-01 2.07776174e-01 4.95942026e-01 -8.23394835e-01 -6.46593750e-01 -6.33350611e-02 -9.15548503e-01 -1.90718964e-01 -6.83382213e-01 5.47785103e-01 -3.86624426e-01 -6.48875609e-02 -7.60423064e-01 -6.33533120e-01 -4.12304580e-01 -8.37436199e-01 9.12932158e-01 1.10314839e-01 -7.32579887e-01 -8.59232903e-01 -1.62095234e-01 3.28457296e-01 6.73498690e-01 -2.32751712e-01 1.29119527e+00 -1.34501970e+00 -3.61298591e-01 -7.26796269e-01 -8.54418427e-02 1.23649023e-01 3.05468738e-01 -3.52762818e-01 -5.70245028e-01 5.18279262e-02 2.99400594e-02 -3.98861855e-01 2.37032264e-01 9.31086019e-02 7.06518352e-01 -5.58265924e-01 -2.90862381e-01 6.39519468e-02 1.16632295e+00 2.07326770e-01 4.56586361e-01 1.06837071e-01 6.11936390e-01 9.88539696e-01 1.04165530e+00 7.00363040e-01 3.51701051e-01 8.42735887e-01 -2.52433363e-02 1.75986066e-01 1.45872131e-01 -5.03084481e-01 4.28505652e-02 6.23581052e-01 5.02922118e-01 -1.40553758e-01 -6.74224079e-01 6.50353789e-01 -1.77800286e+00 -9.08319294e-01 -1.80451095e-01 2.22492957e+00 1.11313772e+00 2.40761057e-01 2.47098431e-01 -2.61582136e-01 2.99119502e-01 -1.64563283e-01 -6.19672656e-01 -5.63367844e-01 2.04609975e-01 -2.73535755e-02 5.77476442e-01 1.79196745e-01 -7.87564635e-01 9.00918961e-01 6.10000563e+00 1.01864207e+00 -8.05593908e-01 7.61677697e-02 5.73765993e-01 -2.61589587e-01 -5.23790419e-01 -1.04222260e-02 -1.11062729e+00 5.45258701e-01 1.21926045e+00 -5.63793540e-01 2.68062830e-01 1.14516187e+00 1.19028717e-01 -1.85545340e-01 -1.38075781e+00 9.17009652e-01 -1.40186220e-01 -1.10081708e+00 5.00672963e-03 4.63986248e-01 4.35967475e-01 -1.55473039e-01 8.79155472e-03 6.23018801e-01 5.13843060e-01 -9.51899171e-01 6.61340296e-01 -1.27791852e-01 7.28156745e-01 -6.37133777e-01 9.74008143e-01 4.06596810e-01 -1.11848056e+00 -1.08978897e-02 -4.99364704e-01 -1.98560029e-01 4.36919987e-01 6.89261317e-01 -1.01122725e+00 3.56290609e-01 4.42909926e-01 6.72521234e-01 -1.87055320e-01 7.41447628e-01 8.84488821e-02 4.33743834e-01 -3.68847370e-01 -5.21946430e-01 3.50326031e-01 -1.93961933e-01 1.77302808e-01 9.96924460e-01 4.42974567e-01 2.18926609e-01 4.43162173e-01 1.02095592e+00 -7.54717663e-02 4.10468996e-01 -6.06703043e-01 -7.87413239e-01 6.36455834e-01 1.37122893e+00 -5.75874388e-01 -4.78311807e-01 -1.02542460e+00 5.26134908e-01 3.79888006e-02 2.53754407e-01 -4.62669253e-01 -5.88978231e-01 9.27213669e-01 5.39054811e-01 5.10961235e-01 -2.68649936e-01 -2.54961461e-01 -8.55373204e-01 -7.38263652e-02 -9.16944861e-01 2.56916940e-01 -5.04499555e-01 -1.60724998e+00 4.88257647e-01 2.18508884e-01 -1.09770560e+00 -7.52798557e-01 -4.17837709e-01 -1.05940953e-01 1.04389691e+00 -1.17248249e+00 -8.75302851e-01 -2.88506802e-02 2.92931527e-01 7.16533542e-01 -1.41099855e-01 9.88857865e-01 3.71137053e-01 -1.23961762e-01 3.92089784e-01 -7.86069632e-02 2.08565891e-01 7.02658713e-01 -1.09986842e+00 6.94973409e-01 1.57159746e-01 2.13929221e-01 7.43500113e-01 9.78275597e-01 -8.30119193e-01 -1.57935154e+00 -7.27987349e-01 1.45468938e+00 -6.15939856e-01 8.06927919e-01 -6.41354799e-01 -9.63353395e-01 6.00734949e-01 1.81656346e-01 -1.05266757e-01 1.09341276e+00 6.38300002e-01 -4.18412209e-01 -1.46513700e-01 -1.04381168e+00 3.75248104e-01 5.92449844e-01 -6.49985194e-01 -5.06311774e-01 4.39441144e-01 9.45155799e-01 -1.78852558e-01 -1.16916156e+00 -1.16313025e-01 5.28317273e-01 -3.87766540e-01 6.99889362e-01 -7.23584712e-01 2.90304035e-01 1.18647762e-01 -3.89006734e-02 -1.38257194e+00 -3.03913265e-01 -7.12782919e-01 5.23659363e-02 1.51452613e+00 9.48475003e-01 -4.09389317e-01 7.91090012e-01 1.47928417e+00 1.93157569e-02 -7.19642341e-01 -4.64734703e-01 -9.33860183e-01 -1.95769534e-01 -3.43940377e-01 8.43567908e-01 7.25126326e-01 3.99783969e-01 9.89056230e-01 -9.60430801e-02 -4.76035565e-01 4.39667016e-01 2.52432972e-01 6.90233231e-01 -1.32513952e+00 -2.05711186e-01 -2.68346846e-01 -1.50107130e-01 -1.04952300e+00 -1.52473778e-01 -6.88447773e-01 4.90647741e-02 -1.58301425e+00 2.38318354e-01 -7.28114367e-01 5.59021756e-02 5.23310602e-01 1.37299031e-01 -1.48626834e-01 1.51620492e-01 1.84940726e-01 -5.45367301e-01 3.85886967e-01 1.06721425e+00 -1.79823950e-01 -5.49111009e-01 1.76486745e-01 -8.36656392e-01 2.03692466e-01 8.54037702e-01 -7.14660347e-01 -7.15695322e-01 -9.27263275e-02 7.45337486e-01 6.80433884e-02 -1.97644323e-01 -2.39931911e-01 2.71623760e-01 -2.64159620e-01 1.04585126e-01 -4.19586509e-01 3.57389241e-01 -8.61026049e-01 1.94424674e-01 -1.63500458e-01 -6.06194735e-01 1.77716687e-01 1.02062643e-01 3.43327582e-01 -3.92313659e-01 -4.88547176e-01 1.78725049e-01 -4.89426672e-01 -8.02181840e-01 2.62964398e-01 -4.80250031e-01 1.67739794e-01 7.73287237e-01 -2.23763183e-01 -3.40229124e-01 -5.98641396e-01 -4.97447997e-01 2.56565839e-01 2.78260738e-01 6.11335039e-01 3.72511268e-01 -1.18759692e+00 -5.98161101e-01 2.44779989e-01 4.51482296e-01 -1.17536239e-01 2.26576552e-02 6.71479702e-01 -9.86954384e-03 6.51048243e-01 -1.56248555e-01 -3.27968419e-01 -1.09630275e+00 8.55539203e-01 -2.42440119e-01 -3.26586366e-01 -5.67049026e-01 5.34066319e-01 -2.42596082e-02 -2.21379027e-01 1.32557794e-01 -1.59474283e-01 -1.16517313e-01 3.73651147e-01 6.32701099e-01 -8.01738799e-02 1.91753939e-01 -3.39108616e-01 1.37741612e-02 6.43867850e-02 -6.24961138e-01 -2.83788353e-01 1.23319900e+00 -2.52014637e-01 2.11602181e-01 4.77319419e-01 1.29506624e+00 -2.40498617e-01 -1.15978646e+00 -5.27462840e-01 4.65200245e-01 -5.45594454e-01 1.72027528e-01 -8.26557934e-01 -8.10178816e-01 6.21475101e-01 6.14976659e-02 5.10156214e-01 7.43481696e-01 3.40499163e-01 1.01490378e+00 3.66441220e-01 6.09739780e-01 -1.35935521e+00 -4.62764315e-02 2.56324857e-01 4.04272467e-01 -1.33702433e+00 1.77431479e-02 -6.81635022e-01 -1.00104034e+00 8.20719540e-01 3.51947039e-01 4.30719107e-01 6.87243164e-01 3.94927800e-01 1.47892490e-01 -4.16897714e-01 -9.50554848e-01 -3.77212077e-01 3.65873575e-01 4.72347826e-01 8.18315625e-01 -8.50540474e-02 -2.85661906e-01 8.16095889e-01 -2.81665981e-01 -1.74498954e-03 3.09602529e-01 1.09083652e+00 -3.82370025e-01 -1.33988857e+00 3.06878239e-01 8.86059701e-01 -6.58154130e-01 -3.60719889e-01 -2.75635213e-01 7.75837362e-01 -2.49767751e-01 1.10030568e+00 1.80910602e-01 -2.90525347e-01 3.40886682e-01 3.83289248e-01 1.40221149e-01 -1.05631185e+00 -7.56359816e-01 2.49415234e-01 6.21805131e-01 -5.73522091e-01 -9.20508057e-02 -6.73985124e-01 -1.03229237e+00 -4.89964187e-01 -4.42684263e-01 4.66212958e-01 9.31633055e-01 8.90660465e-01 6.95230544e-01 1.74786076e-01 3.12107563e-01 -4.29765373e-01 -9.79734302e-01 -1.17169201e+00 -8.49431694e-01 6.87094450e-01 2.56388411e-02 -5.43627977e-01 2.60331947e-02 2.49167979e-02]
[10.001337051391602, 6.233855724334717]
3eba0552-40e2-45d1-be94-1db42e50a63e
co-fusion-real-time-segmentation-tracking-and
1706.06629
null
http://arxiv.org/abs/1706.06629v1
http://arxiv.org/pdf/1706.06629v1.pdf
Co-Fusion: Real-time Segmentation, Tracking and Fusion of Multiple Objects
In this paper we introduce Co-Fusion, a dense SLAM system that takes a live stream of RGB-D images as input and segments the scene into different objects (using either motion or semantic cues) while simultaneously tracking and reconstructing their 3D shape in real time. We use a multiple model fitting approach where each object can move independently from the background and still be effectively tracked and its shape fused over time using only the information from pixels associated with that object label. Previous attempts to deal with dynamic scenes have typically considered moving regions as outliers, and consequently do not model their shape or track their motion over time. In contrast, we enable the robot to maintain 3D models for each of the segmented objects and to improve them over time through fusion. As a result, our system can enable a robot to maintain a scene description at the object level which has the potential to allow interactions with its working environment; even in the case of dynamic scenes.
['Martin Rünz', 'Lourdes Agapito']
2017-06-20
null
null
null
null
['object-slam', 'semantic-slam']
['computer-vision', 'computer-vision']
[ 2.72834837e-01 1.26677901e-01 1.21214271e-01 -3.05033803e-01 -4.83481348e-01 -7.73407638e-01 4.17109460e-01 2.49781728e-01 -4.38689142e-01 2.23902762e-01 -5.50131798e-01 9.19923633e-02 1.47250265e-01 -5.18154383e-01 -6.32904828e-01 -6.33466005e-01 1.28125072e-01 1.06715357e+00 9.27891910e-01 1.67516932e-01 1.76346958e-01 1.04153848e+00 -1.60226309e+00 -1.18694462e-01 4.66176778e-01 7.75644362e-01 6.70972109e-01 9.59141612e-01 -2.97434658e-01 7.56906986e-01 -4.08185184e-01 2.46858686e-01 5.43141603e-01 -1.21261559e-01 -7.53043771e-01 8.46064627e-01 4.78788167e-01 -2.06513554e-01 -4.75320160e-01 1.15439951e+00 -1.05872855e-01 2.63891757e-01 2.20743671e-01 -1.25721049e+00 3.34206223e-01 -3.50858927e-01 -5.70658505e-01 -2.02773452e-01 6.18225753e-01 2.04944849e-01 2.01973826e-01 -6.37964845e-01 9.48968828e-01 1.27342081e+00 6.07993484e-01 4.41994727e-01 -1.44350028e+00 -2.33327419e-01 3.10379595e-01 -1.05734970e-02 -1.31276500e+00 -6.91941440e-01 8.02790165e-01 -5.13307512e-01 8.09425235e-01 1.78613648e-01 7.83520579e-01 4.50852752e-01 3.16175342e-01 7.79641330e-01 7.01936066e-01 -3.61385018e-01 3.59334767e-01 -1.42025221e-02 2.69317348e-02 7.68241227e-01 2.90432245e-01 6.17818795e-02 -4.51583833e-01 5.34746945e-02 8.24011683e-01 1.77682906e-01 -1.49290338e-01 -1.12923861e+00 -1.60868144e+00 1.90124795e-01 4.48976129e-01 2.25386649e-01 -6.53057396e-01 4.02436674e-01 -4.53630723e-02 -8.89681801e-02 2.25112349e-01 1.05218858e-01 -4.24834341e-01 1.19520448e-01 -1.13006055e+00 3.18818390e-01 6.54172778e-01 1.06066978e+00 1.20226610e+00 -6.56102737e-03 4.28118974e-01 1.30430982e-02 5.04221022e-01 6.78588092e-01 1.69070572e-01 -1.50693643e+00 9.35986191e-02 6.22181118e-01 4.72943157e-01 -9.58164692e-01 -5.12555599e-01 1.03396051e-01 -3.34787726e-01 7.91432500e-01 3.97261411e-01 2.19931871e-01 -1.22846413e+00 1.34122002e+00 7.83562958e-01 3.52572381e-01 2.42872342e-01 7.83275187e-01 2.79434532e-01 4.33128089e-01 -2.19287291e-01 -3.00918072e-01 7.91698933e-01 -9.35955822e-01 -7.22306252e-01 -7.52533615e-01 5.46383142e-01 -8.36734474e-01 1.73151448e-01 6.00181937e-01 -1.16741323e+00 -6.52331114e-01 -1.02706170e+00 1.40060652e-02 -2.12431267e-01 -3.10095727e-01 4.98836488e-01 6.84169233e-02 -1.08228469e+00 6.51305497e-01 -1.51743436e+00 -5.86313188e-01 1.31674156e-01 5.16751289e-01 -7.55406737e-01 -3.54410022e-01 -1.65708587e-01 1.19696081e+00 7.10969150e-01 1.96998551e-01 -9.27879691e-01 -1.09069362e-01 -9.38595116e-01 -4.86467004e-01 5.39481997e-01 -7.53451824e-01 1.48334074e+00 -1.05576789e+00 -1.12106049e+00 7.55006015e-01 -5.15664816e-01 -2.44661942e-01 5.25004923e-01 -1.04724444e-01 -2.25896519e-02 1.54010028e-01 1.27996609e-01 8.61298740e-01 7.18376577e-01 -1.86157072e+00 -9.61555600e-01 -6.79569185e-01 -1.30984858e-01 4.53619659e-01 5.74324310e-01 -2.75263131e-01 -8.93980324e-01 -1.76180969e-03 9.84545827e-01 -1.15601683e+00 -6.32916689e-01 3.34576726e-01 -8.22731331e-02 3.92765075e-01 1.18043923e+00 -4.86332148e-01 5.26482701e-01 -2.08987498e+00 4.95494425e-01 1.53273985e-01 5.33498712e-02 4.81342599e-02 1.03629515e-01 8.74128491e-02 3.84883553e-01 -4.07567054e-01 -3.80433083e-01 -9.47673142e-01 -4.14338112e-01 7.58233845e-01 4.07087803e-02 8.44431162e-01 -1.03483580e-01 6.02070332e-01 -1.09627974e+00 -6.36103570e-01 8.07274461e-01 4.22337592e-01 -2.33820334e-01 1.48001999e-01 -3.90787572e-01 7.19732404e-01 -4.81667966e-01 6.51771605e-01 7.97674596e-01 1.56267703e-01 2.13834867e-01 -2.09976152e-01 -2.93049246e-01 -1.73583865e-01 -1.54342699e+00 2.15362453e+00 -1.83238909e-01 6.59538031e-01 7.34730661e-01 -5.90669811e-01 7.35931396e-01 2.29390994e-01 9.61208642e-01 -4.50914562e-01 7.41361752e-02 2.28381068e-01 -2.84427315e-01 -3.83685052e-01 8.73681188e-01 -6.53878376e-02 2.34280229e-02 2.24020541e-01 -1.86005563e-01 -5.97286344e-01 -4.44062799e-02 1.83197737e-01 1.14514458e+00 6.04282796e-01 1.22568384e-01 2.84779102e-01 1.72155648e-01 6.91734135e-01 6.11251116e-01 6.33461177e-01 -3.66159648e-01 7.08526373e-01 -3.31911802e-01 -3.90755862e-01 -1.07357466e+00 -1.10953212e+00 1.20649956e-01 4.52020526e-01 9.48145390e-01 1.99612409e-01 -3.75798494e-01 -4.49180514e-01 5.96344993e-02 6.52185500e-01 -3.62432986e-01 -5.30019328e-02 -6.72988117e-01 -1.43117473e-01 -9.12024528e-02 2.21496180e-01 1.54039815e-01 -9.67814505e-01 -1.35114825e+00 5.86966097e-01 -1.35476530e-01 -1.16931045e+00 -1.60650894e-01 6.40799642e-01 -1.05215228e+00 -1.14838672e+00 -8.47645104e-02 -5.90344250e-01 7.92641997e-01 7.34720707e-01 8.24147105e-01 -1.83481574e-02 -2.37789810e-01 7.36377001e-01 -4.16813433e-01 -2.30352595e-01 -6.88660562e-01 -5.61397731e-01 3.99815030e-02 -1.26534060e-01 6.36671782e-02 -2.70825058e-01 -2.22004354e-01 2.10907981e-01 -9.87042785e-01 2.33546957e-01 2.92090684e-01 1.82030722e-01 9.10882771e-01 2.18205199e-01 -2.86901832e-01 -6.08227253e-01 -3.09838593e-01 -3.10595423e-01 -7.77685225e-01 2.33513579e-01 -2.88211793e-01 -1.36301503e-01 -1.13273978e-01 -3.41224641e-01 -1.01214421e+00 1.01134276e+00 4.18920159e-01 -9.37090516e-01 -4.84576553e-01 7.94020891e-02 -2.82141715e-01 -2.15652898e-01 3.07165146e-01 1.26459971e-01 1.71311080e-01 -4.81977582e-01 4.27302241e-01 3.34687084e-01 1.09227347e+00 -1.79289073e-01 7.37751186e-01 8.76920521e-01 1.11192077e-01 -7.02467859e-01 -7.06578374e-01 -1.02069414e+00 -1.20425987e+00 -5.80813050e-01 8.27374220e-01 -8.95388961e-01 -2.68246502e-01 4.72025096e-01 -1.37738097e+00 -2.66026974e-01 -4.14182007e-01 5.51968694e-01 -6.90416098e-01 4.04244393e-01 -2.94477880e-01 -1.05315232e+00 2.43483305e-01 -1.16079926e+00 1.36141407e+00 1.97523028e-01 -1.64201230e-01 -9.09314930e-01 -2.92761112e-03 1.47451058e-01 -9.22744803e-04 6.04504943e-01 1.20959245e-01 -2.81508654e-01 -8.86289775e-01 -5.40619373e-01 1.67157695e-01 -8.29062685e-02 2.31503993e-01 1.31893521e-02 -9.60374117e-01 -3.21849614e-01 2.70596027e-01 2.39832714e-01 6.15926564e-01 2.67147392e-01 5.02242804e-01 1.75234854e-01 -8.45618427e-01 4.73928034e-01 1.42359006e+00 4.87961382e-01 2.45488957e-01 5.30444443e-01 8.43586922e-01 6.81817830e-01 1.05385220e+00 2.01437265e-01 3.69836807e-01 8.01477909e-01 9.08857763e-01 -9.11842585e-02 -2.62525566e-02 4.55053747e-02 2.71551996e-01 4.42259610e-01 4.24877614e-01 -2.46596009e-01 -1.07532835e+00 6.37642980e-01 -2.22863555e+00 -7.16456890e-01 -5.06101370e-01 2.25260735e+00 2.04418316e-01 1.65546834e-01 -1.04776561e-01 1.06205948e-01 6.80490136e-01 -1.95908427e-01 -8.49229515e-01 -2.68108491e-02 -6.81803674e-02 -3.62750500e-01 9.90864396e-01 7.35627592e-01 -9.47725952e-01 1.05212831e+00 5.94238138e+00 -2.98542511e-02 -9.00639057e-01 3.81658375e-02 -8.09073746e-02 -1.96961835e-02 1.56266112e-02 3.93652916e-01 -5.58893025e-01 1.19048655e-01 6.30805552e-01 -5.55843487e-02 3.58935207e-01 7.04424620e-01 2.40555093e-01 -8.40644300e-01 -1.26351368e+00 9.22310531e-01 4.69864346e-02 -9.11058187e-01 -4.10916835e-01 2.31160164e-01 6.85960889e-01 2.36369982e-01 -3.57333690e-01 -2.35112891e-01 4.40055728e-01 -6.31018579e-01 1.32573438e+00 9.69766498e-01 1.51189417e-01 -3.78889233e-01 6.23573005e-01 9.42564905e-01 -1.03873193e+00 5.37692569e-02 -1.84194654e-01 2.39319005e-03 6.04859769e-01 2.62562245e-01 -1.18173420e+00 7.47025669e-01 5.48622072e-01 5.92388928e-01 -6.31149411e-01 1.41064024e+00 2.59776145e-01 -1.25387222e-01 -7.34537661e-01 4.87480164e-01 1.84567258e-01 -2.26177916e-01 7.78839290e-01 7.96680212e-01 4.26316142e-01 4.00744677e-02 8.28120053e-01 5.03372967e-01 6.16120994e-01 -4.96229231e-01 -6.12568736e-01 3.24810237e-01 5.18896759e-01 1.04928839e+00 -1.15662193e+00 -5.16365051e-01 -2.61523455e-01 1.21578264e+00 6.45031175e-03 2.95781642e-01 -4.20612842e-01 1.69016615e-01 5.17132342e-01 3.82384621e-02 3.44876260e-01 -7.81886578e-01 -1.17712706e-01 -1.11562979e+00 -5.46475239e-02 -2.83859462e-01 5.32024428e-02 -1.22992194e+00 -5.99659562e-01 4.94479567e-01 -4.20773402e-02 -1.25527406e+00 -3.63069773e-01 -3.92620623e-01 -3.26566910e-03 5.92535138e-01 -1.20319831e+00 -1.18576622e+00 -4.47938293e-01 5.46836078e-01 6.41740263e-01 4.12355751e-01 5.55074334e-01 -4.31629010e-02 -1.45058557e-01 -6.05733871e-01 2.58036464e-01 -3.12688559e-01 4.79416698e-01 -1.16506267e+00 8.20718482e-02 1.04047310e+00 4.23347205e-01 2.08464980e-01 1.12569714e+00 -9.14789319e-01 -1.45641541e+00 -1.09859192e+00 4.93949503e-01 -7.14886248e-01 3.66138965e-01 -2.30377436e-01 -9.48177338e-01 9.04744387e-01 -1.75940067e-01 2.58913517e-01 -3.05999834e-02 -5.40042996e-01 2.03867570e-01 3.61749828e-02 -1.24790311e+00 1.07774943e-01 8.27783465e-01 -2.94644654e-01 -5.51925838e-01 2.20086038e-01 6.61540389e-01 -7.46048689e-01 -4.91603553e-01 5.16707897e-01 3.73622030e-01 -1.10976851e+00 8.24772716e-01 -2.26609588e-01 -5.00303686e-01 -9.81152356e-01 -4.15189326e-01 -1.00965190e+00 -2.29255147e-02 -2.98129797e-01 -2.05584705e-01 9.44387078e-01 -9.04704109e-02 -1.59827009e-01 1.06444800e+00 8.43959630e-01 -2.77884573e-01 2.51874607e-02 -1.13816905e+00 -6.74248993e-01 -5.19402564e-01 -8.15931380e-01 2.75618643e-01 6.55149996e-01 -5.33108234e-01 -1.01048753e-01 -2.07063645e-01 6.15000188e-01 9.11047041e-01 3.07900876e-01 1.06080592e+00 -1.32356954e+00 5.05825654e-02 -1.38646349e-01 -7.61419654e-01 -1.00657105e+00 2.41388336e-01 -6.21782899e-01 7.63662457e-01 -1.62257600e+00 -2.15813760e-02 -5.54423094e-01 1.23249210e-01 3.92347813e-01 2.55408853e-01 1.28435120e-01 4.43529129e-01 5.05802214e-01 -9.80951846e-01 2.18542546e-01 1.04597461e+00 -5.37489541e-02 -2.67813951e-01 9.67909098e-02 1.41654715e-01 9.90012050e-01 5.71948349e-01 -6.17823482e-01 -1.28784955e-01 -6.36551559e-01 -2.08128899e-01 4.45312768e-01 5.50670922e-01 -1.30665290e+00 6.77796125e-01 -2.53956288e-01 5.49269259e-01 -1.00495124e+00 7.17979372e-01 -1.52000225e+00 7.99330890e-01 5.00659168e-01 1.00505233e-01 1.37881890e-01 2.99488693e-01 8.92240822e-01 -9.67634544e-02 -3.62944365e-01 7.01219678e-01 -4.51312393e-01 -1.12171960e+00 2.76062161e-01 -5.68735242e-01 -7.86666453e-01 1.45924079e+00 -7.56194890e-01 2.28126392e-01 -1.10052906e-01 -1.11772704e+00 3.68174404e-01 1.34972095e+00 4.50896293e-01 6.86418474e-01 -1.10173094e+00 -1.20540768e-01 3.83886218e-01 6.83350563e-02 7.99325347e-01 1.32625207e-01 7.19555020e-01 -8.62130523e-01 1.87904954e-01 -4.78056259e-02 -1.14616585e+00 -1.34436226e+00 8.28136802e-01 4.69819278e-01 2.31529653e-01 -6.64028287e-01 4.67380375e-01 1.32861733e-01 -3.71090561e-01 3.31326097e-01 -2.86813140e-01 2.58075207e-01 -1.25968978e-01 2.86664724e-01 3.57476324e-01 8.39431435e-02 -1.19763327e+00 -5.01509368e-01 6.92572653e-01 3.47282439e-01 -3.63963187e-01 1.21164262e+00 -7.53020108e-01 -1.66550696e-01 8.56021881e-01 9.85339880e-01 -5.72028086e-02 -1.79640400e+00 -3.95862311e-01 2.92900652e-01 -7.69661844e-01 2.31410474e-01 -4.59594727e-01 -7.78926432e-01 4.83088225e-01 7.16396093e-01 7.07071498e-02 9.36818540e-01 2.62287021e-01 5.07094324e-01 3.62527430e-01 8.26538086e-01 -8.03169310e-01 2.15915926e-02 4.68875259e-01 3.75865549e-01 -1.08193326e+00 1.17000103e-01 -2.71059781e-01 -5.36338031e-01 1.15036869e+00 4.93068963e-01 6.22785389e-02 2.80584872e-01 5.24702847e-01 2.28099808e-01 -1.62677243e-01 -5.16938925e-01 -4.45419222e-01 -1.16448797e-01 7.41159976e-01 -3.95022333e-01 -1.59205288e-01 6.39059246e-01 -3.17882270e-01 2.15642184e-01 -1.72791779e-01 7.10606694e-01 1.44281423e+00 -9.37581956e-01 -1.02646303e+00 -8.21755111e-01 1.05447648e-02 6.52738009e-03 6.44663453e-01 -3.40109706e-01 7.86968470e-01 3.87237221e-01 9.50271845e-01 2.64492810e-01 -2.09472358e-01 4.80054438e-01 6.12505861e-02 6.67922199e-01 -8.23237062e-01 -2.34803140e-01 4.61284190e-01 -2.34193414e-01 -9.07344341e-01 -9.03440356e-01 -1.24476695e+00 -1.61394501e+00 9.45584290e-03 -5.87559700e-01 -7.00849341e-03 1.07764983e+00 9.17618036e-01 1.04210332e-01 4.05224383e-01 4.41260457e-01 -1.55480003e+00 -2.54731011e-02 -5.35422206e-01 -6.12641096e-01 1.74908504e-01 8.49348545e-01 -6.87815785e-01 -1.61748752e-01 4.40055490e-01]
[7.334436893463135, -2.3255414962768555]
957b0224-73ec-4e2b-b78f-6a4913a4b61c
using-duck-net-for-polyp-image-segmentation
null
null
https://www.nature.com/articles/s41598-023-36940-5#Abs1
https://www.nature.com/articles/s41598-023-36940-5.pdf
Using DUCK-Net for polyp image segmentation
This paper presents a novel supervised convolutional neural network architecture, “DUCK-Net”, capable of effectively learning and generalizing from small amounts of medical images to perform accurate segmentation tasks. Our model utilizes an encoder-decoder structure with a residual downsampling mechanism and a custom convolutional block to capture and process image information at multiple resolutions in the encoder segment. We employ data augmentation techniques to enrich the training set, thus increasing our model's performance. While our architecture is versatile and applicable to various segmentation tasks, in this study, we demonstrate its capabilities specifically for polyp segmentation in colonoscopy images. We evaluate the performance of our method on several popular benchmark datasets for polyp segmentation, Kvasir-SEG, CVC-ClinicDB, CVC-ColonDB, and ETIS-LARIBPOLYPDB showing that it achieves state-of-the-art results in terms of mean Dice coefficient, Jaccard index, Precision, Recall, and Accuracy. Our approach demonstrates strong generalization capabilities, achieving excellent performance even with limited training data.
['Catalin Craciun', 'Darius Peteleaza', 'Razvan-Gabriel Dumitru']
2023-06-16
null
null
null
nature-scientific-reports-2023-6
['polyp-segmentation']
['computer-vision']
[ 3.37640822e-01 2.33364761e-01 -2.67903268e-01 -4.64719564e-01 -8.90821815e-01 -4.18675125e-01 1.00620776e-01 5.87893546e-01 -6.41499162e-01 4.69183981e-01 2.37183403e-02 -6.44080698e-01 5.90176657e-02 -7.47591317e-01 -7.51775444e-01 -4.87775922e-01 -3.32387716e-01 3.18012089e-01 3.40647489e-01 -9.75633189e-02 -6.14501238e-02 2.77945191e-01 -8.53881538e-01 6.97077155e-01 1.08647442e+00 1.06724155e+00 1.96657658e-01 9.84712601e-01 3.28207761e-01 8.96317422e-01 -3.81266117e-01 -4.88627136e-01 2.35140726e-01 -3.28511029e-01 -9.61094975e-01 -7.51425177e-02 3.46145183e-01 -4.03064519e-01 -2.62109518e-01 7.90571630e-01 5.46094358e-01 -2.31125802e-01 6.15145028e-01 -2.25702941e-01 -7.15895414e-01 6.42741144e-01 -3.75252545e-01 2.81388074e-01 8.50436687e-02 1.73897997e-01 6.46656632e-01 -3.62514168e-01 5.56750238e-01 6.70372903e-01 1.26737702e+00 4.43775177e-01 -1.02263248e+00 -2.87851989e-01 -1.63751826e-01 -4.40551102e-01 -1.09997284e+00 1.41308010e-01 1.31235078e-01 -5.20496011e-01 7.95722663e-01 2.83498913e-01 7.09081888e-01 8.92684221e-01 4.73186344e-01 9.15962160e-01 7.42348731e-01 -2.86554277e-01 1.34285897e-01 -2.64163665e-03 1.17907003e-01 9.60664511e-01 3.18972409e-01 2.40176152e-02 6.18788190e-02 1.93054713e-02 1.00172412e+00 2.62973178e-03 -4.35784280e-01 -3.72596711e-01 -1.30544090e+00 8.66821945e-01 9.84538019e-01 2.99459666e-01 -4.15869117e-01 5.63956164e-02 7.93843985e-01 1.61662757e-01 3.66746396e-01 7.78834641e-01 -3.72419387e-01 1.10963359e-01 -1.13932621e+00 -7.18993247e-02 9.06756520e-01 1.05381441e+00 -2.18723789e-02 -3.40219706e-01 -3.18478554e-01 7.42429435e-01 -8.08201581e-02 9.35475901e-02 1.00730228e+00 -7.70476282e-01 3.40442657e-01 7.21059382e-01 -9.22236368e-02 -6.55438721e-01 -7.99607337e-01 -8.41334999e-01 -1.12626600e+00 -4.25094785e-03 2.01378003e-01 -2.05481395e-01 -1.24667764e+00 1.32475162e+00 2.64400709e-03 4.95933788e-03 2.06177130e-01 8.47908258e-01 1.27335298e+00 1.40108034e-01 1.94850132e-01 2.08583280e-01 1.41152871e+00 -1.17926192e+00 -3.71822298e-01 -1.91757590e-01 7.71370351e-01 -5.42137563e-01 8.43319654e-01 5.23271143e-01 -1.28223300e+00 -5.66043675e-01 -1.17370224e+00 -2.90828407e-01 -3.07209998e-01 6.35122299e-01 6.66480303e-01 5.41977108e-01 -1.03127420e+00 7.36533523e-01 -1.24345076e+00 -3.69414359e-01 8.53926301e-01 4.86250132e-01 -2.97901481e-01 -1.53943688e-01 -9.08466399e-01 7.39479780e-01 5.31749070e-01 -9.88388583e-02 -8.81128371e-01 -9.31969225e-01 -9.67969537e-01 1.97839662e-01 1.18244633e-01 -9.97712076e-01 1.48811209e+00 -7.67628968e-01 -1.34819269e+00 9.59222257e-01 5.14715433e-01 -1.11111951e+00 7.83236206e-01 -5.07160742e-03 -1.54573128e-01 4.12663162e-01 -6.24421909e-02 1.03241706e+00 4.21264231e-01 -7.82337904e-01 -6.07699633e-01 -1.46876276e-01 8.79007578e-02 1.62947387e-01 -1.66106552e-01 -4.26077902e-01 -5.67154884e-01 -8.41356993e-01 9.21181366e-02 -9.13637280e-01 -7.69464076e-01 1.51474923e-01 -3.70997310e-01 3.19030762e-01 3.83869708e-01 -8.45213532e-01 1.12931418e+00 -2.24911141e+00 -1.72026932e-01 1.18695058e-01 3.73657674e-01 5.87102056e-01 -8.21538717e-02 9.92272142e-03 -1.14212841e-01 1.13650352e-01 -6.62532687e-01 -2.20149860e-01 -5.40302277e-01 1.50236264e-01 3.30023468e-01 4.66478348e-01 2.59336054e-01 1.06965470e+00 -9.79273915e-01 -4.45383221e-01 4.70374376e-01 5.66580176e-01 -7.96491325e-01 1.49012744e-01 -4.01270725e-02 4.79097515e-01 -3.29626590e-01 5.53459167e-01 5.87382257e-01 -8.74971390e-01 1.95329890e-01 -2.40370974e-01 1.12314448e-01 1.59326285e-01 -6.39135957e-01 2.18150377e+00 -6.78275645e-01 5.71720183e-01 1.31426856e-01 -9.92550135e-01 7.00678289e-01 2.75903046e-01 5.95596373e-01 -5.70706725e-01 2.61675417e-01 3.52632165e-01 1.47216797e-01 -6.29202187e-01 3.47251445e-01 1.88196182e-01 -4.97099245e-03 9.22686793e-03 1.55347034e-01 -1.52069151e-01 4.62848544e-01 3.51533815e-02 1.08294261e+00 -1.11555509e-01 6.20791435e-01 -4.47578371e-01 6.80246234e-01 1.73808977e-01 1.29697293e-01 8.18151355e-01 -2.43273959e-01 9.27891374e-01 4.41040486e-01 -6.77452087e-01 -1.01756597e+00 -1.07262015e+00 -4.88765746e-01 6.16813779e-01 7.95994326e-02 -2.28815421e-01 -8.48273337e-01 -8.82440031e-01 -1.89727768e-01 3.94411504e-01 -9.01932478e-01 -7.51413181e-02 -5.95266700e-01 -1.04532254e+00 6.35577321e-01 7.61699677e-01 6.30067408e-01 -9.45333838e-01 -1.02284503e+00 3.57676089e-01 -8.82969201e-02 -1.17342293e+00 -3.65922987e-01 3.19487989e-01 -1.04070449e+00 -1.38632381e+00 -1.01210654e+00 -1.23447418e+00 7.83550382e-01 -1.50859177e-01 1.40140581e+00 1.04709566e-02 -7.40840614e-01 3.24046053e-02 -3.21423978e-01 -3.79821450e-01 -6.88898921e-01 4.51106191e-01 -7.26118147e-01 -5.39442003e-01 1.04948536e-01 -1.59963459e-01 -1.16222739e+00 1.06841281e-01 -9.81362939e-01 1.59924433e-01 8.58146369e-01 1.07975912e+00 7.67165720e-01 -4.00652587e-01 2.30859131e-01 -1.07903790e+00 6.74575746e-01 -3.06184620e-01 -5.32260299e-01 1.53866678e-01 -5.46784520e-01 -3.00472438e-01 7.83476174e-01 -1.68347061e-01 -7.87919700e-01 2.03326672e-01 -3.88419807e-01 -4.42015864e-02 -8.68825912e-02 5.76688647e-01 6.65792763e-01 -1.73411459e-01 9.75471437e-01 2.06885636e-01 3.21864545e-01 -5.23192108e-01 2.12047309e-01 6.68984056e-01 8.37069750e-01 4.44463547e-03 -1.51535068e-02 5.69199026e-01 -3.69763188e-02 -3.94896150e-01 -8.55014086e-01 -6.04301035e-01 -6.04594290e-01 3.56768608e-01 1.03890967e+00 -1.11019123e+00 -5.49379945e-01 3.83645415e-01 -7.90053308e-01 -2.78258473e-01 -5.09389877e-01 6.90835416e-01 -4.76459086e-01 2.38338873e-01 -1.26055825e+00 1.59994930e-01 -9.20445204e-01 -1.68780661e+00 1.01938677e+00 2.41446286e-01 -6.52242154e-02 -1.23552120e+00 6.32004533e-03 6.02922169e-03 6.56437635e-01 5.52580118e-01 8.19903076e-01 -8.81056190e-01 -4.36495125e-01 -4.85082775e-01 -3.61363441e-01 6.74523711e-01 1.73407182e-01 -2.61358142e-01 -5.68052292e-01 -5.13098896e-01 -1.80274412e-01 -3.21259648e-01 1.08364260e+00 9.03536499e-01 1.81718802e+00 -4.92148586e-02 -4.43990022e-01 1.08439136e+00 1.58327651e+00 1.39199361e-01 5.60364068e-01 2.87997603e-01 3.70473862e-01 8.05591047e-02 1.40324742e-01 2.21621811e-01 2.59333313e-01 1.22713834e-01 5.20197034e-01 -5.95937371e-01 -4.25318211e-01 -1.50384696e-03 -3.33067149e-01 7.40285575e-01 1.12783223e-01 -2.23500177e-01 -1.04181230e+00 7.07072318e-01 -1.51399922e+00 -4.20317799e-01 -5.74625954e-02 1.86693931e+00 8.03277910e-01 1.05050660e-03 -9.50722843e-02 -2.43422359e-01 4.68615085e-01 -1.42952710e-01 -5.12577772e-01 -4.68709826e-01 2.17257753e-01 4.71831113e-01 7.34648347e-01 2.19626650e-01 -1.49484587e+00 5.47949731e-01 6.99758911e+00 5.22971451e-01 -1.14065051e+00 -5.76710254e-02 1.15378964e+00 3.84137072e-02 3.21730912e-01 -8.12887728e-01 -2.57000715e-01 3.22828919e-01 8.53924632e-01 2.71674931e-01 2.73012128e-02 9.49930727e-01 -2.11971581e-01 6.10853285e-02 -1.13938904e+00 7.10031092e-01 -4.63479497e-02 -1.78701806e+00 -8.21177214e-02 -2.50423074e-01 8.48862410e-01 4.16931599e-01 3.91434193e-01 2.51182824e-01 2.88742304e-01 -1.32383156e+00 -5.11047281e-02 1.96079656e-01 1.18157470e+00 -5.66801310e-01 1.18215346e+00 -1.30813234e-02 -7.69380569e-01 1.01873249e-01 -1.74928591e-01 4.73533481e-01 -1.86327428e-01 4.73995268e-01 -1.24078488e+00 6.64352715e-01 6.78232074e-01 8.63645434e-01 -7.47751951e-01 1.45194006e+00 -1.57739501e-02 4.59956676e-01 -1.31134495e-01 2.05662638e-01 6.69529080e-01 1.32481784e-01 8.19823667e-02 1.59837639e+00 3.96113843e-01 -7.65710399e-02 1.69396713e-01 5.90068817e-01 -3.21974576e-01 2.81998694e-01 -1.41492441e-01 2.82016754e-01 8.78900662e-02 1.22686100e+00 -8.53167772e-01 -6.27754629e-01 -5.46976328e-01 8.78280520e-01 -2.34402586e-02 3.20202373e-02 -8.20955157e-01 -4.80310321e-01 1.27638265e-01 -8.29799473e-02 5.00096977e-01 3.15546781e-01 -5.90330958e-01 -9.72220659e-01 -1.10478625e-01 -1.03557372e+00 6.31811202e-01 -4.43562299e-01 -1.18442011e+00 8.75051379e-01 -2.88172573e-01 -1.40971935e+00 -1.64408386e-01 -9.28664267e-01 -3.18958640e-01 7.88290620e-01 -1.63868737e+00 -1.09262753e+00 -5.63062549e-01 3.18424553e-01 5.06264150e-01 -2.52950966e-01 9.58202362e-01 3.06196839e-01 -2.66415536e-01 7.18085051e-01 3.65205020e-01 5.44965982e-01 7.12306559e-01 -1.47553086e+00 5.94482780e-01 5.30821145e-01 -3.22631836e-01 4.65639383e-01 2.49921396e-01 -3.13491821e-01 -8.04206252e-01 -1.29484296e+00 2.88920760e-01 -4.11288738e-02 4.16932404e-01 4.44191061e-02 -7.63607442e-01 7.55874991e-01 3.14877480e-01 2.71403491e-01 8.70531082e-01 -2.05573112e-01 -2.06733972e-01 9.35322121e-02 -1.45428443e+00 4.46204394e-01 5.51586747e-01 -3.86703201e-02 -5.07125556e-01 5.86474419e-01 8.74312162e-01 -1.13441789e+00 -1.30006552e+00 6.88896537e-01 4.82962102e-01 -1.03790772e+00 1.11560583e+00 -4.88536954e-01 1.01875985e+00 2.31033824e-02 1.85812369e-01 -1.35444283e+00 -2.15183228e-01 -2.41516352e-01 1.25154614e-01 2.63807476e-01 7.19914138e-01 -4.80986178e-01 9.22444701e-01 1.98243111e-01 -5.65458298e-01 -1.26650298e+00 -6.52941406e-01 -2.67738223e-01 4.42934066e-01 -2.83972733e-02 6.18981302e-01 7.92354584e-01 -6.30379245e-02 -1.49063766e-01 6.80889934e-02 7.45543242e-02 4.21933383e-01 6.40474409e-02 3.12132329e-01 -7.61394620e-01 -3.84729683e-01 -4.84286457e-01 -3.33902329e-01 -1.07386971e+00 -4.31681752e-01 -1.06655884e+00 -1.03531033e-01 -1.87018228e+00 1.65317789e-01 -3.35978061e-01 -3.74396622e-01 3.19125533e-01 -1.61669180e-01 3.97956103e-01 -3.99471447e-02 2.39698961e-02 -5.66348076e-01 -1.19130738e-01 1.70084691e+00 -2.50042051e-01 -2.73016632e-01 2.06116676e-01 -6.77980006e-01 7.77219474e-01 8.34007025e-01 -1.32638112e-01 -2.68353224e-01 -7.13385820e-01 -1.84571713e-01 1.28110588e-01 2.75079519e-01 -1.30742323e+00 4.58802655e-02 4.70545292e-01 7.72174418e-01 -5.11070788e-01 -8.69467780e-02 -6.42955244e-01 -2.33240768e-01 1.11065114e+00 -8.03350151e-01 4.43822332e-02 3.54603410e-01 3.32227767e-01 -4.07513946e-01 -2.20242351e-01 1.16021919e+00 -5.10414839e-01 -4.94944721e-01 4.51267242e-01 -2.47148007e-01 9.43632871e-02 9.38555300e-01 1.27294943e-01 -3.70783687e-01 -1.65035520e-02 -1.00457478e+00 2.92815894e-01 3.48199666e-01 6.05778955e-02 5.75822413e-01 -9.65170264e-01 -8.80771339e-01 2.95224726e-01 1.73051417e-01 4.75834042e-01 3.19158047e-01 9.79471087e-01 -1.41207445e+00 5.81218660e-01 -1.20124549e-01 -9.39805090e-01 -9.57997382e-01 5.98363161e-01 7.69900918e-01 -7.28564978e-01 -9.50952828e-01 1.01958370e+00 1.99923828e-01 -4.80609387e-01 2.16259882e-01 -9.73100960e-01 -8.34525377e-02 -3.26269120e-01 5.09657204e-01 1.69881061e-02 3.68580669e-01 5.93371363e-03 -2.04949602e-01 1.41739592e-01 -4.64566261e-01 5.02821267e-01 1.19663262e+00 1.35370567e-01 4.51728962e-02 -1.85336992e-01 1.41195345e+00 -4.71698105e-01 -1.14465797e+00 -1.59286603e-01 -1.40136957e-01 -1.13431402e-01 1.69714987e-01 -1.30124772e+00 -1.10522604e+00 8.78440261e-01 1.00000870e+00 9.81456712e-02 1.27166867e+00 -1.56358540e-01 1.01609528e+00 2.61153668e-01 -6.92977905e-02 -7.97357917e-01 4.46219034e-02 3.79532129e-01 5.35248041e-01 -1.50151742e+00 1.00013919e-01 -4.36033279e-01 -6.51205361e-01 1.17495942e+00 4.38073009e-01 -4.87377346e-01 5.94565511e-01 4.22912747e-01 2.07806602e-01 -1.29580602e-01 -3.78211200e-01 1.08002692e-01 4.81321931e-01 4.27160323e-01 7.53475189e-01 2.12553352e-01 -3.58911663e-01 3.76906991e-01 -2.43073627e-01 3.11382920e-01 5.79753816e-01 8.68234754e-01 -2.55505502e-01 -7.14469492e-01 1.03472792e-01 7.80088842e-01 -1.08489561e+00 -3.34726036e-01 9.51207429e-02 1.12113190e+00 -1.30949160e-02 4.27975059e-01 1.59204185e-01 1.11212403e-01 2.56812841e-01 -4.22266126e-01 4.65727001e-01 -6.16472960e-01 -1.06425524e+00 5.68734221e-02 3.49029228e-02 -5.22989273e-01 -4.01926160e-01 -3.03672791e-01 -1.06433976e+00 1.89161956e-01 7.02139661e-02 4.80183326e-02 7.44108915e-01 5.53069532e-01 2.29760036e-01 1.14922929e+00 1.24393649e-01 -3.76699626e-01 -6.42521322e-01 -9.19359624e-01 -2.63576299e-01 5.25476336e-01 5.90497077e-01 -3.39663550e-02 8.26741010e-02 8.46248567e-02]
[14.587197303771973, -2.7563531398773193]
64741661-135e-4adf-9067-d727b745ffcd
qualiassistant-extracting-qualia-structures
null
null
https://aclanthology.org/2022.argmining-1.19
https://aclanthology.org/2022.argmining-1.19.pdf
QualiAssistant: Extracting Qualia Structures from Texts
In this paper, we present QualiAssistant, a free and open-source system written in Java for identification and extraction of Qualia structures from any natural language texts having many application scenarios such as argument mining or creating dictionaries. It answers the call for a Qualia bootstrapping tool with a ready-to-use system that can be gradually filled by the community with patterns in multiple languages. Qualia structures express the meaning of lexical items. They describe, e.g., of what kind the item is (formal role), what it includes (constitutive role), how it is brought about (agentive role), and what it is used for (telic role). They are also valuable for various Information Retrieval and NLP tasks. Our application requires search patterns for Qualia structures consisting of POS tag sequences as well as the dataset the user wants to search for Qualias. Samples for both are provided alongside this paper. While samples are in German, QualiAssistant can process all languages for which constituency trees can be generated and patterns are available. Our provided patterns follow a high-precision low-recall design aiming to generate automatic annotations for text mining but can be exchanged easily for other purposes. Our evaluation shows that QualiAssistant is a valuable and reliable tool for finding Qualia structures in unstructured texts.
['Ralf Schenkel', 'Björn Metzler', 'Markus Nilles', 'Lorik Dumani', 'Manuel Biertz']
null
null
null
null
argmining-acl-2022-10
['argument-mining']
['natural-language-processing']
[ 1.26475140e-01 4.80798930e-01 -4.51410353e-01 -2.67392933e-01 -6.40514374e-01 -1.08646154e+00 8.96617889e-01 8.38344991e-01 -5.61315894e-01 1.04107916e+00 5.61191320e-01 -6.34622157e-01 -4.44723278e-01 -8.80704403e-01 -2.85182983e-01 -2.24004954e-01 4.29977179e-01 1.22417974e+00 4.56103593e-01 -6.38201594e-01 2.58470774e-01 4.39352505e-02 -1.86312795e+00 7.69863605e-01 6.74190164e-01 7.08982468e-01 4.79922056e-01 2.20375001e-01 -7.31107831e-01 8.31501007e-01 -7.12026656e-01 -6.58637166e-01 3.75012308e-02 -1.00433424e-01 -1.47160864e+00 -4.01229918e-01 1.89213883e-02 4.44742918e-01 6.42763972e-01 9.32180583e-01 1.79188102e-01 -2.94735402e-01 6.84207976e-01 -9.83375072e-01 1.63426042e-01 1.29536068e+00 7.66442716e-02 1.23635672e-01 8.55237246e-01 -3.91257077e-01 1.45687687e+00 -7.33286381e-01 1.28059506e+00 1.30753338e+00 3.98831546e-01 3.80079478e-01 -9.61288989e-01 -3.30614924e-01 -2.77738780e-01 -8.07636827e-02 -7.62732208e-01 -3.03958297e-01 4.18089896e-01 -5.50337732e-01 1.19273567e+00 7.34944224e-01 5.47532856e-01 1.00146258e+00 -4.11377624e-02 7.86309898e-01 1.23248780e+00 -1.16304696e+00 -5.00600860e-02 3.22249919e-01 7.03534603e-01 3.28897059e-01 5.25099993e-01 -3.80092710e-01 -4.16033924e-01 -3.87495548e-01 2.03743964e-01 -3.80669773e-01 2.73248553e-01 1.13011830e-01 -1.43281877e+00 7.52553523e-01 -4.37776089e-01 9.97428060e-01 -3.84491742e-01 -2.94637740e-01 9.62770760e-01 5.03681242e-01 3.47693288e-03 8.57018709e-01 -1.06760824e+00 -5.00513911e-01 -6.92490757e-01 5.11142135e-01 1.13970125e+00 9.62473094e-01 5.89418352e-01 -4.40446079e-01 -1.10492975e-01 9.58798051e-01 3.10642600e-01 3.24759603e-01 9.63958979e-01 -4.90862489e-01 5.20948648e-01 1.30393231e+00 2.94576943e-01 -4.27058578e-01 -5.17035007e-01 2.46739127e-02 -1.92656249e-01 -2.73269922e-01 7.37902462e-01 -5.49706519e-02 -4.97376829e-01 1.48110938e+00 5.05230844e-01 -7.42001593e-01 2.90508121e-01 4.02962267e-01 1.30916357e+00 6.53777957e-01 2.41672993e-01 -5.02713084e-01 1.98376000e+00 -3.29426527e-01 -8.72194290e-01 -1.23221084e-01 8.96584809e-01 -1.03966045e+00 1.30803514e+00 5.15906870e-01 -1.18756092e+00 -1.62845060e-01 -9.23266053e-01 -1.16182365e-01 -8.45463991e-01 3.31349969e-02 5.78069985e-01 7.02432871e-01 -4.58919942e-01 4.28068519e-01 -4.39282149e-01 -4.40348536e-01 -2.28509277e-01 1.60512999e-01 -4.00655061e-01 6.27976894e-01 -1.43144429e+00 1.00108397e+00 8.70550692e-01 -5.08585811e-01 -1.90049797e-01 -3.85061413e-01 -8.61226141e-01 -1.41920656e-01 6.66594923e-01 -2.65573800e-01 1.32374537e+00 -1.09264171e+00 -1.01751959e+00 1.63398790e+00 -1.58311099e-01 -6.50417924e-01 2.84895301e-01 -9.19124633e-02 -7.09599078e-01 -1.65699124e-01 8.13491702e-01 1.19768687e-01 4.98365968e-01 -8.13857675e-01 -1.03138959e+00 -2.87454188e-01 1.91314027e-01 5.00071049e-02 -1.27908975e-01 8.99879575e-01 -1.04655720e-01 -7.16524184e-01 4.42540050e-02 -7.11002350e-01 1.82687670e-01 -7.26634979e-01 -4.05717403e-01 -8.66110802e-01 6.66190386e-01 -6.15685940e-01 1.86543107e+00 -1.69409490e+00 -1.02914266e-01 5.47003329e-01 1.52774574e-02 3.96959364e-01 2.67932862e-01 9.62182045e-01 -2.19017014e-01 5.87673366e-01 -1.01573020e-01 5.26583552e-01 4.52887714e-01 6.13985360e-01 -4.07615244e-01 -7.37335905e-02 4.65200916e-02 6.67764843e-01 -9.21691000e-01 -7.31718242e-01 5.57117388e-02 -1.15147255e-01 -2.24039078e-01 -6.80029690e-02 -5.53148687e-01 -1.75883234e-01 -5.23733616e-01 4.52072710e-01 2.09367886e-01 -4.95406017e-02 8.23505342e-01 -7.47288764e-02 -5.56126654e-01 1.25133324e+00 -1.47720408e+00 1.07959986e+00 -6.11027300e-01 2.46705607e-01 3.55425812e-02 -8.43620241e-01 8.86734426e-01 4.75101173e-01 2.68757343e-01 -4.47717667e-01 2.28602752e-01 8.40948820e-01 2.50812501e-01 -4.19246197e-01 6.89592183e-01 -2.31308937e-01 -5.46470940e-01 6.06536627e-01 2.48322841e-02 -8.32966268e-02 1.17713571e+00 1.10956311e-01 1.09191668e+00 2.21688554e-01 1.21679246e+00 -7.66283154e-01 9.65451181e-01 4.36805665e-01 5.54060042e-01 3.47480595e-01 3.27982426e-01 -4.68037650e-02 5.32872558e-01 -5.62096775e-01 -1.05457819e+00 -8.24597299e-01 -6.54637814e-01 1.12919366e+00 -2.45324239e-01 -1.08930981e+00 -5.03276110e-01 -1.00381839e+00 -3.34060103e-01 6.01548254e-01 -5.07391691e-01 7.68542469e-01 -8.36214304e-01 -1.48687541e-01 5.87925792e-01 1.22276470e-01 1.76590934e-01 -1.77208352e+00 -9.68707800e-01 6.40004158e-01 -5.05262733e-01 -7.20814526e-01 -2.71888077e-02 7.08249748e-01 -3.54639649e-01 -1.43629634e+00 1.59079060e-01 -8.93114269e-01 8.07107687e-02 -3.72275054e-01 1.59628046e+00 3.75079930e-01 -1.67230681e-01 6.44088909e-02 -7.23046005e-01 -9.32148397e-01 -8.93878341e-01 2.09839761e-01 -1.20196335e-01 -4.07517493e-01 6.87012851e-01 -4.05432135e-01 1.03624143e-01 2.02113047e-01 -1.04322398e+00 -8.76095071e-02 1.62940666e-01 8.42520416e-01 6.86773181e-01 1.94271468e-02 5.39879858e-01 -1.36867869e+00 9.93775725e-01 -5.26920736e-01 -3.64899904e-01 2.77895957e-01 -6.06665671e-01 4.70101804e-01 6.35569692e-01 -1.56819001e-01 -1.01008260e+00 -2.39762574e-01 -6.85441434e-01 7.69082785e-01 -3.50874096e-01 9.79590297e-01 -4.44817901e-01 8.79472375e-01 9.61328208e-01 -2.26068035e-01 -2.55665064e-01 -4.48874682e-01 2.10788339e-01 8.24517488e-01 2.07759753e-01 -1.12820745e+00 5.90447664e-01 2.35203519e-01 -1.94356605e-01 -8.24538767e-01 -1.02994204e+00 -9.67300057e-01 -5.51493585e-01 -5.83902746e-02 5.46737015e-01 -5.24979532e-01 -5.24066269e-01 -3.88417765e-02 -1.08089948e+00 -2.44545490e-01 -7.66071677e-01 5.61839417e-02 -5.42129934e-01 2.12129578e-01 -3.00260603e-01 -7.40331411e-01 -7.37781644e-01 -5.58123410e-01 9.79919195e-01 4.55292016e-02 -1.10222721e+00 -1.16937232e+00 2.14405462e-01 1.97863221e-01 -8.84183347e-02 1.71496734e-01 1.13891637e+00 -1.16290283e+00 2.38920063e-01 1.01009689e-01 1.43394500e-01 8.39207973e-03 1.46639779e-01 9.49443690e-03 -4.26563859e-01 2.60857910e-01 -2.84403056e-01 -5.06530344e-01 2.27282360e-01 -1.23166023e-02 6.32531404e-01 -7.66195953e-01 -2.63548404e-01 -1.16012670e-01 1.16705096e+00 2.89635658e-01 7.41463661e-01 1.07662666e+00 1.33860447e-02 9.56749797e-01 1.06946778e+00 3.59148294e-01 2.68933356e-01 7.51351178e-01 1.51114494e-01 2.28946120e-01 1.69437267e-02 -3.27195585e-01 4.89762455e-01 6.65885568e-01 4.30469364e-02 9.68773197e-03 -1.09239006e+00 7.60935307e-01 -1.91103792e+00 -1.35060668e+00 -7.86617756e-01 1.90307641e+00 1.55994296e+00 3.58361065e-01 4.10651833e-01 7.96243072e-01 5.05032659e-01 -1.36667788e-01 2.16748238e-01 -9.02999103e-01 -3.75756264e-01 8.59566092e-01 3.49082537e-02 4.03329045e-01 -1.04302025e+00 1.14455235e+00 5.11773109e+00 1.09996963e+00 -8.07745218e-01 1.09590456e-01 -7.67180398e-02 5.90788901e-01 -6.30576968e-01 3.94704401e-01 -1.29511833e+00 3.27268451e-01 9.48609948e-01 -3.85240287e-01 2.96687670e-02 9.73902583e-01 1.13676339e-01 -1.23195261e-01 -9.51213658e-01 6.34066105e-01 -3.21203202e-01 -1.63705528e+00 8.67967010e-02 3.57171334e-02 1.73555017e-01 1.66124459e-02 -6.02471888e-01 2.42116183e-01 5.65678477e-01 -6.12323761e-01 1.30116165e+00 6.09211884e-02 8.20304215e-01 -4.47594404e-01 7.48172700e-01 5.52628219e-01 -1.25826097e+00 5.35756424e-02 -8.56885239e-02 -2.70277023e-01 1.57280013e-01 5.30147254e-01 -1.03076267e+00 5.26174724e-01 6.10028982e-01 4.62938964e-01 -5.11029303e-01 6.78225756e-01 -6.38695121e-01 9.78291512e-01 -5.44306040e-01 -5.70826173e-01 2.81783521e-01 -2.40739748e-01 8.44869792e-01 1.62615466e+00 2.18341991e-01 -1.26103133e-01 2.23221570e-01 3.73208851e-01 2.35538989e-01 9.25262868e-01 -5.94320297e-01 -1.32822230e-01 6.12504601e-01 1.43932760e+00 -9.82868850e-01 -5.36004364e-01 2.11790539e-02 3.01322222e-01 3.76926847e-02 -3.28578979e-01 -3.15275937e-01 -4.72888470e-01 2.51212746e-01 7.95176029e-01 5.53716533e-02 7.54032657e-02 -1.28112257e-01 -1.00579906e+00 1.06982104e-01 -1.35666871e+00 9.69608605e-01 -6.06590152e-01 -1.25254297e+00 6.81507051e-01 4.33274031e-01 -1.07708144e+00 -7.32964396e-01 -9.75880444e-01 -3.46580833e-01 6.19443476e-01 -1.00444484e+00 -1.21520543e+00 2.82780826e-01 6.83631182e-01 5.70902467e-01 -2.66465336e-01 1.11164045e+00 1.16375625e-01 -6.06959127e-02 1.17963694e-01 -5.03806174e-01 3.33712220e-01 5.79239428e-01 -1.44753277e+00 2.40536425e-02 7.42262244e-01 6.94269657e-01 1.09871435e+00 8.71954322e-01 -7.78234720e-01 -9.07609105e-01 -6.34248614e-01 1.78498971e+00 -5.41722000e-01 1.12875104e+00 -2.63730168e-01 -6.77545905e-01 5.07432401e-01 4.43078190e-01 -5.89098871e-01 8.99661303e-01 4.67229247e-01 -3.55560511e-01 1.45581543e-01 -1.01389623e+00 5.11524081e-01 9.72996294e-01 -2.35651195e-01 -1.31302607e+00 2.51556277e-01 3.35491210e-01 -2.75771290e-01 -6.64210737e-01 1.15867093e-01 4.27529037e-01 -8.38372290e-01 4.92900431e-01 -7.36664176e-01 3.10452789e-01 -5.52203596e-01 -1.33373842e-01 -9.01178718e-01 2.20224753e-01 -1.03075111e+00 1.47827089e-01 1.38250387e+00 1.01784325e+00 -6.35406494e-01 3.80974978e-01 2.75333613e-01 -2.36144125e-01 -2.18977273e-01 -1.06042194e+00 -6.57982111e-01 -1.32059092e-02 -7.70061135e-01 6.24290347e-01 1.05042017e+00 5.58988035e-01 1.05975759e+00 1.32982045e-01 -5.28733909e-01 7.64551759e-02 3.25617433e-01 7.51284719e-01 -1.74879026e+00 -2.75065690e-01 -3.38228166e-01 -8.17074627e-02 -5.91417789e-01 2.25686222e-01 -1.32253003e+00 -2.97415674e-01 -1.65546262e+00 -2.94864401e-02 -8.59063745e-01 3.53930950e-01 8.50182176e-01 2.50514001e-01 -1.30680144e-01 3.48028820e-03 2.52789676e-01 -4.06511486e-01 -1.59868777e-01 7.26684451e-01 -2.90061235e-02 -3.68640572e-01 -7.09305629e-02 -7.58657277e-01 1.07963908e+00 7.47371614e-01 -6.68084025e-01 -1.80837065e-01 2.43561074e-01 1.20835352e+00 -1.67394429e-01 -8.32696185e-02 -5.40788889e-01 -2.07578287e-01 -4.19332415e-01 -2.21623570e-01 -6.47941053e-01 -1.93392515e-01 -8.28914225e-01 2.27346301e-01 4.92144495e-01 -1.30542159e-01 3.05464089e-01 1.13084629e-01 -2.61838883e-01 -4.08746868e-01 -1.14543295e+00 4.36196774e-01 -4.07432497e-01 -6.03920698e-01 -2.39732757e-01 -7.18201101e-01 5.59462368e-01 7.39220858e-01 -3.85180950e-01 -2.86228389e-01 -2.48574163e-03 -6.43296182e-01 2.57562436e-02 3.81295592e-01 3.22105914e-01 3.96648467e-01 -1.07871187e+00 -6.63825154e-01 2.03632638e-02 4.68544632e-01 -2.88230628e-01 -5.47228634e-01 5.11194944e-01 -5.37495077e-01 3.11160028e-01 -1.78126305e-01 -1.87939018e-01 -1.55250669e+00 3.93566668e-01 -1.37464598e-01 -1.06613338e+00 -5.28978765e-01 3.30602318e-01 -3.06715041e-01 -6.01986468e-01 -1.62918761e-01 -4.86849606e-01 -8.80612969e-01 6.60865188e-01 4.90473717e-01 -7.84436986e-02 3.57616186e-01 -7.91484118e-01 -4.97015953e-01 1.03392750e-01 2.53268462e-02 -4.66667980e-01 1.40575564e+00 8.15057978e-02 -6.36325121e-01 5.93344152e-01 4.59352285e-01 8.83687019e-01 -1.24561675e-01 7.51721859e-02 7.81172574e-01 5.23517504e-02 -5.36126196e-01 -1.04464924e+00 -5.04714511e-02 2.35425293e-01 -2.41958201e-02 8.76247227e-01 6.41416788e-01 4.14227456e-01 5.01941741e-01 4.08093423e-01 5.78765035e-01 -1.47684956e+00 -1.67604655e-01 1.05784726e+00 1.23351097e+00 -7.59207547e-01 -1.56392097e-01 -5.40272772e-01 -6.18342698e-01 1.20438385e+00 7.65013248e-02 1.71030059e-01 7.62523413e-01 6.14187241e-01 4.64145802e-02 -6.11023724e-01 -7.72495687e-01 -6.37300730e-01 1.62044674e-01 6.76810145e-01 8.58431518e-01 3.14632416e-01 -1.51178551e+00 9.91218388e-01 -8.31342697e-01 -4.92867857e-01 4.93828654e-01 9.32376146e-01 -5.46939015e-01 -2.09420681e+00 -4.07520592e-01 5.52608728e-01 -9.81643140e-01 -3.47478539e-01 -7.22382009e-01 1.07798028e+00 4.78807867e-01 9.88080978e-01 -1.95683360e-01 4.81231585e-02 4.34415489e-01 2.40504518e-01 4.42732632e-01 -1.15575027e+00 -1.27483201e+00 1.65428549e-01 1.25919938e+00 -1.38856009e-01 -8.33066046e-01 -1.10811293e+00 -1.65223885e+00 1.51204959e-01 -3.07476848e-01 7.65130162e-01 5.33657312e-01 1.14112270e+00 -1.93151981e-01 -3.08967829e-02 4.23985049e-02 -5.72935939e-02 -2.81282157e-01 -1.12718165e+00 -3.14617336e-01 5.15442431e-01 -3.89770061e-01 -6.89217329e-01 -1.61613315e-01 2.81202376e-01]
[9.951732635498047, 9.560660362243652]
514da470-319f-4d37-b531-55e69b8a396a
dual-encoder-decoder-based-generative
1909.08797
null
https://arxiv.org/abs/1909.08797v1
https://arxiv.org/pdf/1909.08797v1.pdf
Dual Encoder-Decoder based Generative Adversarial Networks for Disentangled Facial Representation Learning
To learn disentangled representations of facial images, we present a Dual Encoder-Decoder based Generative Adversarial Network (DED-GAN). In the proposed method, both the generator and discriminator are designed with deep encoder-decoder architectures as their backbones. To be more specific, the encoder-decoder structured generator is used to learn a pose disentangled face representation, and the encoder-decoder structured discriminator is tasked to perform real/fake classification, face reconstruction, determining identity and estimating face pose. We further improve the proposed network architecture by minimising the additional pixel-wise loss defined by the Wasserstein distance at the output of the discriminator so that the adversarial framework can be better trained. Additionally, we consider face pose variation to be continuous, rather than discrete in existing literature, to inject richer pose information into our model. The pose estimation task is formulated as a regression problem, which helps to disentangle identity information from pose variations. The proposed network is evaluated on the tasks of pose-invariant face recognition (PIFR) and face synthesis across poses. An extensive quantitative and qualitative evaluation carried out on several controlled and in-the-wild benchmarking datasets demonstrates the superiority of the proposed DED-GAN method over the state-of-the-art approaches.
['Xiao-Jun Wu', 'Zhen-Hua Feng', 'Cong Hu', 'Josef Kittler']
2019-09-19
null
null
null
null
['robust-face-recognition']
['computer-vision']
[ 4.35854346e-01 4.62840289e-01 1.43785134e-01 -4.47397411e-01 -8.49209845e-01 -5.70157111e-01 6.63575649e-01 -8.54319334e-01 2.23035365e-02 6.47192657e-01 7.96039030e-02 2.46179700e-01 3.23524833e-01 -6.97894156e-01 -8.35952103e-01 -1.00448048e+00 1.72084183e-01 4.34897691e-01 -5.86098194e-01 -2.34359019e-02 -2.33428836e-01 6.27567708e-01 -1.37350023e+00 5.33330701e-02 5.10792136e-01 1.18764663e+00 -3.27550143e-01 4.72505748e-01 5.31871021e-01 7.15757191e-01 -6.10092402e-01 -9.59852874e-01 5.74326158e-01 -6.70679688e-01 -2.21796364e-01 2.37118393e-01 6.51059270e-01 -5.60765386e-01 -6.62788928e-01 1.05223835e+00 8.61085773e-01 -2.88691133e-01 6.49683237e-01 -1.43744671e+00 -8.44063222e-01 1.84731960e-01 -5.88019490e-01 -3.08013529e-01 3.16627622e-01 3.10983568e-01 8.16866636e-01 -1.02934337e+00 5.36607146e-01 1.51117814e+00 5.26079416e-01 9.72027540e-01 -1.40333033e+00 -1.00076342e+00 -1.04535773e-01 -1.39374778e-01 -1.58661175e+00 -7.62499273e-01 1.12774920e+00 -4.74115252e-01 8.74778032e-02 2.94554204e-01 3.52997959e-01 1.50669062e+00 3.88459004e-02 5.85235953e-01 9.62181449e-01 -1.26743674e-01 9.46842600e-03 1.54589310e-01 -8.90360475e-01 8.91259909e-01 1.59250736e-01 4.01010305e-01 -5.31693935e-01 -1.29476376e-02 1.00983858e+00 5.10235764e-02 -3.57009143e-01 -6.06713176e-01 -7.45101571e-01 8.95328104e-01 5.25141418e-01 -3.76223385e-01 -2.82682747e-01 2.26264983e-01 3.10545921e-01 1.75454155e-01 6.53832614e-01 1.41467214e-01 -9.21248496e-02 2.90600240e-01 -9.25998509e-01 3.27711552e-01 6.49807334e-01 9.22192514e-01 6.12746537e-01 3.70801836e-01 -3.80788654e-01 6.76778376e-01 6.87687457e-01 6.58147395e-01 2.01026648e-01 -7.08241045e-01 5.64870834e-01 4.16242331e-01 -1.04058407e-01 -1.15505552e+00 2.11200237e-01 -5.51097810e-01 -9.03776288e-01 4.57045168e-01 1.94311529e-01 -3.17355216e-01 -8.89255941e-01 2.20102763e+00 3.26290429e-01 4.35061276e-01 4.89603877e-02 9.14197683e-01 8.04556251e-01 3.98990214e-01 -8.12022910e-02 -1.46663785e-01 1.29830539e+00 -7.85776615e-01 -7.27151453e-01 -2.23743930e-01 -1.29863039e-01 -6.59448147e-01 5.42221963e-01 9.73114371e-02 -1.15453339e+00 -5.20500422e-01 -1.22519207e+00 -1.42446578e-01 1.76041033e-02 5.39263189e-01 1.73293903e-01 7.76719809e-01 -8.71771097e-01 3.77185404e-01 -7.73877144e-01 2.30091050e-01 8.45849693e-01 4.67531770e-01 -6.44156516e-01 -1.20020732e-01 -1.12636435e+00 6.11891270e-01 -6.50649611e-03 4.62281227e-01 -1.47922170e+00 -6.77607119e-01 -1.09208417e+00 -5.85112274e-02 3.03173423e-01 -8.70710790e-01 8.25411618e-01 -1.10229516e+00 -1.68622136e+00 1.15115047e+00 4.14981022e-02 -1.62493482e-01 8.29794407e-01 -5.27775921e-02 -3.24544281e-01 1.34332895e-01 -2.91284099e-02 6.66613996e-01 1.35021532e+00 -1.58614552e+00 2.86802314e-02 -8.53724182e-01 -8.08072016e-02 1.54162869e-01 -2.03326195e-01 -2.30309978e-01 -5.43082595e-01 -8.82202387e-01 -9.81146172e-02 -1.04069233e+00 2.41820857e-01 3.41958582e-01 -5.35373986e-01 3.30635190e-01 1.05827844e+00 -1.09877133e+00 7.83792078e-01 -2.26833725e+00 4.32536215e-01 -4.61608469e-02 3.67087722e-01 2.79193103e-01 -3.47321779e-01 2.43443474e-01 -3.41341168e-01 -2.02370569e-01 -2.13526025e-01 -9.58545446e-01 9.78932530e-02 1.76918209e-01 -2.66155601e-01 7.98040628e-01 5.25776803e-01 9.33352172e-01 -6.22379124e-01 -1.64251387e-01 2.10967641e-02 1.00379288e+00 -6.45215452e-01 5.64320505e-01 -1.48540840e-01 9.49465752e-01 -2.95269161e-01 5.60532928e-01 9.53666687e-01 2.62821496e-01 1.12522431e-01 -4.52608466e-01 3.54575783e-01 4.66591008e-02 -9.28061604e-01 1.78706622e+00 -5.54972112e-01 4.62195694e-01 1.94135621e-01 -8.08178186e-01 1.07235718e+00 6.20583773e-01 2.85414815e-01 -4.19272035e-01 2.96610951e-01 7.12248981e-02 -1.04966640e-01 -2.97271639e-01 -7.73195103e-02 -2.63049126e-01 8.72709900e-02 2.25508273e-01 2.82125205e-01 -7.15520307e-02 -3.76462519e-01 -8.29699412e-02 7.86286950e-01 4.22204942e-01 -1.83916111e-02 5.19546680e-02 7.33438075e-01 -7.56962001e-01 6.37884080e-01 -4.81572226e-02 -1.34991920e-02 8.75498950e-01 7.97203660e-01 -1.67621553e-01 -1.01295149e+00 -1.23612130e+00 8.29348117e-02 7.82379866e-01 -8.19691122e-02 2.95686945e-02 -9.90165114e-01 -8.11706662e-01 5.35926372e-02 4.30417061e-01 -9.47851360e-01 -3.77211481e-01 -4.62684333e-01 -4.46227252e-01 8.57588053e-01 4.93034542e-01 6.78498924e-01 -8.13812256e-01 -1.79932430e-01 -2.33573005e-01 9.69470069e-02 -1.15529919e+00 -7.83919573e-01 -2.59473354e-01 -4.01932180e-01 -1.01287425e+00 -7.37954795e-01 -8.10276747e-01 9.62544918e-01 -2.94706702e-01 8.49084139e-01 -3.49692196e-01 -4.40873086e-01 8.16030949e-02 -9.17074829e-02 -2.68252075e-01 -3.34854126e-01 -2.61062682e-01 -6.30674139e-02 6.79970503e-01 -2.49698712e-03 -8.18955898e-01 -9.68960106e-01 3.56180161e-01 -8.26985300e-01 1.08340625e-02 6.43679321e-01 1.08299541e+00 5.72349906e-01 -2.94811726e-01 6.14700019e-01 -9.82383549e-01 3.43036056e-01 -3.65389168e-01 -5.67484379e-01 1.66903749e-01 -3.52226228e-01 2.04284355e-01 5.84534287e-01 -3.40294540e-01 -1.12954247e+00 3.12862784e-01 -3.06104928e-01 -8.24574113e-01 8.48992765e-02 -3.40292491e-02 -1.06982684e+00 -3.15479666e-01 4.47627485e-01 3.70593101e-01 3.53678524e-01 -3.52204531e-01 4.88693565e-01 5.52619040e-01 6.38914108e-01 -4.79721308e-01 1.05845284e+00 4.54471409e-01 2.08484888e-01 -4.09754604e-01 -6.35003209e-01 2.48954102e-01 -5.95609784e-01 -9.62816775e-02 9.16327000e-01 -1.18487287e+00 -8.50203097e-01 4.86050963e-01 -1.26689708e+00 5.45603670e-02 -2.40609854e-01 4.75307554e-02 -7.01460361e-01 -6.88278079e-02 -4.28180724e-01 -7.86272466e-01 -3.96965235e-01 -1.43614995e+00 1.56810331e+00 6.96551427e-02 1.33182600e-01 -1.02261615e+00 1.82995498e-02 4.97161806e-01 2.56391376e-01 9.88375962e-01 6.02432191e-01 -5.89094520e-01 -5.64812958e-01 -4.86562908e-01 -1.56014755e-01 7.09024489e-01 2.36970589e-01 -2.46913910e-01 -1.34037030e+00 -5.41272044e-01 2.65589118e-01 -5.20794392e-01 6.32188976e-01 -1.49626508e-01 1.18931484e+00 -7.60849118e-01 3.10385991e-02 1.02034843e+00 1.38550389e+00 2.84892004e-02 8.48903596e-01 -4.10416216e-01 1.01336432e+00 6.81276321e-01 7.96589851e-02 3.76614332e-01 2.64109939e-01 8.83122981e-01 7.16318250e-01 -8.76226202e-02 -3.57485324e-01 -7.29096889e-01 5.59998572e-01 4.96221215e-01 -1.94161776e-02 -2.19520137e-01 -4.81005132e-01 4.67441417e-02 -1.44710875e+00 -9.03494716e-01 4.34590906e-01 2.08708119e+00 8.06710780e-01 -3.22712302e-01 -3.14288437e-02 1.02058582e-01 6.15861058e-01 3.20540696e-01 -6.72875345e-01 -6.59588650e-02 3.13753635e-02 4.17710692e-01 2.19831809e-01 4.05066490e-01 -9.71523345e-01 7.04128683e-01 5.04956675e+00 7.58687556e-01 -1.34669018e+00 2.59884119e-01 7.80997336e-01 -1.03768088e-01 -2.28322670e-01 -3.20304453e-01 -5.32743573e-01 4.04488951e-01 5.78825712e-01 9.39592794e-02 5.62446833e-01 8.14081907e-01 8.85623172e-02 4.94521260e-01 -1.48969209e+00 1.16102922e+00 5.09698391e-01 -9.26504016e-01 2.39407718e-01 2.12299347e-01 7.70153701e-01 -5.01449883e-01 6.80385888e-01 7.54339695e-02 -4.20729332e-02 -1.49143791e+00 7.99064338e-01 5.51688790e-01 1.44685400e+00 -9.06226873e-01 5.55119038e-01 1.69166118e-01 -1.07531512e+00 1.28538340e-01 -1.52138469e-03 3.40191603e-01 -3.29121314e-02 2.85396487e-01 -6.63939178e-01 7.28617966e-01 -3.95543408e-03 6.58590019e-01 -4.56279188e-01 3.42249453e-01 -4.92574066e-01 2.04749733e-01 1.51443809e-01 5.93737066e-01 -1.64835185e-01 -2.58606851e-01 5.62788308e-01 7.52671540e-01 1.50313124e-01 -1.57600850e-01 -7.32651651e-02 1.24889624e+00 -5.39205968e-01 -2.42219433e-01 -6.91431820e-01 -1.21380337e-01 3.93298298e-01 1.30346167e+00 -2.18754530e-01 1.18960805e-01 -1.84125230e-01 1.29777944e+00 1.68554515e-01 3.45314056e-01 -9.52765822e-01 -2.98325568e-01 9.44565654e-01 1.68476582e-01 3.24324042e-01 6.41809031e-02 -1.55976549e-01 -1.13878596e+00 1.15581609e-01 -1.07241797e+00 -9.39121991e-02 -5.36535263e-01 -1.30820894e+00 9.23488259e-01 -1.85679421e-01 -1.09483194e+00 -4.61370856e-01 -6.24993801e-01 -6.13019288e-01 1.18411922e+00 -1.26689482e+00 -1.70828128e+00 -3.60533714e-01 6.50577068e-01 3.79962027e-01 -4.63686854e-01 9.11806762e-01 5.35290122e-01 -7.87923396e-01 1.25993979e+00 -1.50535196e-01 5.39398670e-01 5.70832908e-01 -8.71733367e-01 3.27719986e-01 7.84368455e-01 6.09011650e-02 5.92586696e-01 5.88281810e-01 -4.60966438e-01 -1.65217757e+00 -1.50861251e+00 3.20436746e-01 -3.75386596e-01 2.62146741e-01 -8.44402850e-01 -4.70823258e-01 7.25251317e-01 1.19630760e-02 4.96698827e-01 6.06355965e-01 -5.06049693e-01 -6.98024929e-01 -3.42421651e-01 -1.57043564e+00 4.36171234e-01 1.04418504e+00 -8.60360444e-01 -8.22759047e-02 1.47117585e-01 6.25811279e-01 -5.01112282e-01 -7.45158195e-01 3.86300772e-01 8.40987861e-01 -9.33740437e-01 1.03538489e+00 -3.15969020e-01 6.82054698e-01 -1.50376812e-01 -1.79025874e-01 -1.31995261e+00 1.02863483e-01 -8.20285261e-01 -2.88398981e-01 1.42984247e+00 2.24036366e-01 -6.73094869e-01 9.11223412e-01 2.57424355e-01 1.41601056e-01 -1.00912404e+00 -9.96375680e-01 -6.23370409e-01 -1.39488861e-01 4.55494523e-02 8.14923644e-01 5.91384172e-01 -6.79266930e-01 4.08335477e-01 -6.52737081e-01 3.51609707e-01 7.71403730e-01 -2.58396327e-01 7.43025780e-01 -8.53638887e-01 -5.07254362e-01 -7.93968365e-02 -6.75643444e-01 -8.94863904e-01 4.46289659e-01 -8.32529664e-01 3.49833426e-04 -8.67400587e-01 1.93307385e-01 -1.48882791e-01 1.20642833e-01 3.50937605e-01 -3.28445947e-03 6.41698539e-01 1.58475265e-01 -2.65998468e-02 -7.81907588e-02 1.03931320e+00 1.56328750e+00 -1.81244761e-01 2.60166883e-01 9.81646925e-02 -7.40245402e-01 5.08208394e-01 3.58473331e-01 -3.75956565e-01 -6.40369236e-01 -4.89028931e-01 -4.42326032e-02 3.48377347e-01 6.92203343e-01 -8.16095650e-01 -2.86636930e-02 2.40726769e-01 6.36391878e-01 -5.69174550e-02 7.17992842e-01 -8.75760257e-01 4.02613461e-01 3.94309402e-01 -3.36100012e-01 -5.95656969e-02 8.31666514e-02 6.38963342e-01 -2.80745715e-01 2.01275080e-01 9.98682797e-01 1.76957265e-01 4.12602462e-02 8.04077983e-01 5.25808275e-01 8.84787813e-02 1.14646089e+00 -9.53504294e-02 -4.59200591e-02 -3.84577841e-01 -7.53624201e-01 -2.00059578e-01 3.80451232e-01 4.19295073e-01 7.13641167e-01 -1.68728077e+00 -9.58264291e-01 6.91792846e-01 1.61032304e-01 5.40307499e-02 2.00823694e-01 5.97652137e-01 -5.85383594e-01 1.44492432e-01 -3.29678804e-01 -3.07871670e-01 -1.19916081e+00 4.20321494e-01 5.45474112e-01 -1.77932322e-01 -2.00414479e-01 1.03417504e+00 4.99232829e-01 -4.19714004e-01 2.92032987e-01 3.40924263e-01 -3.21294405e-02 -5.14603127e-03 3.66292268e-01 2.63923466e-01 -6.36748155e-04 -9.49900031e-01 -2.97892809e-01 4.41866606e-01 1.11263067e-01 -2.57474601e-01 1.44386363e+00 3.99384052e-02 -2.01630935e-01 8.85580331e-02 1.64589679e+00 6.21833950e-02 -1.69503760e+00 -4.76729237e-02 -6.77470207e-01 -5.79859674e-01 -8.05972666e-02 -7.13453531e-01 -1.53977847e+00 1.01940119e+00 8.16864550e-01 -3.36436778e-01 1.19446671e+00 -2.27801338e-01 7.05621362e-01 -3.81609827e-01 2.71082461e-01 -5.06532907e-01 3.17740619e-01 1.07453428e-01 1.32627499e+00 -1.13276732e+00 -2.15692207e-01 -5.12463510e-01 -6.80540919e-01 9.41963077e-01 7.01013446e-01 -3.15660655e-01 6.29628122e-01 3.31560820e-01 -1.92169547e-01 -1.12148479e-01 -4.91097987e-01 2.23605111e-01 6.12591505e-01 5.92248321e-01 3.87868077e-01 8.55737627e-02 1.29887283e-01 6.53935194e-01 -2.94253349e-01 -1.23483397e-01 -8.01057816e-02 4.74519461e-01 3.09087962e-01 -1.20736837e+00 -3.13690335e-01 5.82721131e-03 -3.09059024e-01 1.11804955e-01 -4.57712501e-01 5.97388029e-01 5.10549426e-01 6.02769494e-01 7.73191527e-02 -5.31518042e-01 2.32575059e-01 2.17798613e-02 7.41156459e-01 -6.56231344e-01 -4.33250010e-01 -7.61329383e-02 -3.15177917e-01 -5.88043511e-01 -6.74486682e-02 -5.40370405e-01 -7.19984353e-01 -2.56861180e-01 -1.03364810e-01 -1.49352506e-01 6.29211545e-01 7.51706660e-01 4.04229969e-01 7.01119840e-01 1.12609041e+00 -1.02215791e+00 -9.67194319e-01 -9.63930130e-01 -5.40078640e-01 5.94275355e-01 5.34168661e-01 -7.62449205e-01 -3.28058630e-01 1.18827701e-01]
[12.900186538696289, 0.16244354844093323]
08b3b2c3-113d-4ada-97a4-6a4aba11c1d7
rico-regularizing-the-unobservable-for-indoor
2303.08605
null
https://arxiv.org/abs/2303.08605v1
https://arxiv.org/pdf/2303.08605v1.pdf
RICO: Regularizing the Unobservable for Indoor Compositional Reconstruction
Recently, neural implicit surfaces have become popular for multi-view reconstruction. To facilitate practical applications like scene editing and manipulation, some works extend the framework with semantic masks input for the object-compositional reconstruction rather than the holistic perspective. Though achieving plausible disentanglement, the performance drops significantly when processing the indoor scenes where objects are usually partially observed. We propose RICO to address this by regularizing the unobservable regions for indoor compositional reconstruction. Our key idea is to first regularize the smoothness of the occluded background, which then in turn guides the foreground object reconstruction in unobservable regions based on the object-background relationship. Particularly, we regularize the geometry smoothness of occluded background patches. With the improved background surface, the signed distance function and the reversedly rendered depth of objects can be optimized to bound them within the background range. Extensive experiments show our method outperforms other methods on synthetic and real-world indoor scenes and prove the effectiveness of proposed regularizations.
['Yong liu', 'Yiyi Liao', 'Mengmeng Wang', 'Yuanyuan Ding', 'Xiaoyang Lyu', 'Zizhang Li']
2023-03-15
null
null
null
null
['object-reconstruction']
['computer-vision']
[ 4.94058847e-01 1.44791557e-02 1.55812621e-01 -3.04408878e-01 -2.06010580e-01 -4.21901047e-01 3.53191793e-01 -5.18351495e-01 8.07344168e-02 7.08735049e-01 3.94311339e-01 7.19177201e-02 -3.07805873e-02 -1.01330471e+00 -7.69693375e-01 -8.58010471e-01 4.69446033e-01 1.78377941e-01 3.83858591e-01 -6.67686239e-02 2.11443618e-01 5.91999650e-01 -1.55737817e+00 3.93374562e-01 1.00754285e+00 7.36723304e-01 4.53248858e-01 4.64572459e-01 -1.66198000e-01 7.44184732e-01 -1.57265842e-01 -8.13563243e-02 5.53543866e-01 -1.26684338e-01 -2.93401778e-01 5.88989437e-01 7.01082349e-01 -5.80277264e-01 -3.66383940e-01 1.19761074e+00 1.88824072e-01 2.95286745e-01 3.93305540e-01 -8.35811973e-01 -6.78550124e-01 4.25554141e-02 -8.37750018e-01 -7.94394985e-02 3.23574752e-01 -1.52134476e-02 6.72023177e-01 -1.04799271e+00 5.87261081e-01 1.43063569e+00 3.36637288e-01 5.31988084e-01 -1.16506958e+00 -6.27663970e-01 1.01018178e+00 -6.28473014e-02 -1.39013827e+00 -5.36359251e-01 1.16281617e+00 -3.55409175e-01 3.48760277e-01 6.28439367e-01 5.82774818e-01 6.82905555e-01 -2.14313548e-02 7.22078621e-01 1.03767371e+00 -1.49421170e-01 -4.02111514e-03 2.77836680e-01 4.75412421e-02 7.00803757e-01 3.00391525e-01 -7.66094849e-02 -3.28492343e-01 -1.75764546e-01 1.39308548e+00 5.04330933e-01 -6.76497936e-01 -6.62697494e-01 -1.28628838e+00 3.85836482e-01 4.36182857e-01 -5.51090688e-02 -2.78014839e-01 1.31453767e-01 2.13623466e-03 -1.41770750e-01 8.45659316e-01 7.69102648e-02 -4.40086633e-01 5.46875477e-01 -5.73559940e-01 2.61933595e-01 3.23875725e-01 1.24086869e+00 8.47687244e-01 1.21453680e-01 -4.08064760e-02 8.79973710e-01 4.96915847e-01 4.93684977e-01 -3.35944116e-01 -1.28525817e+00 6.42674685e-01 6.83689237e-01 5.39562225e-01 -1.12932217e+00 8.68783817e-02 -3.59022528e-01 -1.00961745e+00 3.01189601e-01 3.25712085e-01 4.83099520e-02 -9.48058009e-01 1.44955778e+00 7.52429664e-01 4.07767534e-01 -1.99388713e-01 1.11810637e+00 7.71502197e-01 8.17119598e-01 -3.79895151e-01 -2.67003834e-01 1.10052192e+00 -1.13612092e+00 -7.61825800e-01 -2.94658303e-01 -3.96381021e-02 -7.95118272e-01 1.14729047e+00 3.68071795e-01 -1.16546273e+00 -5.25677323e-01 -8.93429339e-01 -2.83203810e-01 1.71494544e-01 6.36780336e-02 7.35271513e-01 3.54673177e-01 -7.03317165e-01 2.45766982e-01 -8.79565120e-01 3.75020914e-02 4.35520351e-01 2.36130893e-01 -2.21304834e-01 -3.74306083e-01 -5.44780493e-01 2.59906918e-01 1.07710235e-01 4.37171519e-01 -9.91784811e-01 -8.13749492e-01 -8.25460434e-01 3.77691798e-02 6.41022146e-01 -9.84913468e-01 5.64080596e-01 -1.18710434e+00 -1.62338173e+00 8.14599633e-01 -2.77590096e-01 2.37431571e-01 5.99003613e-01 -3.88761669e-01 8.25540423e-02 -8.61679465e-02 1.33902356e-01 2.27563888e-01 7.58115172e-01 -1.82690787e+00 -5.48717618e-01 -4.69522595e-01 4.81914043e-01 6.09112799e-01 5.76053225e-02 -2.94807076e-01 -6.39804065e-01 -6.54615223e-01 5.98065495e-01 -6.40496016e-01 -5.29905677e-01 5.49747527e-01 -3.55729938e-01 2.94629067e-01 9.81887817e-01 -6.67334914e-01 7.83578873e-01 -2.36697340e+00 5.03840923e-01 -4.02369164e-02 2.25926563e-01 -2.89019763e-01 -3.22273634e-02 -6.76345900e-02 1.10167891e-01 -1.45099476e-01 -3.17970157e-01 -4.20738548e-01 -3.93277913e-01 3.18099439e-01 -3.28076750e-01 7.76508510e-01 -2.44443059e-01 4.34702635e-01 -8.82149637e-01 -2.40693077e-01 4.57918137e-01 6.04888141e-01 -9.17257011e-01 5.01972795e-01 -3.31762612e-01 1.03919411e+00 -8.27807307e-01 5.62855899e-01 1.21651840e+00 -1.66713580e-01 -5.77900223e-02 -2.08064854e-01 -2.54404426e-01 1.65542185e-01 -1.58527768e+00 1.80668914e+00 -6.02898836e-01 3.17093432e-01 8.08032215e-01 -7.60462523e-01 6.93279743e-01 3.19485404e-02 4.40971226e-01 -3.38209718e-01 -1.53493732e-01 -1.20813631e-01 -4.12614256e-01 -3.30779254e-01 3.12078297e-01 -1.30771846e-01 4.93175566e-01 -1.13254540e-01 -6.29327416e-01 -2.52164364e-01 -4.18183386e-01 8.40367079e-02 5.95096350e-01 4.75197107e-01 1.80623874e-01 -3.77745926e-01 8.22278798e-01 -3.63642931e-01 8.33413005e-01 2.47536600e-01 1.75640553e-01 9.56679881e-01 1.21052794e-01 -4.95003402e-01 -9.00896668e-01 -1.34403074e+00 -1.08172037e-01 9.06541944e-01 7.78503656e-01 -2.34802775e-02 -5.85918784e-01 -5.06558836e-01 -2.02877864e-01 6.15038693e-01 -4.76222515e-01 2.55913377e-01 -8.96276772e-01 -5.64231813e-01 -2.68592596e-01 3.30645323e-01 6.11253619e-01 -7.42352545e-01 -3.75807732e-01 3.62412855e-02 -4.15817797e-01 -1.14540267e+00 -5.99445343e-01 -1.74185634e-01 -9.42941964e-01 -1.03393912e+00 -6.86146438e-01 -6.26802802e-01 1.11798823e+00 7.98426926e-01 9.57648039e-01 3.16740066e-01 -1.45694822e-01 3.69359136e-01 -1.74851008e-02 -6.88098818e-02 5.52971959e-02 -5.26113987e-01 -9.29617882e-02 3.80736768e-01 -4.85119343e-01 -9.43750441e-01 -7.85179436e-01 5.64114213e-01 -8.83462787e-01 7.25127876e-01 7.51021579e-02 6.69734716e-01 9.69252825e-01 -8.63797963e-02 -4.21731882e-02 -9.36961651e-01 -2.60576308e-01 -3.47390175e-01 -9.30844963e-01 2.12741986e-01 -1.45822689e-01 -2.28192613e-01 5.84542632e-01 -4.62837636e-01 -1.59489405e+00 8.79229140e-03 1.27317131e-01 -6.15619004e-01 -1.43744156e-01 -2.96121180e-01 -8.45668674e-01 -2.85572372e-02 1.48688748e-01 2.53480434e-01 -5.99032402e-01 -4.43662286e-01 4.85479653e-01 1.69530526e-01 2.62356997e-01 -7.08430111e-01 8.00200939e-01 1.15362573e+00 6.61781523e-03 -9.60312068e-01 -9.17873442e-01 -3.83735865e-01 -5.65569401e-01 -2.15565845e-01 9.87668812e-01 -1.09594560e+00 -5.44308364e-01 2.74589449e-01 -1.38816071e+00 -3.52949381e-01 -2.13356525e-01 2.85949737e-01 -5.09961247e-01 5.36999583e-01 -5.17607927e-01 -9.77583587e-01 1.54236248e-02 -1.18793988e+00 1.33295023e+00 1.77481219e-01 3.49640310e-01 -9.19787705e-01 -2.95109779e-01 4.73518401e-01 1.42701432e-01 3.83690476e-01 8.73813987e-01 2.83032417e-01 -1.43472278e+00 4.01502281e-01 -4.59696203e-01 1.67168468e-01 5.09187579e-01 -2.99807899e-02 -1.07327247e+00 -1.22179978e-01 4.29556817e-01 1.84955969e-01 7.50750244e-01 6.38792694e-01 1.34957385e+00 -4.13227737e-01 -4.78271008e-01 1.19065475e+00 1.46873450e+00 1.46670938e-01 6.76675200e-01 1.63226902e-01 1.21597219e+00 7.60054708e-01 6.67618215e-01 3.72530490e-01 3.86641502e-01 7.04238176e-01 7.19373643e-01 -1.86954111e-01 -1.47462904e-01 -1.93239197e-01 1.53753042e-01 8.98145020e-01 -3.50190997e-01 -4.03456867e-01 -4.12363142e-01 3.20044935e-01 -1.83223569e+00 -8.31008136e-01 -3.80559087e-01 2.12825370e+00 4.26556617e-01 -6.38618544e-02 -4.26296085e-01 -3.20920259e-01 5.97958624e-01 5.13972104e-01 -5.00472069e-01 1.10955618e-01 -1.37916178e-01 -2.86557406e-01 4.89252567e-01 9.19519722e-01 -8.49274635e-01 8.11973214e-01 5.45323896e+00 6.65745378e-01 -8.98438573e-01 2.11275965e-01 7.48268962e-01 -1.53169706e-01 -8.37416291e-01 8.77322853e-02 -8.02959919e-01 2.51648337e-01 1.18211126e-02 8.76797885e-02 7.57842958e-01 8.68546784e-01 4.38473433e-01 -2.19189510e-01 -1.05811930e+00 1.05018687e+00 -4.01462466e-02 -1.36269498e+00 1.53348878e-01 7.48196468e-02 1.00963640e+00 -2.97572136e-01 1.56328809e-02 -7.93815330e-02 7.01223388e-02 -7.59204388e-01 8.47138762e-01 6.85804307e-01 6.32861316e-01 -5.20089447e-01 2.27829099e-01 6.65937245e-01 -1.40124261e+00 1.92246467e-01 -6.18711591e-01 -1.62006125e-01 3.53419453e-01 6.49391472e-01 -1.29382223e-01 6.28550053e-01 4.90628570e-01 7.88434148e-01 -1.78541541e-01 6.47756815e-01 -1.47718787e-01 1.35039017e-01 -2.27569506e-01 5.06199479e-01 -4.20255847e-02 -8.43303502e-01 7.59405077e-01 7.82102585e-01 9.08112079e-02 5.92543125e-01 4.01618451e-01 1.28979576e+00 1.17792733e-01 -4.45817073e-04 -6.76125109e-01 4.12647814e-01 4.50608097e-02 1.13573313e+00 -8.57656598e-01 -3.25438619e-01 -4.81318593e-01 1.34961581e+00 3.40925872e-01 8.52142990e-01 -1.03582132e+00 1.50937393e-01 8.46211731e-01 5.26007056e-01 1.85088098e-01 -3.97047937e-01 -4.70641255e-01 -1.55378997e+00 2.72102088e-01 -5.94170570e-01 -1.35095134e-01 -7.96471834e-01 -1.08924341e+00 3.07590574e-01 1.36171371e-01 -1.08409512e+00 4.86805439e-01 -4.85693663e-01 -5.15237033e-01 8.76315236e-01 -1.43187368e+00 -1.15982437e+00 -5.74750245e-01 6.63171530e-01 1.02212429e+00 3.78264666e-01 5.12306631e-01 4.78432924e-01 -4.74572271e-01 -8.94738510e-02 4.32100557e-02 -2.11675569e-01 4.35354769e-01 -8.96099448e-01 1.64813504e-01 9.54187810e-01 -8.79237428e-02 7.87444830e-01 5.54983616e-01 -8.13043654e-01 -1.39902973e+00 -1.12428892e+00 -1.74508858e-02 -4.65812892e-01 1.45230979e-01 -5.81980288e-01 -9.64504480e-01 8.17350030e-01 5.44704050e-02 3.38673323e-01 1.95940316e-01 -2.10229725e-01 -2.39479497e-01 -2.11744219e-01 -1.08603728e+00 7.44694650e-01 1.49449420e+00 -3.82636428e-01 -1.85927242e-01 4.16420817e-01 1.01983511e+00 -8.27211201e-01 -3.85275036e-01 5.63317180e-01 4.76446539e-01 -1.30757463e+00 1.29546630e+00 -3.48419189e-01 4.10293996e-01 -6.59314930e-01 -5.32213867e-01 -9.59264278e-01 -3.48852187e-01 -3.67028028e-01 -1.05791807e-01 1.13854873e+00 -9.51742157e-02 -5.50875366e-01 9.04835343e-01 5.94682813e-01 -3.29409331e-01 -7.45989442e-01 -7.83001006e-01 -3.55640769e-01 -3.76308233e-01 -3.34802866e-01 3.73243600e-01 1.00034964e+00 -9.00890112e-01 2.03685060e-01 -4.93971437e-01 6.87793612e-01 8.11186492e-01 4.03535813e-01 1.03391612e+00 -9.65279698e-01 -4.52051044e-01 -3.17809405e-03 -6.12269267e-02 -1.64748943e+00 5.93421608e-02 -5.67199528e-01 9.09820944e-02 -1.59160340e+00 3.23205382e-01 -6.27309561e-01 5.72933406e-02 -2.17591748e-01 -3.75507385e-01 -5.07800467e-02 3.63613702e-02 5.35299927e-02 -4.63391781e-01 8.46005797e-01 1.75722182e+00 -4.54912707e-02 -3.50000829e-01 1.29200250e-01 -4.72032040e-01 1.25498152e+00 2.60805160e-01 -1.99280277e-01 -6.90708876e-01 -8.75751197e-01 9.10925567e-02 1.41601369e-01 4.74144846e-01 -6.85922265e-01 -1.96360588e-01 -7.80883491e-01 5.06470382e-01 -7.39021242e-01 7.43432283e-01 -1.20085013e+00 5.00782788e-01 1.04415148e-01 3.58704850e-02 -2.57319450e-01 5.30756973e-02 9.26215053e-01 1.14936411e-01 1.45729586e-01 9.39426661e-01 -3.05617005e-01 -3.77640545e-01 5.49871564e-01 5.04982136e-02 -1.19142480e-01 9.97006059e-01 -4.05419230e-01 -1.69989467e-02 -1.92678630e-01 -5.84906876e-01 8.55023116e-02 8.02061558e-01 2.09759757e-01 8.53966355e-01 -1.29320681e+00 -5.12264073e-01 4.91612583e-01 -1.58961192e-01 6.99800909e-01 5.68284035e-01 7.43410766e-01 -7.46587992e-01 1.06732018e-01 -1.95039269e-02 -7.49652326e-01 -1.38906121e+00 6.41740501e-01 5.51363885e-01 -2.91426219e-02 -9.44815278e-01 9.75987077e-01 1.36261928e+00 -4.91921246e-01 3.14715564e-01 -6.42064869e-01 1.43578216e-01 -4.92853969e-01 5.62735796e-01 5.66527784e-01 -5.24751961e-01 -5.66663027e-01 -2.41902694e-01 8.44432354e-01 1.01664685e-01 3.03940568e-02 1.36553693e+00 -5.61723590e-01 -4.12348479e-01 5.47146797e-01 8.28689516e-01 6.87551498e-01 -1.59175169e+00 -1.42996743e-01 -5.84611714e-01 -1.04454708e+00 -5.16678989e-02 -1.42927811e-01 -1.14769971e+00 8.75795186e-01 5.06832242e-01 -4.59622890e-02 1.13534248e+00 -1.50888383e-01 6.09144747e-01 1.53553307e-01 6.35465860e-01 -5.92482567e-01 5.48445247e-02 3.89850557e-01 1.05555940e+00 -9.84764397e-01 4.53068584e-01 -1.15360749e+00 -3.57310802e-01 9.15513396e-01 9.56828475e-01 -3.67089957e-01 6.39791608e-01 2.32454315e-01 -1.48106635e-01 -2.48310149e-01 -3.36216152e-01 1.83798298e-01 3.48645806e-01 4.58763093e-01 3.78751576e-01 9.54799950e-02 3.86372209e-02 3.96070600e-01 3.36192906e-01 -4.18708831e-01 3.44186902e-01 4.89588976e-01 -4.52987432e-01 -8.01032841e-01 -7.15017021e-01 1.11001253e-01 -2.88378179e-01 -9.00496617e-02 -1.01242915e-01 5.86302221e-01 4.47177202e-01 7.07747996e-01 -9.87267494e-02 2.24577174e-01 4.46800351e-01 -4.35207546e-01 6.42521501e-01 -9.25842285e-01 -8.67793709e-02 4.49275076e-01 -8.02255869e-02 -7.98850358e-01 -5.94843924e-01 -5.25792599e-01 -1.23608029e+00 -1.36916324e-01 -5.05135119e-01 -2.72666901e-01 2.90371209e-01 8.30558658e-01 7.63161480e-02 6.74030900e-01 6.27354145e-01 -1.15426826e+00 8.14227536e-02 -5.31065702e-01 -6.08617604e-01 2.31772393e-01 5.32802224e-01 -7.55574405e-01 -3.79880220e-01 1.26102716e-01]
[9.098954200744629, -3.021789073944092]
f45665ae-cf62-4730-b975-0ba01b1d1b6e
improving-the-numerical-reasoning-skills-of
2205.06733
null
https://arxiv.org/abs/2205.06733v2
https://arxiv.org/pdf/2205.06733v2.pdf
Arithmetic-Based Pretraining -- Improving Numeracy of Pretrained Language Models
State-of-the-art pretrained language models tend to perform below their capabilities when applied out-of-the-box on tasks that require understanding and working with numbers. Recent work suggests two main reasons for this: (1) popular tokenisation algorithms have limited expressiveness for numbers, and (2) common pretraining objectives do not target numeracy. Approaches that address these shortcomings usually require architectural changes or pretraining from scratch. In this paper, we propose a new extended pretraining approach called Arithmetic-Based Pretraining that jointly addresses both in one extended pretraining step without requiring architectural changes or pretraining from scratch. Arithmetic-Based Pretraining combines contrastive learning to improve the number representation, and a novel extended pretraining objective called Inferable Number Prediction Task to improve numeracy. Our experiments show the effectiveness of Arithmetic-Based Pretraining in three different tasks that require improved numeracy, i.e., reading comprehension in the DROP dataset, inference-on-tables in the InfoTabs dataset, and table-to-text generation in the WikiBio and SciGen datasets.
['Iryna Gurevych', 'Nafise Sadat Moosavi', 'Dominic Petrak']
2022-05-13
null
null
null
null
['table-to-text-generation']
['natural-language-processing']
[ 3.42618555e-01 9.70296934e-02 -2.63211787e-01 -3.91279280e-01 -7.55164564e-01 -5.37668467e-01 7.53658295e-01 7.32639611e-01 -8.27264607e-01 1.05220079e+00 2.53975540e-01 -7.50143170e-01 3.26248407e-02 -1.17535841e+00 -1.11126959e+00 1.54172936e-02 2.80355185e-01 7.08899260e-01 -6.19919859e-02 -4.37233359e-01 5.34654438e-01 3.10004652e-01 -1.66133916e+00 6.83765411e-01 1.26749730e+00 8.40424657e-01 2.04006553e-01 8.03723395e-01 -8.14863145e-01 1.26094115e+00 -8.08136284e-01 -6.03304863e-01 1.60857469e-01 -2.59323299e-01 -9.44952309e-01 -6.91895783e-01 5.48222244e-01 -5.08510709e-01 -5.42230085e-02 8.08265746e-01 4.35582578e-01 7.84856826e-02 7.28483617e-01 -1.30270004e+00 -8.43611717e-01 1.46132147e+00 -1.27936646e-01 2.86030453e-02 4.81493384e-01 1.01592779e-01 1.13471413e+00 -6.82339013e-01 2.45350718e-01 1.46668100e+00 8.66306841e-01 5.91259718e-01 -1.22828591e+00 -6.36583269e-01 1.07293695e-01 2.33787850e-01 -1.34862483e+00 -3.58087182e-01 3.52336794e-01 -2.42515370e-01 1.55904031e+00 2.85110116e-01 4.87826526e-01 6.58774316e-01 1.57671813e-02 8.65802109e-01 1.11820471e+00 -9.05315161e-01 -5.82766682e-02 1.14087284e-01 9.18446034e-02 6.23340547e-01 6.95491612e-01 -2.24961624e-01 -5.03152192e-01 2.64828026e-01 9.25826669e-01 -4.62567031e-01 1.99128345e-01 1.14164621e-01 -1.60495198e+00 7.42980063e-01 1.29116669e-01 3.00229400e-01 -1.62134305e-01 2.55761385e-01 6.86104834e-01 4.07598615e-01 1.97031111e-01 1.03670192e+00 -8.86045158e-01 -4.05284315e-01 -1.01591074e+00 5.03546357e-01 1.02771533e+00 1.11547112e+00 7.56298959e-01 1.29568890e-01 -3.16674054e-01 5.93603134e-01 -6.51260391e-02 3.95589948e-01 6.11171305e-01 -6.33452296e-01 1.00133801e+00 6.85020149e-01 -5.46425618e-02 -5.74618697e-01 -4.78064030e-01 -2.11976334e-01 -8.10242534e-01 5.33478670e-02 9.16902423e-01 -2.06526354e-01 -1.10601914e+00 1.91520083e+00 -6.58381656e-02 -1.79918915e-01 2.45259732e-01 3.42082828e-01 1.28370631e+00 9.47741032e-01 2.38553450e-01 9.19628590e-02 1.42504358e+00 -9.20737147e-01 -8.38966668e-01 -2.57936686e-01 1.44499564e+00 -6.58330381e-01 1.57742774e+00 5.98323941e-01 -1.52916873e+00 -7.77967691e-01 -1.25743508e+00 -6.91570878e-01 -1.10865331e+00 2.73738474e-01 1.01536870e+00 7.36827135e-01 -9.90233481e-01 6.65366709e-01 -5.43496847e-01 1.49168754e-02 2.60085553e-01 5.12018919e-01 -1.53322741e-01 -1.69632196e-01 -1.61478615e+00 1.40334523e+00 1.07442105e+00 -1.74480438e-01 -2.12406456e-01 -1.19602275e+00 -1.22561908e+00 4.36108232e-01 5.00548005e-01 -8.52068961e-01 1.33710074e+00 -7.13306129e-01 -1.58354342e+00 7.70799577e-01 1.21013418e-01 -6.80417717e-01 4.64966595e-01 -4.85334486e-01 -8.29779878e-02 -3.40666890e-01 6.29671216e-02 9.62220192e-01 3.26571882e-01 -8.45991850e-01 -3.99525583e-01 3.62475924e-02 2.13565156e-01 4.30039763e-01 -1.38893694e-01 -1.20373949e-01 -1.15697116e-01 -8.61921191e-01 -2.51945525e-01 -2.33650789e-01 -3.09869628e-02 -3.27299833e-01 -4.28737074e-01 -7.39997029e-01 5.89302145e-02 -5.83165407e-01 1.51048684e+00 -1.48185599e+00 4.78635868e-03 3.64634395e-02 4.51673679e-02 4.53035235e-01 -3.35714579e-01 3.22044611e-01 -1.93113491e-01 4.35013115e-01 6.96694572e-03 -2.91596472e-01 2.72649109e-01 1.72432259e-01 -8.37294608e-02 -2.46456578e-01 5.03957748e-01 1.05533588e+00 -1.06333160e+00 -7.62404799e-01 3.42526704e-01 7.54573867e-02 -8.61293912e-01 1.28233761e-01 -6.54309034e-01 -5.01960181e-02 2.54600018e-01 4.01014805e-01 6.57464981e-01 -1.60746723e-01 1.32796854e-01 -3.98722058e-03 -2.57517427e-01 1.01978171e+00 -1.43420100e+00 1.88450480e+00 -8.01900983e-01 3.26774746e-01 -5.94567537e-01 -1.01233113e+00 6.47593200e-01 2.56600082e-02 -9.62296054e-02 -8.40609133e-01 1.19943455e-01 4.96162206e-01 3.65855277e-01 -3.10302883e-01 8.91361117e-01 -1.56164646e-01 -1.50143847e-01 5.12573421e-01 2.58868635e-01 -5.52314341e-01 8.59536648e-01 1.92973703e-01 7.69292057e-01 3.54741037e-01 5.97385168e-01 -1.38769269e-01 7.25925684e-01 -5.66711137e-03 1.98828027e-01 9.57722187e-01 4.14214641e-01 2.67665893e-01 5.38662612e-01 -4.37772125e-01 -1.09670365e+00 -8.46393764e-01 1.74516416e-03 1.46059871e+00 -5.45641482e-01 -8.32096159e-01 -6.82487726e-01 -5.87886095e-01 -5.17756976e-02 1.27106082e+00 -5.37813246e-01 -3.82799134e-02 -8.70297551e-01 -5.47348082e-01 8.90303552e-01 9.84919667e-01 5.89614093e-01 -1.24514472e+00 -6.49323702e-01 3.34209025e-01 -3.24832588e-01 -1.21837103e+00 -8.50128457e-02 3.86505187e-01 -9.24956739e-01 -7.65751421e-01 -7.20992982e-01 -8.65797460e-01 7.14571536e-01 -3.61629367e-01 1.52819514e+00 2.64328688e-01 7.84638152e-02 -8.43976997e-03 -1.97750047e-01 -8.73645544e-01 -5.48966825e-01 4.83839035e-01 -1.66896701e-01 -7.78730989e-01 3.71592224e-01 -2.69326627e-01 2.26304442e-01 -3.48571241e-01 -8.61040175e-01 5.12030244e-01 9.71898973e-01 9.10303652e-01 3.59663695e-01 1.62356019e-01 7.14121461e-01 -1.07813859e+00 6.46442652e-01 -1.47610292e-01 -6.01933837e-01 5.48877239e-01 -4.93535340e-01 5.20982444e-01 1.06066775e+00 -4.28383946e-01 -8.75119269e-01 -1.61874875e-01 -2.20220745e-01 9.29391831e-02 -1.45505428e-01 9.33268309e-01 -3.76037002e-01 3.38927954e-01 6.77741885e-01 2.83096999e-01 -1.03373729e-01 -2.14099154e-01 5.96576452e-01 2.14371726e-01 5.55769920e-01 -9.81176972e-01 7.54004896e-01 -1.81403443e-01 1.77471191e-02 -6.57653749e-01 -1.06457651e+00 -1.35972932e-01 -7.36032367e-01 2.33749390e-01 6.05433166e-01 -9.06539083e-01 -8.80532801e-01 7.07001030e-01 -1.37846291e+00 -8.42270017e-01 -5.33995986e-01 5.29164732e-01 -4.95125562e-01 1.55953303e-01 -5.95575809e-01 -7.38345623e-01 -5.14574587e-01 -9.02368784e-01 6.48319006e-01 1.51302591e-01 -5.79050660e-01 -1.16953897e+00 -3.13668132e-01 8.62702280e-02 6.51697159e-01 -2.53026448e-02 1.62500846e+00 -8.39035809e-01 -4.26559865e-01 -6.64134100e-02 -5.48115015e-01 4.42983776e-01 -1.60546247e-02 -1.24295093e-01 -5.66274345e-01 2.18105271e-01 -4.40410256e-01 -8.46887887e-01 8.32431793e-01 2.31818140e-01 1.50085723e+00 -5.93211174e-01 1.37998417e-01 5.25726199e-01 1.22238195e+00 -1.70821413e-01 9.05615747e-01 5.41682065e-01 7.99723923e-01 4.50041354e-01 3.91776383e-01 5.14218092e-01 8.52661788e-01 3.44861984e-01 1.13550298e-01 3.18491235e-02 -1.26721501e-01 -4.96796399e-01 2.03206763e-01 9.04507577e-01 -1.01600662e-01 -2.18581468e-01 -1.02296305e+00 6.08956277e-01 -1.46983659e+00 -6.61454499e-01 -5.24594858e-02 2.26036644e+00 1.59070027e+00 4.81404990e-01 1.15633704e-01 6.12588406e-01 3.47748280e-01 -3.35514992e-02 -1.91473097e-01 -7.98766673e-01 -2.80501284e-02 7.36243308e-01 5.11076391e-01 4.25110340e-01 -8.61603737e-01 1.28233278e+00 5.47360086e+00 9.38466191e-01 -9.53453600e-01 -3.40044558e-01 6.28229916e-01 2.96789080e-01 -3.01106691e-01 -1.71056911e-01 -1.12125528e+00 1.23409614e-01 1.16153944e+00 -1.50362462e-01 4.59403455e-01 6.83138371e-01 -3.11434209e-01 -1.92333207e-01 -1.56891775e+00 9.99898732e-01 1.34773150e-01 -1.45273876e+00 6.37426555e-01 -4.75896448e-01 4.60417300e-01 -6.13465548e-01 1.06467165e-01 8.99288177e-01 2.61365384e-01 -1.57770598e+00 8.77331436e-01 4.92109507e-01 1.03032255e+00 -8.71882021e-01 8.03370953e-01 3.94523144e-01 -1.06221485e+00 2.72735924e-01 -3.34183633e-01 -7.15544999e-01 -5.19206971e-02 4.54385161e-01 -9.83343422e-01 4.20508087e-01 2.10519299e-01 4.35072958e-01 -9.25279200e-01 9.95343685e-01 -5.64349115e-01 3.43511611e-01 -3.88656408e-01 -4.80912417e-01 2.29086593e-01 2.50747740e-01 5.04105650e-02 1.35594296e+00 3.03343147e-01 7.12814331e-02 -1.18046463e-01 1.04187393e+00 -3.39282721e-01 1.57932058e-01 -2.21142173e-01 -3.57949942e-01 5.78764081e-01 9.35865641e-01 -6.69195175e-01 -7.75892496e-01 -4.43071365e-01 5.47470152e-01 5.20903170e-01 -7.85170719e-02 -8.12659562e-01 -8.47705245e-01 1.25494957e-01 2.42886394e-01 2.41532668e-01 -3.57045352e-01 -1.04804647e+00 -1.02891839e+00 -1.76458910e-01 -1.18255305e+00 4.57522273e-01 -7.60468423e-01 -1.10102713e+00 1.72793373e-01 3.62955689e-01 -6.99448228e-01 -5.01441836e-01 -9.32793081e-01 -6.34638011e-01 1.02320552e+00 -1.56288028e+00 -1.11266935e+00 -1.96020842e-01 4.18654621e-01 3.59488308e-01 -6.44838959e-02 1.08739066e+00 1.63196072e-01 -2.84987003e-01 1.05149639e+00 -3.40863973e-01 3.70120406e-01 7.04055250e-01 -1.67551422e+00 6.31676912e-01 7.78072357e-01 -1.30571276e-01 7.92625964e-01 3.62667590e-01 -5.76423109e-01 -1.37569726e+00 -9.14417803e-01 1.59749126e+00 -3.54334414e-01 5.99009752e-01 -5.95612764e-01 -8.63858163e-01 7.82887340e-01 1.62503570e-01 -3.48849237e-01 5.66069663e-01 3.87329578e-01 -4.78206754e-01 3.64468340e-03 -9.77393091e-01 7.90811300e-01 7.98514962e-01 -2.79363394e-01 -9.33294475e-01 3.66088033e-01 7.21968710e-01 -9.01413739e-01 -7.28054285e-01 2.76236326e-01 5.82758307e-01 -4.66929555e-01 1.03069293e+00 -7.77139187e-01 1.06449962e+00 5.26022241e-02 9.13966596e-02 -1.36839402e+00 -3.05553321e-02 -6.09704971e-01 -3.19254369e-01 1.28821909e+00 6.83075309e-01 -5.81931889e-01 6.50097191e-01 4.49373960e-01 -2.43551731e-01 -7.40340889e-01 -5.02708852e-01 -7.41760552e-01 6.82587981e-01 -6.47150755e-01 9.71545696e-01 9.15963888e-01 7.06364512e-02 5.96606195e-01 -1.37568370e-01 -3.50068748e-01 1.97886989e-01 -3.35346729e-01 9.92624402e-01 -1.09578335e+00 -3.05994689e-01 -5.38876712e-01 1.43693928e-02 -1.10059643e+00 2.06876040e-01 -1.13150156e+00 3.66890691e-02 -1.78337610e+00 -5.42274453e-02 -7.39056826e-01 -8.95852074e-02 8.29755545e-01 -4.78899658e-01 -2.47608483e-01 4.31067973e-01 -3.71999085e-01 -4.74142134e-01 2.17738137e-01 1.27587509e+00 2.71466817e-03 -2.26587489e-01 -1.12744123e-01 -8.18233728e-01 6.35693729e-01 9.32257116e-01 -3.05784225e-01 -3.50324005e-01 -7.67480254e-01 7.73100853e-01 -1.07424997e-03 1.09592266e-01 -1.28457761e+00 2.37609759e-01 -1.34345725e-01 6.82939351e-01 -8.91477346e-01 1.08061716e-01 -3.38826805e-01 -6.42054558e-01 5.32566786e-01 -7.06326663e-01 4.36305791e-01 7.63709903e-01 -1.66823268e-01 2.00588796e-02 -5.90752721e-01 5.80813289e-01 -3.65830094e-01 -5.93962967e-01 -2.55509436e-01 -4.53101456e-01 6.15558684e-01 3.94817531e-01 -5.21425568e-02 -4.79611456e-01 -2.73188442e-01 -2.01382518e-01 1.91423044e-01 3.45880538e-02 2.00456887e-01 4.07459110e-01 -1.14851999e+00 -9.36533809e-01 2.58734643e-01 -1.04340881e-01 3.87784779e-01 -3.79110761e-02 6.22118950e-01 -9.04697955e-01 9.25688386e-01 -4.35682833e-01 -1.88489377e-01 -1.08181500e+00 5.81813037e-01 3.42024267e-02 -1.08318460e+00 -1.30342394e-01 1.13889015e+00 -1.88705891e-01 -9.79805350e-01 4.41357255e-01 -1.17391264e+00 -1.36350527e-01 9.83348265e-02 6.29197478e-01 4.45928246e-01 3.67489249e-01 -4.89344494e-03 -1.35702595e-01 1.43173873e-01 -3.07787865e-01 -1.33679748e-01 1.22552156e+00 4.48732048e-01 -2.33347654e-01 5.32221794e-01 8.78329158e-01 -4.88325283e-02 -5.98146379e-01 -2.75414854e-01 1.46311030e-01 -1.01889402e-01 -2.82234460e-01 -1.26568162e+00 -4.60102111e-01 1.12066936e+00 -1.84232637e-01 -1.15969755e-01 9.21550512e-01 -4.86542493e-01 8.81880701e-01 8.55206549e-01 8.39444920e-02 -1.07849932e+00 7.32664838e-02 1.19516969e+00 8.38774264e-01 -1.17894387e+00 2.64323324e-01 -4.97254729e-01 -4.80465055e-01 1.40325868e+00 9.29103434e-01 2.29013458e-01 2.18450621e-01 4.69270498e-01 -2.57495016e-01 2.60627240e-01 -8.46079528e-01 -1.45732611e-01 5.70964992e-01 5.18570960e-01 9.54720795e-01 2.00101770e-02 -3.31025600e-01 8.52306426e-01 -9.21661675e-01 2.40401328e-01 5.83746254e-01 1.15356612e+00 -1.28875077e-01 -1.12230659e+00 -4.24659699e-01 7.44951010e-01 -4.62592065e-01 -9.07976627e-01 -3.26364845e-01 1.09902060e+00 3.30536097e-01 6.05762005e-01 2.39680097e-01 -1.44878194e-01 1.71099499e-01 4.13877696e-01 1.02374339e+00 -9.30776536e-01 -8.11699033e-01 -5.50575554e-01 3.14137489e-01 -9.16809365e-02 -6.95725307e-02 -4.20637697e-01 -1.42038691e+00 -6.62993729e-01 -3.39690536e-01 5.17770573e-02 4.57731336e-01 1.16852117e+00 2.07017660e-02 8.23061466e-01 -2.76804060e-01 -6.82442725e-01 -7.56877720e-01 -1.22455513e+00 -1.13047145e-01 -5.84444543e-03 -9.04515013e-03 -3.77979428e-01 -1.36282533e-01 -1.66382492e-01]
[9.672320365905762, 7.432043075561523]
5491aac3-e854-4dec-8464-b452a384a08f
large-language-models-can-be-easily
2302.00093
null
https://arxiv.org/abs/2302.00093v3
https://arxiv.org/pdf/2302.00093v3.pdf
Large Language Models Can Be Easily Distracted by Irrelevant Context
Large language models have achieved impressive performance on various natural language processing tasks. However, so far they have been evaluated primarily on benchmarks where all information in the input context is relevant for solving the task. In this work, we investigate the distractibility of large language models, i.e., how the model problem-solving accuracy can be influenced by irrelevant context. In particular, we introduce Grade-School Math with Irrelevant Context (GSM-IC), an arithmetic reasoning dataset with irrelevant information in the problem description. We use this benchmark to measure the distractibility of cutting-edge prompting techniques for large language models, and find that the model performance is dramatically decreased when irrelevant information is included. We also identify several approaches for mitigating this deficiency, such as decoding with self-consistency and adding to the prompt an instruction that tells the language model to ignore the irrelevant information.
['Denny Zhou', 'Nathanael Schärli', 'Ed Chi', 'David Dohan', 'Nathan Scales', 'Kanishka Misra', 'Xinyun Chen', 'Freda Shi']
2023-01-31
null
null
null
null
['arithmetic-reasoning']
['reasoning']
[ 1.27499759e-01 1.04963601e-01 -1.81773808e-02 -3.23561847e-01 -5.93519032e-01 -5.74705958e-01 4.24482524e-01 8.38582814e-01 -4.22074139e-01 3.54862809e-01 4.13053960e-01 -7.68820405e-01 -1.58886671e-01 -7.28860259e-01 -7.17181623e-01 -2.23836407e-01 4.26294744e-01 2.23192513e-01 3.68222862e-01 -4.37036484e-01 7.55753577e-01 1.59431860e-01 -1.42550731e+00 6.12780511e-01 1.30763268e+00 2.24335879e-01 3.85353446e-01 6.36059284e-01 -5.79816043e-01 1.53997660e+00 -8.96976113e-01 -1.73011094e-01 -3.25731598e-02 -3.58741403e-01 -8.08823645e-01 -4.17857319e-01 8.52452219e-01 -3.38408768e-01 -1.91308185e-01 1.12307429e+00 4.33943063e-01 1.53991625e-01 3.73977154e-01 -9.99985158e-01 -6.37421191e-01 9.74249840e-01 -3.08541417e-01 5.79289615e-01 5.73551416e-01 3.96749258e-01 5.12360632e-01 -5.84231973e-01 3.24404359e-01 1.53455758e+00 3.25160295e-01 5.10288775e-01 -1.39833510e+00 -6.38761342e-01 5.21774232e-01 3.73488337e-01 -1.43910289e+00 -4.57382470e-01 7.18990564e-01 -6.68580651e-01 1.24014437e+00 3.65751773e-01 2.85597056e-01 9.71477866e-01 7.68129766e-01 5.55583000e-01 1.36702049e+00 -7.74667799e-01 2.96968728e-01 8.41351673e-02 9.88921881e-01 6.43116891e-01 5.70074797e-01 -1.03475731e-02 -5.73588908e-01 1.15244242e-03 2.76478469e-01 -4.20551270e-01 -5.10103479e-02 1.26971513e-01 -9.65431154e-01 5.04343927e-01 -5.27931936e-02 3.33094090e-01 -7.45511279e-02 1.43950865e-01 1.68000147e-01 7.44074345e-01 3.07535648e-01 8.35435092e-01 -3.99503171e-01 -3.64385724e-01 -6.81187868e-01 5.42199790e-01 7.88047433e-01 9.54259634e-01 4.87251371e-01 1.17636994e-01 -6.27584159e-01 7.23532498e-01 2.30674654e-01 3.34728062e-01 7.04704106e-01 -1.01554811e+00 8.27871025e-01 9.61840868e-01 -2.58735642e-02 -1.05900717e+00 -7.03038335e-01 -3.90225172e-01 -3.84434730e-01 6.69500791e-04 8.87929499e-01 -5.23357615e-02 -3.34568381e-01 2.14464927e+00 -9.30968747e-02 -2.16971543e-02 1.93143398e-01 5.77949703e-01 9.68147278e-01 4.60108310e-01 5.59396684e-01 -7.21985623e-02 1.27729332e+00 -1.07974100e+00 -9.97892082e-01 -8.20895493e-01 1.38126254e+00 -1.00778222e+00 1.47514153e+00 4.76840854e-01 -1.29587066e+00 -6.51328921e-01 -1.02284050e+00 -4.26829517e-01 -3.19359869e-01 -2.16509447e-01 3.07866395e-01 4.60413486e-01 -1.05658293e+00 4.64767456e-01 -4.63841289e-01 -1.07773365e-02 -1.87333286e-01 1.03898220e-01 -8.44808891e-02 -5.94237983e-01 -1.11193037e+00 1.23197937e+00 4.93004203e-01 -4.19481397e-01 -5.37031949e-01 -8.89392972e-01 -9.62236047e-01 5.16383827e-01 6.20712936e-01 -4.73852992e-01 1.38664961e+00 -1.02237475e+00 -1.26001585e+00 6.52209759e-01 -3.29142630e-01 -1.36900917e-01 2.83739626e-01 -4.07360584e-01 -3.05932164e-01 -3.00713450e-01 6.30636141e-02 4.03524905e-01 4.11538839e-01 -9.26523805e-01 -1.19992845e-01 -4.07356173e-01 3.56674582e-01 2.28204504e-02 -2.72136956e-01 2.13721022e-01 -1.58117533e-01 -9.04552996e-01 3.50092828e-01 -7.66596258e-01 -1.93645343e-01 -4.20992315e-01 -1.36211231e-01 -4.71482962e-01 3.55937570e-01 -6.77375853e-01 1.87292325e+00 -1.99749744e+00 -5.86960651e-02 -2.33578384e-02 2.69817322e-01 3.22405457e-01 -7.17887521e-01 4.26807612e-01 -2.61891842e-01 3.99087876e-01 1.74613744e-01 7.31172636e-02 5.86736836e-02 1.88626554e-02 -2.60292143e-01 2.50425600e-02 8.84894803e-02 9.71717417e-01 -8.62323225e-01 -3.85577589e-01 4.83430959e-02 -7.22624082e-03 -8.94883037e-01 2.30489433e-01 -5.53439975e-01 6.75861016e-02 -2.88917333e-01 3.01959813e-01 7.20781207e-01 -2.53463387e-01 1.54338524e-01 4.23386961e-01 -5.41994981e-02 8.66832078e-01 -1.15867388e+00 1.48552537e+00 -3.49988371e-01 6.57178342e-01 -5.18999919e-02 -6.09660745e-01 8.34414005e-01 1.37025550e-01 -2.76384920e-01 -1.25306034e+00 -1.31252035e-01 -1.97930671e-02 8.40539634e-01 -7.75045514e-01 6.23611152e-01 2.17157856e-01 -4.10972536e-03 6.23308063e-01 -3.33657503e-01 -9.97547731e-02 3.26316416e-01 4.03722823e-01 1.17677546e+00 -1.26997679e-01 4.11094934e-01 -7.18963087e-01 7.11159468e-01 1.37746772e-02 3.05213094e-01 1.17273235e+00 -2.21359432e-01 1.47168934e-01 7.10640311e-01 -2.75062680e-01 -6.11121356e-01 -6.33142471e-01 2.91067868e-01 1.47467756e+00 5.02004959e-02 -1.08483744e+00 -8.53524029e-01 -4.07979310e-01 -6.07750602e-02 1.34882677e+00 -1.40325874e-01 -7.60297000e-01 -8.52240980e-01 -3.92502725e-01 5.67268372e-01 5.73461235e-01 2.81931251e-01 -7.32668221e-01 -3.93113732e-01 2.17985258e-01 -2.53468990e-01 -1.06618214e+00 -2.94339478e-01 1.01731859e-01 -9.42743719e-01 -9.36413884e-01 1.23116694e-01 -5.96028864e-01 7.28752732e-01 4.37287927e-01 1.60006130e+00 7.50703275e-01 1.12842582e-01 3.34325343e-01 -1.39888274e-02 -5.03506839e-01 -6.24185026e-01 -1.99670196e-02 -1.84999064e-01 -8.09752226e-01 8.07183743e-01 -1.19886391e-01 -4.49852496e-02 9.27294940e-02 -7.94590890e-01 4.22276407e-01 3.06247681e-01 5.28356314e-01 5.22623174e-02 -1.23267561e-01 4.64476645e-01 -8.43449771e-01 1.06487656e+00 -4.94996071e-01 -6.70285702e-01 3.37553233e-01 -5.49997985e-01 4.17326391e-01 8.95676374e-01 -5.22123277e-01 -8.86813760e-01 -5.08123934e-01 -2.43454859e-01 1.10528193e-01 -3.25497836e-01 5.89534402e-01 -1.69796094e-01 -1.73786715e-01 6.36077881e-01 2.00640216e-01 -4.35749441e-01 -4.24953938e-01 -9.10288170e-02 2.65451699e-01 1.40964836e-01 -1.28222334e+00 4.06517923e-01 -3.35524738e-01 -6.14764988e-02 -6.63268209e-01 -1.02200127e+00 -2.04155445e-01 -3.19401056e-01 -8.06937367e-03 4.93669331e-01 -9.04725909e-01 -7.07598329e-01 3.38655770e-01 -1.34859264e+00 -6.09242201e-01 1.13005474e-01 5.08247375e-01 -3.47978413e-01 3.40442806e-01 -7.85930097e-01 -7.16292858e-01 2.41025370e-02 -1.56163883e+00 5.97678542e-01 1.28929004e-01 -6.19985402e-01 -8.72282922e-01 -1.17702827e-01 5.17013013e-01 6.22149110e-01 -3.04579943e-01 1.71818006e+00 -1.07166791e+00 -7.49321699e-01 2.06939161e-01 -1.76200807e-01 7.28079826e-02 -3.04465860e-01 -2.04842597e-01 -1.00756907e+00 -9.71375182e-02 2.17027217e-01 -3.03757131e-01 6.29453301e-01 6.56785965e-02 1.43101072e+00 -4.38845247e-01 3.83229144e-02 2.18114674e-01 1.31898892e+00 1.90124378e-01 4.90190148e-01 2.85584778e-01 6.81708932e-01 7.22888231e-01 5.38042307e-01 1.32873729e-02 6.01913810e-01 6.34045184e-01 -1.13096558e-01 4.04595375e-01 -1.83361262e-01 -4.18565691e-01 4.88941222e-01 1.10990167e+00 3.87904823e-01 -2.68549800e-01 -1.36598301e+00 3.57084841e-01 -1.62279999e+00 -7.15036094e-01 -4.25958455e-01 2.18523765e+00 1.00521338e+00 3.58968079e-01 -5.73604584e-01 4.21205461e-02 2.61626273e-01 1.17280513e-01 -2.62001067e-01 -7.02379048e-01 2.13723108e-02 1.46249667e-01 1.53722897e-01 1.03871703e+00 -5.13085723e-01 1.12361956e+00 6.73127508e+00 9.47815001e-01 -1.00890398e+00 9.56855640e-02 5.35512030e-01 -1.49608567e-01 -5.35620093e-01 1.08508551e-02 -8.35394025e-01 5.26395500e-01 1.26532519e+00 -3.09758395e-01 2.73573786e-01 6.43498898e-01 4.54394639e-01 -5.47160089e-01 -1.55211961e+00 6.59122348e-01 9.92181748e-02 -1.01519191e+00 2.62111783e-01 -3.08027774e-01 5.66984475e-01 -4.75984037e-01 -1.64083689e-01 7.69412935e-01 2.88378179e-01 -1.12566745e+00 6.35987759e-01 5.65458000e-01 2.13308148e-02 -8.06726158e-01 6.00383759e-01 1.04467142e+00 -7.13199794e-01 -1.49901003e-01 -3.62694144e-01 -8.95770669e-01 -4.67328906e-01 2.70990491e-01 -5.70145965e-01 -2.41354797e-02 2.07708076e-01 4.29818667e-02 -1.28009808e+00 7.89936304e-01 -4.91860390e-01 7.89137602e-01 -6.13713264e-02 -1.48575768e-01 1.03257455e-01 -5.68527058e-02 3.99773926e-01 1.15248859e+00 5.38601354e-02 3.82497281e-01 3.88411313e-01 1.02399433e+00 1.21116728e-01 2.84929425e-01 -7.03835130e-01 -3.54267806e-02 6.13018334e-01 6.67018056e-01 -5.03335774e-01 -5.64895570e-01 -5.36584079e-01 3.33456039e-01 6.50563061e-01 2.98348665e-01 -5.21726906e-01 -1.59031190e-02 6.38800740e-01 2.53803760e-01 -3.25580806e-01 -3.41080576e-01 -7.28860676e-01 -1.23041475e+00 -1.95764080e-02 -1.24516439e+00 2.50237733e-01 -1.01022339e+00 -8.23103845e-01 -8.02981332e-02 1.80947021e-01 -3.67121100e-01 -1.86845347e-01 -6.29461110e-01 -6.51056945e-01 1.05895090e+00 -1.46969461e+00 -4.96967494e-01 -3.47371668e-01 3.33467484e-01 7.37990379e-01 -6.22501373e-02 7.12052166e-01 2.26227641e-01 -6.16213500e-01 7.68161714e-01 -1.94252312e-01 4.06834409e-02 9.19311583e-01 -1.14043939e+00 2.92889208e-01 1.08254099e+00 -1.92090541e-01 1.13456273e+00 8.27290893e-01 -7.25339890e-01 -1.58562207e+00 -8.09153855e-01 1.50522411e+00 -7.66153634e-01 5.38479686e-01 -4.18239683e-01 -1.34630692e+00 8.85328650e-01 2.59393662e-01 -3.19514424e-01 7.27819026e-01 1.07121805e-03 -4.76553023e-01 1.51219293e-01 -9.69403803e-01 1.00293040e+00 8.88019681e-01 -4.90276039e-01 -1.01761270e+00 3.22162330e-01 1.05113912e+00 -5.46109140e-01 -5.07788241e-01 1.06681548e-01 1.59392223e-01 -8.52809072e-01 9.32348013e-01 -9.30993438e-01 8.47193539e-01 -4.76668775e-02 -2.51508564e-01 -1.24380457e+00 -5.58213651e-01 -2.39320397e-01 -1.41662642e-01 1.12450075e+00 2.07586333e-01 -4.79576021e-01 3.37332428e-01 1.13090169e+00 -2.61698037e-01 -5.02889037e-01 -5.18855512e-01 -5.87041020e-01 5.99719763e-01 -5.34925878e-01 6.53693020e-01 9.98562753e-01 2.15874359e-01 4.83540982e-01 1.11630023e-01 7.47260004e-02 2.08376259e-01 8.01898837e-02 6.24445438e-01 -1.10342455e+00 -9.90748107e-02 -6.29632711e-01 3.21079530e-02 -1.08098555e+00 4.78344351e-01 -9.84551787e-01 -1.21382318e-01 -1.21425545e+00 2.02212155e-01 -3.34849626e-01 -1.48663431e-01 4.42254215e-01 -5.96064866e-01 -3.77685905e-01 6.04568899e-01 -6.77511841e-02 -8.53837967e-01 9.74826589e-02 1.24868488e+00 -2.12654948e-01 -8.76578242e-02 -3.40907514e-01 -9.44069862e-01 1.02134871e+00 8.05447876e-01 -5.59006274e-01 -5.30757368e-01 -7.85765827e-01 4.75284606e-01 1.02046646e-01 2.35951245e-01 -1.10297692e+00 4.69531029e-01 -4.66835022e-01 2.33482614e-01 -3.24694872e-01 -3.75725180e-02 -8.49676430e-01 -3.76908749e-01 6.01856172e-01 -1.00983226e+00 6.64043188e-01 7.90983975e-01 5.29423989e-02 -8.38694870e-02 -7.42970824e-01 5.39836586e-01 -3.28867704e-01 -8.75273824e-01 -5.75865328e-01 -7.26143301e-01 5.84173501e-01 8.25547457e-01 2.11902469e-01 -8.56801331e-01 -1.22421175e-01 -3.40560347e-01 3.15673739e-01 3.64641368e-01 5.88000894e-01 4.51190740e-01 -1.14209294e+00 -7.26688027e-01 3.71556371e-01 2.40045801e-01 -2.06622332e-01 2.69166797e-01 8.26468706e-01 -5.04616678e-01 9.26222026e-01 -1.09151348e-01 -4.09945875e-01 -1.54858816e+00 8.88845384e-01 1.40652090e-01 -4.37992513e-01 -2.60720283e-01 5.50327837e-01 3.06641102e-01 -4.59528476e-01 3.40166032e-01 -9.49644625e-01 -2.42093593e-01 -1.47940457e-01 8.18441391e-01 2.82809645e-01 3.96512866e-01 -2.01392114e-01 -1.33830279e-01 3.06647599e-01 -1.94529846e-01 3.42933774e-01 8.56646001e-01 -5.12026995e-02 -3.03229719e-01 6.55933261e-01 6.46568716e-01 2.76452363e-01 -5.13690829e-01 -3.39382589e-01 4.07626599e-01 -4.77875799e-01 3.76013517e-02 -1.15612030e+00 -5.04815459e-01 9.46621001e-01 2.61508524e-01 2.86134444e-02 8.10030580e-01 -3.33855391e-01 1.38262764e-01 9.30598378e-01 2.89164186e-01 -1.07943892e+00 5.43326810e-02 1.18505704e+00 8.46550405e-01 -1.23156297e+00 -2.79096086e-02 -4.39898580e-01 -1.67203978e-01 1.07744503e+00 1.35853803e+00 1.01535007e-01 2.05036759e-01 5.01782596e-01 9.59055796e-02 -8.04041848e-02 -1.35060394e+00 3.07224274e-01 1.65402129e-01 2.69522727e-01 9.75405216e-01 -7.51470029e-03 -7.31509686e-01 1.01566529e+00 -4.95648324e-01 5.98084852e-02 8.25719059e-01 1.10746586e+00 -5.91411531e-01 -8.68347704e-01 -7.70426989e-01 3.89758945e-01 -4.26632106e-01 -5.52820265e-01 -6.09537542e-01 7.89060175e-01 -2.71764733e-02 1.06391478e+00 5.41453697e-02 -8.91398117e-02 3.28630120e-01 5.93440354e-01 5.30268788e-01 -8.83324385e-01 -1.10869575e+00 -3.11342269e-01 1.10240079e-01 -7.24247277e-01 2.69320190e-01 -5.63308559e-02 -1.15531969e+00 -6.18337512e-01 -1.33224994e-01 7.03703091e-02 2.80538440e-01 1.18290818e+00 3.83384228e-01 8.31108034e-01 -3.04137200e-01 -1.43408865e-01 -9.55741942e-01 -9.68669593e-01 -5.14311641e-02 2.21553624e-01 3.68503183e-01 -4.73473787e-01 -2.80999094e-01 -3.14069569e-01]
[9.855569839477539, 7.513607025146484]
8076d460-6b01-4ef6-bbe7-899a717e6d4a
the-behavior-and-convergence-of-local
2305.15572
null
https://arxiv.org/abs/2305.15572v1
https://arxiv.org/pdf/2305.15572v1.pdf
The Behavior and Convergence of Local Bayesian Optimization
A recent development in Bayesian optimization is the use of local optimization strategies, which can deliver strong empirical performance on high-dimensional problems compared to traditional global strategies. The "folk wisdom" in the literature is that the focus on local optimization sidesteps the curse of dimensionality; however, little is known concretely about the expected behavior or convergence of Bayesian local optimization routines. We first study the behavior of the local approach, and find that the statistics of individual local solutions of Gaussian process sample paths are surprisingly good compared to what we would expect to recover from global methods. We then present the first rigorous analysis of such a Bayesian local optimization algorithm recently proposed by M\"uller et al. (2021), and derive convergence rates in both the noisy and noiseless settings.
['Jacob R. Gardner', 'Roman Garnett', 'Kyurae Kim', 'Kaiwen Wu']
2023-05-24
null
null
null
null
['bayesian-optimization']
['methodology']
[-1.70236990e-01 -8.14921632e-02 1.65854543e-01 -3.36104065e-01 -1.31764674e+00 -3.99242967e-01 7.68456221e-01 -1.01468235e-01 -4.43800628e-01 9.41821873e-01 2.23203406e-01 -3.17709923e-01 -6.54544950e-01 -5.13653517e-01 -5.57997882e-01 -1.19688976e+00 -1.27668872e-01 6.11034751e-01 1.17326893e-01 3.93263638e-01 4.01488274e-01 3.46860051e-01 -1.09714162e+00 -6.29638612e-01 5.82271159e-01 8.14373970e-01 1.29369363e-01 1.03844404e+00 3.00171286e-01 2.71149069e-01 -4.64663893e-01 -3.07012409e-01 2.22645879e-01 -5.12354255e-01 -5.49423039e-01 3.13443355e-02 2.44441718e-01 -1.97887719e-01 -2.36807331e-01 1.44676435e+00 8.09726715e-01 4.06159997e-01 8.77123356e-01 -6.41884685e-01 -3.40568006e-01 4.65429664e-01 -4.11208183e-01 3.16659182e-01 1.66032419e-01 3.22381347e-01 1.11470270e+00 -8.03924382e-01 4.41918999e-01 1.51928079e+00 8.78440976e-01 5.79602271e-02 -1.46155190e+00 -2.70114481e-01 6.24976978e-02 -1.99043125e-01 -1.60582685e+00 -7.49955475e-01 3.03495288e-01 -6.71254516e-01 3.23936999e-01 3.56205143e-02 2.90679991e-01 1.12704349e+00 6.30362093e-01 7.04184651e-01 1.27193952e+00 -2.95250982e-01 4.78157759e-01 1.63463317e-02 3.90155643e-01 5.99737167e-01 3.19962651e-01 4.75530863e-01 -8.17312002e-01 -5.84414005e-01 6.57847166e-01 -4.98171687e-01 -3.82476628e-01 -3.15436542e-01 -1.08212650e+00 9.13523257e-01 -1.07638426e-01 1.71113461e-01 -3.21387172e-01 5.86493552e-01 9.26214531e-02 5.54417679e-03 7.03780293e-01 2.66083807e-01 -5.49069166e-01 -6.46090925e-01 -8.42553794e-01 6.53860211e-01 9.48264480e-01 7.59103894e-01 5.44237554e-01 4.04636078e-02 -9.39234197e-02 6.81672752e-01 8.87689292e-01 8.59369040e-01 5.78510612e-02 -1.33098459e+00 3.64918202e-01 -4.34215993e-01 5.91592908e-01 -9.68310356e-01 -3.76125753e-01 -6.78194940e-01 -5.32659411e-01 1.88779995e-01 7.87219286e-01 -5.21122396e-01 -4.75738764e-01 1.58951533e+00 3.00079346e-01 3.56171615e-02 -1.72916457e-01 6.53506696e-01 9.09383520e-02 7.21879423e-01 -2.52852976e-01 -6.29511416e-01 7.18497932e-01 -5.11946201e-01 -9.24218416e-01 -1.71592981e-01 4.11967635e-01 -8.85892093e-01 7.38993347e-01 5.36758065e-01 -1.22189057e+00 -2.11912140e-01 -8.73231053e-01 3.49062026e-01 1.72959477e-01 -3.04346353e-01 5.14873981e-01 1.01340389e+00 -9.60080504e-01 7.99995542e-01 -1.14375687e+00 -3.24839205e-01 2.43438482e-01 1.34581953e-01 2.11688995e-01 -4.41825315e-02 -4.77265090e-01 8.80356133e-01 1.44507110e-01 3.44516695e-01 -1.24141479e+00 -5.92770576e-01 -4.72213060e-01 -2.05800813e-02 6.86350346e-01 -5.74576974e-01 1.55238938e+00 -3.14859033e-01 -1.77212846e+00 1.73982918e-01 -6.55780792e-01 -2.52237231e-01 5.42316437e-01 -4.67849255e-01 6.39597997e-02 -1.09805986e-01 -1.01161678e-03 -1.44940048e-01 9.45214927e-01 -1.26643348e+00 -3.67881924e-01 -4.20151442e-01 -5.33445716e-01 2.35189959e-01 2.13907138e-01 3.41741174e-01 -3.98899943e-01 -5.19530833e-01 3.75716329e-01 -9.56859350e-01 -6.25688851e-01 -1.67067930e-01 -4.88002986e-01 -1.94973916e-01 4.08511937e-01 -5.47990322e-01 1.02035499e+00 -2.10207462e+00 2.51979470e-01 5.71605623e-01 1.22553699e-01 -4.28860456e-01 1.12892643e-01 3.18628371e-01 1.25854552e-01 1.45124808e-01 -3.88892353e-01 -5.80702364e-01 3.54478717e-01 1.53555125e-01 -3.92931730e-01 9.78414834e-01 -3.18788648e-01 7.63640046e-01 -9.36710179e-01 -3.25879276e-01 6.19165897e-02 1.19591825e-01 -4.37994480e-01 -5.68869263e-02 2.50213109e-02 2.84382135e-01 -5.09734511e-01 5.63632786e-01 4.63382930e-01 -3.93342316e-01 1.87084123e-01 1.59627259e-01 -1.93707958e-01 3.53400648e-01 -1.49957812e+00 1.34969521e+00 -7.33979642e-02 6.56182945e-01 6.04899526e-01 -9.08175409e-01 4.67683077e-01 4.18131888e-01 5.00037789e-01 5.60163632e-02 3.53850693e-01 1.92793906e-01 -1.22146197e-01 -1.15155622e-01 2.30251148e-01 -4.55725223e-01 -6.24682829e-02 4.36038524e-01 8.55963081e-02 -3.87626886e-01 -1.28156409e-01 -4.73365150e-02 1.09652436e+00 3.36503863e-01 3.19323331e-01 -7.70683169e-01 -4.54880893e-02 -1.90099925e-01 5.10490060e-01 1.59966969e+00 -2.67188877e-01 5.83888292e-01 4.30822223e-01 1.94493309e-02 -7.83440650e-01 -1.42741621e+00 -6.01631939e-01 8.94755304e-01 -1.19328052e-01 -6.17469072e-01 -6.56050384e-01 -3.38336825e-01 -1.04072124e-01 6.48290813e-01 -3.32927525e-01 5.10365181e-02 -3.36261332e-01 -1.55848718e+00 3.76725703e-01 3.45961362e-01 1.62956059e-01 -4.12967950e-01 -2.95121461e-01 3.91277254e-01 1.79633740e-02 -1.02409446e+00 -3.44883800e-01 2.20215663e-01 -1.21034288e+00 -7.90794969e-01 -5.43650687e-01 1.28711894e-01 1.43790409e-01 -1.12380214e-01 8.48688006e-01 -5.28998971e-01 -1.51721686e-01 5.84039092e-01 1.11393958e-01 -3.89987648e-01 -2.92888582e-01 -2.59416163e-01 4.56683367e-01 -1.04303010e-01 2.85613716e-01 -4.54386383e-01 -3.76244783e-01 2.80283570e-01 -2.78521329e-01 -7.56869256e-01 4.92249161e-01 6.83931828e-01 3.63323659e-01 3.18807006e-01 2.11874887e-01 -4.93034601e-01 8.63566637e-01 -4.90983456e-01 -1.01076746e+00 -2.29036286e-02 -7.04505563e-01 2.34731376e-01 8.07417408e-02 -1.90207556e-01 -1.21870267e+00 -1.87821761e-01 -4.75363769e-02 8.99895746e-03 -8.29125941e-02 7.08103836e-01 -4.85395864e-02 -1.18862726e-01 6.53063476e-01 4.09807526e-02 -6.50200620e-02 -7.27905750e-01 3.29590529e-01 4.34708208e-01 5.20193815e-01 -1.20454109e+00 8.76827955e-01 5.12531281e-01 3.44499826e-01 -1.14124346e+00 -9.82233942e-01 -4.99198288e-01 -3.37700188e-01 -1.06828213e-01 7.83241987e-01 -6.51729643e-01 -8.54310036e-01 3.90075952e-01 -1.06621277e+00 -3.32206309e-01 -3.74296695e-01 7.51233399e-01 -8.43982995e-01 6.14698946e-01 -3.79758209e-01 -1.33345437e+00 3.42834741e-01 -1.28024244e+00 9.50515270e-01 -6.75145583e-03 -3.20217580e-01 -1.26430583e+00 3.44409466e-01 3.75520289e-02 4.19297189e-01 -8.56200457e-02 6.55941486e-01 -4.72023636e-01 -7.49204814e-01 -2.89706886e-01 -6.43378422e-02 3.28432381e-01 -1.02166913e-01 1.36396244e-01 -9.14001107e-01 -1.46125779e-01 6.19033217e-01 1.06076412e-01 7.01851070e-01 1.03220952e+00 8.53543043e-01 -1.82049483e-01 -2.74094731e-01 5.58166981e-01 1.53182602e+00 7.01051056e-02 2.82525212e-01 7.70858824e-02 3.46092761e-01 5.56950390e-01 3.91332656e-01 6.63640976e-01 -6.60766885e-02 6.55724883e-01 2.84862821e-03 5.75905740e-01 1.94855079e-01 -1.88949220e-02 5.74152827e-01 8.53377998e-01 -2.44579107e-01 -1.87889099e-01 -9.40599740e-01 5.08135498e-01 -2.15265274e+00 -1.05613494e+00 -2.44714081e-01 2.46703339e+00 8.40013444e-01 1.94197819e-01 -4.89490256e-02 -2.11327583e-01 6.22541249e-01 1.88949957e-01 -2.91943073e-01 -2.89587099e-02 -9.25296843e-02 1.96957961e-01 8.63793910e-01 1.07587266e+00 -1.19811654e+00 6.12349093e-01 8.79669857e+00 1.10388672e+00 -1.74568743e-01 3.76243711e-01 4.55837607e-01 -1.86220646e-01 8.57845768e-02 1.83458507e-01 -1.30444932e+00 4.37556654e-01 1.37773561e+00 -9.61347967e-02 3.16832632e-01 7.41841853e-01 6.10985518e-01 -5.51968455e-01 -1.02743196e+00 8.76465440e-01 -5.62216192e-02 -1.01310909e+00 -4.02971566e-01 5.70797801e-01 9.66975152e-01 1.69303507e-01 7.66138881e-02 -3.26938927e-02 9.43433642e-01 -1.18052506e+00 4.78454411e-01 1.04785323e+00 -6.18306808e-02 -7.37275779e-01 8.13259661e-01 5.44430733e-01 -7.66217113e-01 -2.34774485e-01 -5.24472117e-01 -1.06327673e-02 6.77570701e-01 1.14708650e+00 -5.04216850e-01 2.71523714e-01 7.80441403e-01 5.23348868e-01 -2.74258584e-01 1.39365363e+00 -1.99809507e-01 9.32114005e-01 -9.83596325e-01 -6.54963478e-02 3.21497411e-01 -7.62525737e-01 1.21240413e+00 1.10340118e+00 5.44008076e-01 -6.25587553e-02 3.48101296e-02 7.57519126e-01 3.39718223e-01 -9.55374092e-02 -6.29085124e-01 -1.80592299e-01 2.51532406e-01 8.91242862e-01 -4.55665171e-01 -1.43477678e-01 -2.92962074e-01 2.69143343e-01 -1.30297914e-01 5.97689271e-01 -4.72609699e-01 -1.21632703e-01 7.09629774e-01 -1.75588936e-01 5.01790106e-01 -7.93564975e-01 -6.07680500e-01 -1.12919247e+00 -1.61011249e-01 -8.17463815e-01 1.85303316e-01 -6.88749254e-01 -1.50727642e+00 -2.02761278e-01 3.01431030e-01 -4.53029603e-01 -3.92126858e-01 -8.59573960e-01 -4.21834439e-01 9.78342116e-01 -7.27589309e-01 -4.87746090e-01 2.74952382e-01 3.34238499e-01 4.38524127e-01 -4.37777191e-02 4.09307241e-01 -1.85448170e-01 -4.41201210e-01 3.35688323e-01 9.83475208e-01 -2.45596468e-01 6.12487316e-01 -1.37006724e+00 -7.87039194e-03 1.08822668e+00 4.62318175e-02 1.05758917e+00 1.36586320e+00 -7.25073159e-01 -1.56315792e+00 -4.22952533e-01 3.92511934e-01 -1.01504707e+00 1.04579341e+00 -1.28958449e-01 -5.33100605e-01 7.17022896e-01 2.01451909e-02 -3.91882956e-01 4.36154634e-01 6.57815874e-01 -1.94452349e-02 9.06882510e-02 -8.39595020e-01 6.92396700e-01 7.51124918e-01 -4.53344375e-01 -5.05656242e-01 5.62857270e-01 2.82802641e-01 -9.61131677e-02 -9.57514644e-01 3.25740218e-01 5.14771938e-01 -8.94418299e-01 1.13892901e+00 -4.03223336e-01 8.00640602e-03 -1.81562141e-01 -6.73808277e-01 -1.12899125e+00 -1.26451641e-01 -1.12061775e+00 -2.82431126e-01 9.66506839e-01 2.69268721e-01 -6.72761559e-01 5.77378690e-01 5.27642131e-01 -1.31429642e-01 -3.84103686e-01 -1.31528556e+00 -1.10342610e+00 3.74384582e-01 -7.97784209e-01 -1.08149238e-02 3.85771424e-01 -2.61609346e-01 2.53075063e-01 -3.87308240e-01 4.34099436e-01 1.24130285e+00 -1.54673010e-01 8.18109572e-01 -1.22445750e+00 -8.27589691e-01 -6.90684974e-01 -2.04197198e-01 -1.62510955e+00 -3.62298153e-02 -5.68337500e-01 6.27689123e-01 -1.07497752e+00 3.63414049e-01 -1.59225047e-01 -2.23394074e-02 -2.12112039e-01 -2.78629333e-01 2.68512242e-03 -1.17287926e-01 1.76581562e-01 -4.52658683e-01 6.83402956e-01 9.99389291e-01 2.57663339e-01 -5.50766401e-02 3.43541265e-01 -7.08868802e-01 1.10386741e+00 5.29450059e-01 -8.12152028e-01 -8.56207088e-02 -1.88001245e-01 6.28191948e-01 -1.00822583e-01 4.76701409e-01 -7.36017227e-01 4.06137675e-01 -3.39651465e-01 1.95133999e-01 -6.67451322e-01 5.30020595e-01 -4.68729079e-01 -1.70959845e-01 9.32224244e-02 -1.18789226e-01 -9.77971554e-02 -1.96885481e-01 1.02871001e+00 2.70932429e-02 -8.56136322e-01 8.80922318e-01 -1.34726256e-01 -4.78343889e-02 2.32354939e-01 -7.96838224e-01 1.38158038e-01 5.58350265e-01 1.55504629e-01 -7.58311525e-02 -6.27015769e-01 -9.14909601e-01 7.12016411e-03 3.04930151e-01 -3.57648849e-01 2.95072883e-01 -1.01203120e+00 -7.24235773e-01 -2.70828813e-01 -3.58363420e-01 -8.41129646e-02 -2.24692717e-01 1.10536122e+00 -2.67281950e-01 3.57182860e-01 7.35041678e-01 -8.74113441e-01 -8.22179735e-01 2.95138657e-01 4.37572449e-01 -1.08453212e-02 -4.97043669e-01 8.57652366e-01 9.05015841e-02 -4.96999115e-01 1.53752431e-01 -4.33789007e-02 4.93667930e-01 -7.08546415e-02 4.13881421e-01 7.70985246e-01 -2.75292218e-01 -2.32429460e-01 -1.07617915e-01 7.35697091e-01 1.11733481e-01 -9.78472352e-01 1.17796600e+00 -4.24898922e-01 -3.02560061e-01 1.07641697e+00 1.03963816e+00 2.80496508e-01 -1.31934607e+00 -3.61532480e-01 1.65920734e-01 -6.64293468e-01 3.27140808e-01 -3.50916833e-01 -4.05777216e-01 8.50517809e-01 5.83380640e-01 2.21585706e-01 3.55292410e-01 2.08189070e-01 2.04824299e-01 8.86843324e-01 2.34222054e-01 -1.27246201e+00 -1.29004389e-01 7.95470595e-01 6.91587269e-01 -1.04574060e+00 3.34335834e-01 -1.00388974e-01 -1.41644344e-01 9.57151234e-01 -2.07905754e-01 -2.62131512e-01 1.10261512e+00 2.53358960e-01 -2.89300054e-01 -2.85084635e-01 -6.35214627e-01 9.36160907e-02 1.82947502e-01 5.36244750e-01 8.50842446e-02 -1.76025614e-01 -9.34286043e-02 4.23761487e-01 -2.25569859e-01 -3.49026829e-01 5.23362458e-01 8.28144968e-01 -7.57655621e-01 -7.59712994e-01 -7.73164272e-01 3.79319906e-01 -6.47563517e-01 -2.30840579e-01 1.83350388e-02 6.37331665e-01 -4.48601574e-01 1.15096319e+00 -2.68944502e-01 3.33057910e-01 -2.12628633e-01 4.05793250e-01 5.79111338e-01 -5.37763476e-01 -9.45738405e-02 6.57081664e-01 2.90364146e-01 -7.05808520e-01 -4.06306922e-01 -1.28246677e+00 -4.17167336e-01 -4.03292775e-01 -4.37922925e-01 2.23629907e-01 8.33894253e-01 1.41883349e+00 -6.26197606e-02 2.86729693e-01 2.89523035e-01 -9.73349810e-01 -1.24397898e+00 -9.26036537e-01 -8.14916074e-01 -3.29548448e-01 4.74320024e-01 -4.73826379e-01 -1.04063964e+00 -8.86577815e-02]
[6.876373767852783, 3.8305795192718506]
c0761ef4-eb61-4a60-ac6a-c53b2ab01bbf
offline-reinforcement-learning-with-adaptive
2211.08251
null
https://arxiv.org/abs/2211.08251v1
https://arxiv.org/pdf/2211.08251v1.pdf
Offline Reinforcement Learning with Adaptive Behavior Regularization
Offline reinforcement learning (RL) defines a sample-efficient learning paradigm, where a policy is learned from static and previously collected datasets without additional interaction with the environment. The major obstacle to offline RL is the estimation error arising from evaluating the value of out-of-distribution actions. To tackle this problem, most existing offline RL methods attempt to acquire a policy both ``close" to the behaviors contained in the dataset and sufficiently improved over them, which requires a trade-off between two possibly conflicting targets. In this paper, we propose a novel approach, which we refer to as adaptive behavior regularization (ABR), to balance this critical trade-off. By simply utilizing a sample-based regularization, ABR enables the policy to adaptively adjust its optimization objective between cloning and improving over the policy used to generate the dataset. In the evaluation on D4RL datasets, a widely adopted benchmark for offline reinforcement learning, ABR can achieve improved or competitive performance compared to existing state-of-the-art algorithms.
['Qingyu Qu', 'Xijun Li', 'Yunfan Zhou']
2022-11-15
null
null
null
null
['d4rl']
['robots']
[ 1.35862723e-01 -2.48815101e-02 -4.17661130e-01 -2.51489580e-01 -8.81738603e-01 -5.46735823e-01 4.59066123e-01 4.90481973e-01 -9.00532126e-01 1.12282264e+00 -1.25010550e-01 -1.09963492e-01 -2.76651472e-01 -7.73199141e-01 -8.91527712e-01 -9.38736796e-01 -1.52759477e-01 6.01844072e-01 2.18539149e-01 -5.29439636e-02 3.85874003e-01 4.21401680e-01 -1.65355170e+00 -1.52424708e-01 1.01192474e+00 1.02257574e+00 3.54119569e-01 4.03340459e-01 -1.23233506e-02 8.90840471e-01 -7.47998238e-01 1.26480341e-01 3.05640012e-01 -5.90896249e-01 -4.84008402e-01 2.72316616e-02 -1.69081241e-01 -5.82454920e-01 -2.10315501e-03 1.17635727e+00 5.99706888e-01 4.95186031e-01 3.68788302e-01 -9.93831635e-01 -1.80662885e-01 7.65672743e-01 -5.64769506e-01 1.08613789e-01 3.51030797e-01 4.99718130e-01 8.70304346e-01 -1.51982531e-01 5.31409264e-01 1.19571662e+00 6.26240149e-02 7.20061362e-01 -1.42515445e+00 -6.53779149e-01 4.87096518e-01 2.14980930e-01 -1.04779565e+00 -3.93696487e-01 8.75184476e-01 -1.74595013e-01 6.57954454e-01 -8.33298545e-03 7.43252933e-01 1.14478242e+00 1.79433897e-02 8.92096043e-01 1.51790369e+00 -3.25010538e-01 8.46752107e-01 1.67391002e-01 -2.95258731e-01 4.76105839e-01 1.55844182e-01 5.48869789e-01 -3.82193178e-01 -2.15507701e-01 6.19260013e-01 -3.21923107e-01 -1.32306531e-01 -6.68930590e-01 -8.49850953e-01 6.02015734e-01 2.51969248e-01 9.41227078e-02 -6.68689311e-01 2.96062946e-01 4.53959227e-01 3.88977557e-01 2.11680353e-01 6.46472454e-01 -5.46830237e-01 -5.93951762e-01 -5.95296860e-01 5.92245698e-01 4.71815914e-01 4.92150009e-01 8.32330287e-01 2.35461563e-01 -4.72563028e-01 7.97737837e-01 2.10113794e-01 4.62238252e-01 6.36547923e-01 -7.90471911e-01 5.60364366e-01 6.30843937e-01 5.64481616e-01 -6.49658680e-01 -1.15154646e-01 -4.71837103e-01 -2.66410500e-01 6.16726577e-01 4.89012688e-01 -4.60807621e-01 -6.78046584e-01 2.14242411e+00 6.99465752e-01 3.26421261e-01 1.45714015e-01 8.80473673e-01 5.08770309e-02 4.58613604e-01 2.13483840e-01 -5.81266344e-01 6.58941865e-01 -7.36354828e-01 -6.82262659e-01 -2.75000095e-01 3.48813325e-01 -2.52680182e-01 1.25880694e+00 5.06817520e-01 -1.06346142e+00 -4.36998218e-01 -1.09314704e+00 6.25621200e-01 -1.33230045e-01 1.32606655e-01 4.25600171e-01 3.76773357e-01 -6.85097396e-01 8.42186093e-01 -8.99009466e-01 3.71585526e-02 3.57642859e-01 4.38696980e-01 -4.60527055e-02 3.01126927e-01 -1.01802218e+00 7.99382627e-01 5.59650958e-01 -9.61663350e-02 -1.50536954e+00 -6.04378760e-01 -4.28108811e-01 -2.08805921e-03 1.10764527e+00 -9.74579677e-02 1.55715668e+00 -1.47884929e+00 -2.12261438e+00 4.82701540e-01 2.50176132e-01 -8.16616774e-01 8.22645366e-01 -3.46964210e-01 -1.35529697e-01 -8.90559852e-02 -1.62249252e-01 3.86647612e-01 1.21891308e+00 -1.46845138e+00 -8.61967325e-01 -3.86081308e-01 3.07681829e-01 4.67483759e-01 -3.16429347e-01 -2.84443557e-01 -2.47190237e-01 -5.12451351e-01 -5.04204154e-01 -8.51914883e-01 -4.66128230e-01 -2.85367399e-01 -1.39650881e-01 -3.10393691e-01 8.27838659e-01 -2.76757240e-01 1.39491069e+00 -1.91366446e+00 3.60241473e-01 2.41822168e-01 -2.19038665e-01 4.66227293e-01 -1.93349510e-01 3.46310198e-01 2.65424132e-01 -2.05593154e-01 -2.19418883e-01 -2.33359948e-01 -1.98150799e-02 4.01721179e-01 -4.25269336e-01 5.28389335e-01 -5.23782596e-02 5.55467904e-01 -1.17747045e+00 -2.52254009e-01 1.46895170e-01 -1.45206362e-01 -6.56469941e-01 7.17556000e-01 -8.71533930e-01 8.73011470e-01 -7.98228383e-01 4.05341148e-01 4.58409935e-01 1.62054494e-01 5.31364739e-01 2.25909293e-01 -1.74485728e-01 -4.31640297e-02 -1.21941900e+00 1.54399097e+00 -4.47849452e-01 -5.34968786e-02 1.35640562e-01 -1.12400687e+00 9.94633317e-01 5.61570972e-02 6.90697312e-01 -8.43732417e-01 3.27319771e-01 1.84232071e-01 5.14990538e-02 -5.26860237e-01 2.67017663e-01 6.13874756e-02 -2.43108661e-04 5.39004803e-01 -1.22173943e-01 -1.80746671e-02 2.40063563e-01 -3.28079611e-01 1.13586581e+00 5.18615961e-01 5.03460050e-01 -1.07901596e-01 8.25991809e-01 -1.33251384e-01 7.47281671e-01 1.00729358e+00 -4.50238645e-01 -1.93157420e-01 7.08105445e-01 -2.22498611e-01 -7.75104284e-01 -8.72252226e-01 3.11447322e-01 1.06426013e+00 1.28886119e-01 -1.17874695e-02 -8.05001915e-01 -1.17656457e+00 2.49616176e-01 8.62749875e-01 -7.07740724e-01 -3.60272884e-01 -6.63035691e-01 -3.52281153e-01 2.71098435e-01 3.33914489e-01 5.61499774e-01 -1.33394015e+00 -1.01867211e+00 4.48004544e-01 2.00564727e-01 -8.28823268e-01 -3.06998700e-01 3.85903746e-01 -8.37995470e-01 -1.07109201e+00 -4.74034935e-01 -2.75885403e-01 7.86130667e-01 -1.59553066e-01 8.64168406e-01 1.86889097e-02 -1.28007829e-02 4.06628579e-01 -4.12263930e-01 -3.68996233e-01 -5.35188735e-01 4.16017659e-02 1.95342034e-01 2.22676575e-01 6.34041950e-02 -5.32982409e-01 -5.84982276e-01 1.24516025e-01 -8.93775463e-01 -3.23192239e-01 6.82790518e-01 8.90643835e-01 8.52417469e-01 2.58820176e-01 8.95010591e-01 -7.75471151e-01 9.45871890e-01 -4.85253930e-01 -1.23166275e+00 3.77684861e-01 -7.55925119e-01 5.73130906e-01 1.12137008e+00 -8.17055464e-01 -1.06488669e+00 1.16326533e-01 5.30567691e-02 -6.66635990e-01 -7.68353194e-02 2.81723022e-01 -1.62272617e-01 -2.44222227e-02 4.14395750e-01 4.67891484e-01 1.51744485e-01 -4.12803769e-01 3.19231331e-01 4.86113578e-01 2.08812460e-01 -1.19741523e+00 5.87900281e-01 2.07005635e-01 1.79132714e-03 -3.03720504e-01 -9.03513730e-01 -1.37402043e-01 -3.05013388e-01 -5.07835627e-01 6.11402810e-01 -5.04661143e-01 -1.06171477e+00 4.03794169e-01 -5.11588633e-01 -9.10109341e-01 -6.03625119e-01 4.20571715e-01 -9.67690766e-01 9.41070616e-02 2.14655607e-04 -1.16340840e+00 -8.21218342e-02 -1.27648747e+00 5.76090932e-01 4.37105030e-01 2.34969556e-01 -8.04705858e-01 3.69910866e-01 -2.55732867e-03 4.36660200e-01 3.69577050e-01 9.56352651e-01 -6.14992976e-01 -3.81217271e-01 6.27795830e-02 1.89406529e-01 4.47087228e-01 1.94426239e-01 -2.25874349e-01 -5.75249732e-01 -7.53000617e-01 -6.84509948e-02 -8.59803259e-01 4.11385596e-01 8.54624361e-02 1.61679459e+00 -5.23791969e-01 -2.43640356e-02 3.43659043e-01 1.50932288e+00 5.97716808e-01 3.84504318e-01 6.02613389e-01 8.62478092e-02 3.58098358e-01 1.28222442e+00 1.04970002e+00 8.47582966e-02 7.17621624e-01 6.92252219e-01 3.28336239e-01 3.63648832e-01 -6.51938319e-01 4.97660160e-01 2.60569215e-01 8.09748620e-02 -1.01289511e-01 -5.86776495e-01 2.33577684e-01 -2.18029261e+00 -9.05429304e-01 7.57944763e-01 2.66188931e+00 1.12326860e+00 3.34351271e-01 3.11404169e-01 -2.86134742e-02 4.91174042e-01 1.38397843e-01 -1.24393880e+00 -4.25048262e-01 3.36020350e-01 1.50413409e-01 5.79103708e-01 4.59092021e-01 -9.65042770e-01 1.09128034e+00 6.12449789e+00 1.11317158e+00 -1.34292650e+00 -1.20283671e-01 6.36187494e-01 -1.74594447e-01 -2.49255806e-01 -4.30943407e-02 -8.23793411e-01 5.19247472e-01 8.91946375e-01 -2.67793477e-01 9.89145815e-01 1.01479483e+00 6.41147316e-01 -4.25243437e-01 -1.15466058e+00 7.18015015e-01 -2.18011051e-01 -9.22273219e-01 -9.17211697e-02 -2.68575214e-02 6.87053323e-01 -3.01264644e-01 8.43359381e-02 7.44855285e-01 5.66184223e-01 -7.86604702e-01 8.67814422e-01 5.97203553e-01 6.12607539e-01 -1.09453917e+00 4.20305938e-01 7.41540432e-01 -9.34857786e-01 -5.42421699e-01 -2.73489952e-01 2.03835756e-01 -2.32880861e-01 1.29970223e-01 -7.77969599e-01 3.44323844e-01 4.25846457e-01 4.49932963e-01 -3.71192634e-01 7.94995964e-01 -4.27364439e-01 6.52729690e-01 -8.55535939e-02 -4.18227583e-01 4.92789268e-01 -4.69576627e-01 6.65074468e-01 6.11469626e-01 9.49463546e-02 -1.07485361e-01 6.04984105e-01 7.52314150e-01 6.57353699e-02 1.78870365e-01 -4.22148138e-01 -3.38422477e-01 5.65048575e-01 1.08938897e+00 -5.09899557e-01 -2.15304360e-01 7.81223699e-02 7.25184143e-01 9.21408117e-01 2.29078338e-01 -1.11951768e+00 -3.08255777e-02 5.99240482e-01 -1.12663507e-01 3.50479215e-01 -1.89326689e-01 1.88509032e-01 -8.46693099e-01 -1.36431247e-01 -1.26459122e+00 4.06453371e-01 -2.21467257e-01 -1.14117014e+00 3.19524825e-01 1.44964948e-01 -1.41042721e+00 -2.13985011e-01 -2.63846457e-01 -3.02006543e-01 3.62427443e-01 -1.63266802e+00 -5.89024544e-01 -5.17126918e-02 5.86281598e-01 6.59039557e-01 -3.01436663e-01 5.65670192e-01 5.24090901e-02 -6.05225205e-01 6.60088599e-01 3.20704252e-01 -3.83820117e-01 5.96192062e-01 -1.35753417e+00 -3.77185762e-01 5.22631288e-01 -2.60422856e-01 2.32903585e-01 7.89373279e-01 -5.60413599e-01 -1.68862915e+00 -9.31816161e-01 -1.11303434e-01 1.57418117e-01 6.76477909e-01 -1.87753499e-01 -7.45256007e-01 3.91748458e-01 -3.49479541e-02 -4.95196283e-02 4.00692225e-01 -1.79430932e-01 2.85085235e-02 -5.52237153e-01 -1.42790079e+00 7.78077066e-01 8.18471134e-01 2.33495161e-02 -3.62696856e-01 -1.83623601e-02 6.50947154e-01 -5.34587204e-01 -6.95860386e-01 4.20184165e-01 3.77754211e-01 -8.93134594e-01 6.73802137e-01 -9.24000084e-01 2.03812897e-01 -1.84881330e-01 -9.26770736e-03 -1.57867324e+00 1.53476289e-02 -9.05205846e-01 -4.69251335e-01 1.11491489e+00 7.99028948e-02 -7.06407487e-01 8.74277115e-01 2.51786619e-01 2.44872585e-01 -1.16225362e+00 -9.24479485e-01 -8.25946808e-01 1.29824001e-02 -1.80349082e-01 8.00981343e-01 5.78727663e-01 -2.39188537e-01 -9.55631360e-02 -5.47358036e-01 2.19047479e-02 7.42432475e-01 1.60670340e-01 9.80058670e-01 -6.72942102e-01 -7.25829780e-01 -4.69602138e-01 4.89358529e-02 -1.25684738e+00 4.43615139e-01 -5.18740118e-01 3.66007626e-01 -1.06894767e+00 -3.90080884e-02 -8.17543209e-01 -6.07312381e-01 4.48471725e-01 -2.33588040e-01 -5.59109211e-01 1.60495505e-01 4.17735353e-02 -8.47493768e-01 1.00544024e+00 1.52216125e+00 3.53408419e-02 -6.35575354e-01 2.66733706e-01 -4.81496781e-01 5.95265329e-01 9.61268723e-01 -6.80525005e-01 -7.92889178e-01 -1.38047501e-01 1.04949847e-01 5.15621722e-01 -5.34451157e-02 -9.97299731e-01 1.85100343e-02 -7.81234443e-01 1.04953043e-01 -4.32096809e-01 1.30481467e-01 -9.66115713e-01 -1.33543968e-01 7.69262791e-01 -7.65113950e-01 1.05084851e-01 1.49946839e-01 9.95167613e-01 6.58335257e-03 -3.95708233e-01 1.05441773e+00 -2.62036711e-01 -6.51436865e-01 3.74179959e-01 -4.74808156e-01 2.80691594e-01 1.46436548e+00 2.20243961e-01 -1.53900892e-01 -3.20449501e-01 -4.68216270e-01 5.56785226e-01 2.77815700e-01 3.74412775e-01 5.65170467e-01 -1.16148138e+00 -3.76350373e-01 1.17112689e-01 -1.40505834e-02 -2.58501053e-01 1.16041183e-01 5.55217028e-01 -1.82320818e-01 1.17745630e-01 -2.03249007e-01 -3.47621560e-01 -1.09111762e+00 7.23718882e-01 4.82453763e-01 -7.87351668e-01 -4.13756341e-01 3.67507339e-01 -4.00971144e-01 -4.39879298e-01 5.33184409e-01 -2.42300615e-01 -3.88154328e-01 -1.17696352e-01 3.96562248e-01 4.18844700e-01 -2.01987579e-01 -9.42351595e-02 -1.58833876e-01 2.16244027e-01 2.14685555e-04 -3.86631906e-01 1.20920885e+00 1.34129161e-02 1.80881977e-01 4.99433607e-01 7.77122378e-01 9.63873193e-02 -1.82766521e+00 -1.73096687e-01 1.60014048e-01 -6.78497553e-01 4.43432555e-02 -9.74763036e-01 -9.34336066e-01 3.03581834e-01 7.77494073e-01 2.15010211e-01 1.13336134e+00 -4.20097381e-01 5.85290790e-01 6.12841010e-01 6.79846466e-01 -1.64589989e+00 5.01835823e-01 4.25246358e-01 8.80968213e-01 -1.21121681e+00 -5.20806946e-03 3.35711896e-01 -8.18243682e-01 9.49115336e-01 1.03620780e+00 -4.08303678e-01 4.80424881e-01 1.94893137e-01 -4.88954261e-02 2.28278443e-01 -1.00865793e+00 -3.11174065e-01 -4.60452400e-02 4.01872754e-01 -7.90559649e-02 4.34757955e-02 -5.27404547e-01 3.21189702e-01 2.31936276e-01 2.91393287e-02 2.15458453e-01 1.20687604e+00 -6.73314810e-01 -1.46358263e+00 -2.60687858e-01 3.36019039e-01 -4.22322094e-01 5.43162704e-01 -2.00285107e-01 7.13686049e-01 8.68723691e-02 7.92513430e-01 -1.09259769e-01 -1.40970692e-01 3.12314570e-01 -2.77949005e-01 6.45904541e-01 -4.35500115e-01 -7.09090352e-01 9.17199925e-02 2.88863629e-02 -8.60113978e-01 -3.90542239e-01 -6.00935042e-01 -1.43316197e+00 1.09921776e-01 -1.53380245e-01 3.31017256e-01 5.73810756e-01 1.01967323e+00 1.92379117e-01 5.76986730e-01 1.10242677e+00 -6.18806303e-01 -1.43143964e+00 -6.36358440e-01 -5.18369734e-01 4.12286878e-01 3.51019442e-01 -9.84250486e-01 -3.92741114e-01 -4.87828791e-01]
[4.136989116668701, 2.243880271911621]
d5ade06b-3fb8-4cb3-8f0b-4866a0e1d15c
combining-domain-specific-meta-learners-in
2011.00179
null
https://arxiv.org/abs/2011.00179v1
https://arxiv.org/pdf/2011.00179v1.pdf
Combining Domain-Specific Meta-Learners in the Parameter Space for Cross-Domain Few-Shot Classification
The goal of few-shot classification is to learn a model that can classify novel classes using only a few training examples. Despite the promising results shown by existing meta-learning algorithms in solving the few-shot classification problem, there still remains an important challenge: how to generalize to unseen domains while meta-learning on multiple seen domains? In this paper, we propose an optimization-based meta-learning method, called Combining Domain-Specific Meta-Learners (CosML), that addresses the cross-domain few-shot classification problem. CosML first trains a set of meta-learners, one for each training domain, to learn prior knowledge (i.e., meta-parameters) specific to each domain. The domain-specific meta-learners are then combined in the \emph{parameter space}, by taking a weighted average of their meta-parameters, which is used as the initialization parameters of a task network that is quickly adapted to novel few-shot classification tasks in an unseen domain. Our experiments show that CosML outperforms a range of state-of-the-art methods and achieves strong cross-domain generalization ability.
['Martin Ester', 'Weilian Song', 'Shuman Peng']
2020-10-31
null
null
null
null
['cross-domain-few-shot']
['computer-vision']
[ 0.2649662 -0.22927329 -0.34258425 -0.6069214 -0.9491261 -0.14793411 0.606577 0.16013847 -0.48739862 0.7868103 -0.04563992 0.36543408 -0.20595272 -0.84173816 -0.5315592 -0.5546582 0.18737268 0.6448753 0.7286123 -0.40124276 0.17445986 0.03473043 -1.8661116 0.51267886 1.0734807 0.8986523 0.41100976 0.49322724 -0.27701622 0.7056975 -0.7748822 -0.2130924 -0.09623282 -0.68515694 -0.74704325 0.06427025 0.2613447 -0.17732893 -0.20451431 0.9586049 0.5459979 1.1789172 0.9767866 -1.2570058 -0.78164446 0.351746 -0.29760516 0.6528966 -0.08082735 0.09335273 0.6239473 -0.918922 0.70281863 1.0344448 0.45018524 1.0666298 -1.0546362 -0.6666407 0.21844031 0.5867831 -1.1720029 -0.45707172 0.87893903 -0.27410302 1.1137028 -0.31385285 0.37955728 1.3619127 0.12711519 0.7553273 0.7332988 -0.43254575 0.92369026 0.37850213 0.39469218 0.2606884 0.12779863 0.05404966 -0.55571043 -0.30748186 0.19724774 0.35827252 0.0389531 -0.6726179 -0.87513167 1.0121038 0.33142018 0.4980214 -0.3451272 -0.23924525 0.68233323 0.46068683 0.8896774 0.6290229 -0.46363285 -0.05402556 -0.68591756 0.24615364 0.6193185 0.96594465 0.9895578 0.128934 -0.11866039 1.4693109 -0.13511114 0.12008458 1.1228008 -0.58125734 0.5180885 0.56947774 0.04285065 -0.3658418 -0.3128185 -0.497518 -0.5729896 -0.01465307 -0.20072515 -0.39012167 -1.092624 1.7379198 0.5708909 0.78940356 0.34701067 0.6853721 0.9891974 0.87535584 0.34410477 -0.1647623 0.9142519 -1.3313036 -0.30176067 -0.6774888 0.6078038 -0.17266874 0.92289674 0.06737193 -0.79968005 -0.92001116 -1.3121887 0.27334782 -0.8749326 -0.54424095 0.15374038 0.39313507 -0.4267196 0.6981477 -0.4115456 -0.6386295 0.54782164 -0.05837315 -0.14270245 -0.38737017 -1.395914 1.134013 1.0129048 -0.7353494 -1.2384096 -0.99718165 -1.055906 0.19806926 0.54131377 -0.76070803 1.5003657 -0.91085505 -1.4322102 0.7631397 -0.0079149 -0.4289185 0.11693779 0.01460832 -0.877813 0.05603423 0.12709343 0.5801845 1.1568928 -1.2059597 -1.0921931 -0.31182694 -0.01865074 0.33686176 -0.6584256 -0.3458747 -0.21273513 -0.6187002 -0.2994135 -0.63854796 -0.22048336 -0.35576054 0.2170558 -0.4335036 0.9631784 -0.01323738 1.2477303 -2.1498508 0.35370392 -0.33106428 -0.01761533 0.6626428 -0.53394717 0.3999208 -0.14802256 -0.36254436 -0.21664675 -0.358134 -0.20937653 0.21295793 -0.4164822 0.13626574 0.17950042 0.88073343 -1.367714 -0.31910583 0.4830481 0.19567347 -0.18673931 0.40474582 -0.50468946 0.22117737 -0.48105818 0.4541679 0.59934866 -0.33030695 0.06412698 0.24947253 0.23064508 -0.1829982 -1.0995523 1.961676 -0.7403032 0.28542924 -0.54295665 -1.4469397 1.0756433 0.153263 0.27462566 -0.7387243 0.30434927 0.24072911 -0.24116921 -0.56964797 0.26754597 -0.56018025 -0.21462233 0.6865321 0.87108123 -0.09797355 0.2598026 -0.10444064 0.9335586 -0.1348432 0.7346624 0.01497945 0.44012088 0.14007735 0.5999692 0.93830585 -0.42993316 0.5966874 -0.14482206 -0.69391423 -1.1072266 -1.0691712 -0.03994144 1.8376958 0.30013236 -0.19163503 -0.5451552 -0.8662849 0.14022554 1.182282 -0.8560827 -0.88439614 -0.23639917 -0.76346725 0.0181736 0.58756447 0.573232 -1.1012887 -0.9017111 0.56980425 0.16062279 -0.8273673 -0.05005216 0.45641088 -1.1020833 -0.9426786 -1.1266336 -0.94543445 0.40930095 0.66002864 0.92549706 -0.330326 -0.3683208 0.34762505 -0.6290854 -0.5797764 -0.31515682 0.22870182 0.14447376 0.04758784 0.80411035 -0.70009494 -0.29728422 0.3863176 -0.9093773 -0.33199403 0.31946072 1.2679017 0.5443415 0.1922651 1.023341 -1.1467805 0.70039195 -1.0011988 -0.27618408 0.59257853 -0.37413517 -0.07067433 0.92374647 -0.9131228 -1.3182757 -0.21874027 0.32875478 -0.61046 -0.2610928 0.4857439 -0.13548537 0.05502869 1.1392385 0.42601386 -0.23229891 -0.5304542 0.5762835 0.74602836 0.37271714 -0.43562394 0.6781007 0.31983408 -0.5060647 -1.0142517 -1.5096315 -0.8420835 -0.71059644 -0.14258055 0.7086163 -0.8926887 0.18299739 0.53475165 -0.81512624 -0.41621393 -0.5985796 0.31225613 -0.7820674 -0.03051288 0.01720604 -0.5700135 -0.17998332 -0.6799458 0.69908917 0.55984944 -0.16467313 -1.3393247 0.5145286 0.18819481 0.4526478 0.10824019 1.0335164 -1.1086215 0.09173056 -0.32704714 0.01942663 0.4504996 0.12714727 -0.5561462 -1.2403388 -0.454845 0.27277455 -0.7712717 1.1202091 0.3321499 1.230794 0.05704531 -0.572782 0.5515418 1.4246427 0.4390174 0.30885732 0.49839526 0.13051325 0.24702501 0.8546795 0.6282561 0.19040684 0.48062572 0.04940702 0.4599301 -0.10312628 -0.14162181 -0.0527699 0.51479536 0.12851886 -0.14597507 -0.8184983 0.6492737 -2.1626432 -1.2587663 0.81462896 2.0541773 0.5464772 0.30071354 0.23917408 -0.22212206 0.8672563 0.44789767 -1.2149706 -0.29997233 0.17343597 0.3078454 -0.14121656 0.07600603 -1.3240229 0.97428995 5.9992414 0.9275485 -0.9378473 0.41857228 0.1730795 -0.18560252 0.12348292 -0.18522207 -0.8684111 0.46901846 1.0397966 -0.8155642 0.25957042 1.3736793 -0.42795092 0.05895224 -1.102511 0.8779245 0.5247658 -1.3963777 0.26109508 -0.43637154 1.2639378 0.15751065 0.17684266 1.1395682 0.31421393 -0.5248881 0.26803473 0.48032632 0.5151773 -0.9217244 0.5169444 0.80936784 -0.9331532 -0.6674055 -1.0609789 0.0961583 0.01460282 0.31233063 -0.7626188 0.35559294 0.5275179 0.9668875 -0.46520007 1.2969787 -0.07582521 0.22905429 0.18481345 -0.19383684 0.29251996 0.24329466 0.48687187 0.9645066 0.3387629 0.40911862 0.2357681 0.64392304 -0.18087462 -0.15952404 -0.63523495 -0.0125264 0.725634 1.0963002 -0.34606203 -0.8195749 -0.51236963 1.0264591 0.68594646 0.40760604 -0.6499791 -0.650908 0.64949036 -0.12961552 0.5873647 0.17389315 0.02233135 -1.4418876 -0.28453916 -0.6875764 0.9026223 -0.6353461 -1.7432401 0.5967933 0.29917076 -1.6394091 -0.5300566 -0.40894836 -1.0063897 0.63768107 -1.6210874 -0.9904244 -0.51792413 0.65472317 1.17745 -0.7690499 1.042881 -0.01878731 -0.44154954 0.5841964 0.576289 -0.16221432 0.89373565 -0.9558369 0.53516597 0.40959546 -0.07356188 0.50520253 0.6031381 -0.6056267 -1.0574329 -1.3748599 0.5802525 -0.37380126 0.6173828 -0.21748173 -1.3086439 0.58251846 -0.11773149 0.28124014 0.9966567 0.26062888 -0.6380641 -0.16473275 -1.2845263 0.39475524 1.0063969 -0.4272277 -1.3036706 0.37716272 0.73698026 -0.1270596 -0.70203465 0.2468881 0.3512506 -0.8073879 1.0305306 -1.3280491 0.44755003 0.08213197 -0.23894268 -2.1468499 -0.6549948 -0.01332689 -0.43117997 1.0166916 0.26403183 -0.5395931 0.70945984 0.4068664 -0.2193848 -0.4847976 -1.0178069 -1.3569925 0.24432641 -0.1871741 0.59371185 1.1780059 0.2689977 0.6939011 -0.36212972 -0.2317319 0.8803435 0.32052952 0.7100101 -1.5982084 -0.24736334 -0.22104684 -0.38509655 -0.5327101 0.4938586 -0.93009526 0.01693179 -1.4384711 0.45413548 -0.09061272 -0.7494037 0.37532818 -0.39476177 -0.19793248 0.20211492 0.08516084 -1.1369393 0.82072055 1.0876907 -0.40730318 -0.49720153 0.19649999 -0.6104252 0.7229406 0.6926183 -0.6759179 -0.7842184 -0.38585222 -0.41880447 -0.02499596 0.12218115 -1.4697481 0.47920993 -0.44411054 0.6223117 -0.40998802 0.6360784 -0.60588115 -0.24589571 0.3542816 -0.43021628 -0.5241234 0.16458252 0.82765913 -0.20233682 -0.7480764 1.2659411 -0.50083244 -1.5702782 0.5059874 -0.1330462 0.5546604 1.4245772 -0.26924598 -0.57402927 -0.1755765 -0.9519587 0.35650334 0.37338647 0.84489834 0.79744065 -1.4614388 -0.6471557 0.1029132 0.89263266 -0.19528955 0.8236202 0.02331259 0.21922156 0.07489409 -0.5435846 -0.43038973 -0.82249546 0.970392 0.37027562 -0.03256926 -0.6325272 0.9554126 0.01436325 -0.56323653 0.21049422 0.13409579 -0.23175988 0.44401598 1.0514356 0.6234364 0.01475258 -0.26117122 -0.25716493 0.40308702 -0.4599378 0.02362704 1.7084104 0.0328494 0.5077436 0.888525 1.3915173 -1.0958447 -1.324534 -0.72668314 0.04509245 -0.46341097 -0.11066238 -0.9353584 -0.55161166 1.081116 0.7033917 -0.07638709 0.8896136 0.07136784 0.8528745 0.77348995 0.5782987 -1.7063192 0.5752786 0.83025503 0.55035895 -1.4640266 -0.05294111 0.16949384 -0.8187309 1.0569191 0.9839832 -0.28402233 0.826901 -0.3976993 -0.17790815 0.01150966 -1.0649388 -0.19182576 0.2899903 1.0484118 -0.06814156 -0.09826894 0.16378519 0.8338007 0.31682062 0.3769132 0.26945928 1.2952576 -1.1989006 -0.902439 -0.0118501 0.5121156 0.27923027 0.28580403 -0.14942339 0.50957084 0.24110855 0.8746444 0.08251946 -0.48994413 0.5791357 0.5479776 0.41583458 -1.3129938 -0.33325085 -0.42141232 -0.19957867 -0.17171048 -0.14542691 -0.43444824 -0.8807213 0.04478073 -0.24967837 0.16990466 0.40203717 1.1901606 0.36716607 0.6418169 0.7611871 -0.8776704 -1.0468929 -0.99561584 -0.7972935 0.52107674 0.26747537 -1.1124372 -0.40504098 -0.1848855 ]
[10.048171997070312, 3.1461851596832275]
89b55108-6f75-4a57-9cfe-ab622569c055
ciff-net-contextual-image-feature-fusion-for
2303.03672
null
https://arxiv.org/abs/2303.03672v1
https://arxiv.org/pdf/2303.03672v1.pdf
CIFF-Net: Contextual Image Feature Fusion for Melanoma Diagnosis
Melanoma is considered to be the deadliest variant of skin cancer causing around 75\% of total skin cancer deaths. To diagnose Melanoma, clinicians assess and compare multiple skin lesions of the same patient concurrently to gather contextual information regarding the patterns, and abnormality of the skin. So far this concurrent multi-image comparative method has not been explored by existing deep learning-based schemes. In this paper, based on contextual image feature fusion (CIFF), a deep neural network (CIFF-Net) is proposed, which integrates patient-level contextual information into the traditional approaches for improved Melanoma diagnosis by concurrent multi-image comparative method. The proposed multi-kernel self attention (MKSA) module offers better generalization of the extracted features by introducing multi-kernel operations in the self attention mechanisms. To utilize both self attention and contextual feature-wise attention, an attention guided module named contextual feature fusion (CFF) is proposed that integrates extracted features from different contextual images into a single feature vector. Finally, in comparative contextual feature fusion (CCFF) module, primary and contextual features are compared concurrently to generate comparative features. Significant improvement in performance has been achieved on the ISIC-2020 dataset over the traditional approaches that validate the effectiveness of the proposed contextual learning scheme.
['Shaikh Anowarul Fattah', 'Tanvir Mahmud', 'Bishmoy Paul', 'Md Awsafur Rahman']
2023-03-07
null
null
null
null
['melanoma-diagnosis', 'skin-cancer-classification']
['computer-vision', 'medical']
[ 6.64720535e-01 -2.52240151e-01 -1.81246057e-01 -2.13966221e-01 -1.19057488e+00 6.69204369e-02 7.47547328e-01 5.30547261e-01 -5.94648838e-01 6.26866043e-01 4.50088829e-01 6.08892441e-02 -4.74538863e-01 -4.91674900e-01 -1.27091765e-01 -1.04965448e+00 3.51683259e-01 -3.61337990e-01 1.45578161e-01 -1.49097115e-01 4.52170312e-01 4.58620608e-01 -1.46382320e+00 6.58386230e-01 1.13612795e+00 1.00305879e+00 4.73258525e-01 9.84593570e-01 -1.56954169e-01 7.14850247e-01 -5.07278204e-01 -2.04225630e-01 -6.58583790e-02 -6.21144474e-01 -9.12268579e-01 1.32658318e-01 4.46496218e-01 -1.24715470e-01 2.85939071e-02 1.03632963e+00 7.06192136e-01 7.49160443e-03 7.46941924e-01 -9.10749733e-01 -5.21169364e-01 7.96002075e-02 -8.31467628e-01 5.10398805e-01 3.87714177e-01 2.30446160e-01 4.80748177e-01 -7.45185435e-01 4.65758324e-01 8.98075640e-01 7.45687366e-01 5.59985638e-01 -6.27859294e-01 -2.00668931e-01 1.15890786e-01 5.68001986e-01 -1.29273808e+00 -6.58457503e-02 6.79293334e-01 -1.29787609e-01 8.26633036e-01 6.37753665e-01 7.00342953e-01 8.39249372e-01 7.41759956e-01 7.70033479e-01 1.41796732e+00 -7.63496339e-01 -4.40480784e-02 2.00619102e-01 2.29757726e-01 7.12789476e-01 -1.59385130e-01 -3.89867313e-02 -4.24253821e-01 7.31047988e-02 4.77515787e-01 4.20474410e-01 -1.76876217e-01 4.49055821e-01 -9.77080107e-01 6.23158336e-01 7.96569169e-01 5.67825317e-01 -6.50565147e-01 2.21229523e-01 4.59165335e-01 -1.66111849e-02 3.93697292e-01 5.41838296e-02 -1.87320471e-01 2.44036630e-01 -6.66909277e-01 -1.12949207e-01 -3.94858643e-02 2.63906658e-01 5.96753836e-01 -2.17501119e-01 -5.02348065e-01 6.98747039e-01 1.44551292e-01 3.04257303e-01 8.68674576e-01 -4.17612582e-01 -5.35158161e-03 1.04436195e+00 -3.88313144e-01 -9.56234574e-01 -5.11139393e-01 -7.92751610e-01 -1.05663049e+00 1.12568982e-01 -1.01477489e-01 -2.02416465e-01 -1.29643059e+00 1.42022717e+00 4.71867442e-01 3.92891109e-01 2.15967789e-01 8.27809989e-01 8.66528094e-01 3.09079230e-01 4.89191294e-01 3.93868387e-02 1.42779231e+00 -1.04068029e+00 -6.39903963e-01 1.40648827e-01 5.77014744e-01 -7.57307291e-01 7.23909974e-01 1.15081668e-01 -7.25538552e-01 -6.26333356e-01 -1.02532387e+00 6.04457632e-02 -6.40071571e-01 5.06677963e-02 6.52307928e-01 4.25043762e-01 -1.10173619e+00 2.37046182e-01 -6.18901491e-01 -8.56670082e-01 5.72954059e-01 4.16551650e-01 -6.47459030e-01 -1.86481789e-01 -9.50806916e-01 9.15729880e-01 2.51814574e-01 1.54180929e-01 -8.65255177e-01 -7.56802261e-01 -7.66963243e-01 -2.95209914e-01 1.73443958e-01 -8.07801306e-01 7.97887862e-01 -1.37592196e+00 -1.11914611e+00 4.80878115e-01 -3.95190030e-01 -3.58714938e-01 7.32816830e-02 -1.00362316e-01 -5.56281388e-01 4.74685609e-01 1.48525769e-02 6.97209179e-01 8.05730760e-01 -1.10343397e+00 -1.03301930e+00 -5.40094674e-01 -5.97386137e-02 7.29947805e-01 -3.02807301e-01 2.66222209e-02 -4.60517108e-01 -5.72251022e-01 -3.20580184e-01 -6.26099229e-01 -5.12026072e-01 -3.21379863e-02 -4.24853772e-01 -1.00468114e-01 1.00992393e+00 -8.47541749e-01 1.31223035e+00 -2.05597043e+00 2.04935875e-02 1.89832106e-01 1.93267372e-02 4.41282928e-01 -2.72480875e-01 4.08588558e-01 -1.81898445e-01 1.12057351e-01 -2.09914237e-01 -1.95297569e-01 -5.26265204e-01 -8.24431330e-02 5.20188808e-01 3.67109776e-01 6.55898631e-01 1.03494787e+00 -7.16805875e-01 -5.53714931e-01 5.41920841e-01 7.92623222e-01 -2.01196149e-01 -4.00109701e-02 1.77685246e-01 4.86975849e-01 -3.24408978e-01 1.05630863e+00 6.25343680e-01 -2.42769912e-01 -2.63061523e-01 -5.02102315e-01 1.80788375e-02 -6.83559239e-01 -8.66969407e-01 1.78273630e+00 -5.15947759e-01 4.43306208e-01 -1.62060529e-01 -6.25822365e-01 4.60787147e-01 3.64532948e-01 4.86609817e-01 -8.61606956e-01 1.98633090e-01 -7.66763687e-02 1.64837584e-01 -7.67268419e-01 2.21343026e-01 -1.54220730e-01 1.03169091e-01 -8.31211172e-03 6.84513450e-02 4.04034227e-01 -4.09920141e-02 1.44008726e-01 1.23240864e+00 -1.32103935e-01 7.06408679e-01 -2.01960132e-01 1.06221461e+00 1.69860765e-01 5.07354975e-01 5.10294616e-01 -5.22449315e-01 4.52695817e-01 8.74106959e-02 -4.52756554e-01 -6.49049401e-01 -7.45908976e-01 -1.94855571e-01 7.40861654e-01 1.97643906e-01 -1.44910544e-01 -7.38262713e-01 -9.06869948e-01 1.51506746e-02 2.06973851e-01 -1.16116095e+00 -1.78973317e-01 -5.72514273e-02 -1.03825009e+00 2.68827677e-01 7.14022994e-01 8.22454035e-01 -9.95173991e-01 -7.52036691e-01 1.18415698e-01 2.49378115e-01 -5.21414697e-01 -3.39935482e-01 1.23723194e-01 -5.47489703e-01 -1.33715260e+00 -8.99217188e-01 -7.11716413e-01 7.86955893e-01 2.30350465e-01 4.02601212e-01 8.67665783e-02 -1.06045008e+00 3.94142687e-01 -4.64947015e-01 -4.78175849e-01 -9.08932388e-02 -1.66661859e-01 -2.92549193e-01 3.54941905e-01 4.01552647e-01 -7.45773017e-02 -8.38530183e-01 -1.77650452e-01 -1.01925945e+00 1.78549692e-01 1.18837929e+00 1.27280045e+00 7.27895677e-01 2.54925579e-01 7.44506121e-01 -8.61452937e-01 7.32070267e-01 -4.96725291e-01 8.77121910e-02 4.60974514e-01 -3.97533268e-01 -2.28287235e-01 4.63519245e-01 -9.68968421e-02 -1.31460023e+00 2.42770538e-02 -1.62245959e-01 -1.23437494e-01 -4.05413836e-01 7.36634731e-01 -2.05102433e-02 -3.32566857e-01 3.95061493e-01 3.91945511e-01 1.95526451e-01 -1.58375323e-01 6.34414181e-02 8.94166589e-01 5.80676436e-01 6.39426708e-02 2.37615705e-01 3.98173243e-01 1.88203394e-01 -7.44548380e-01 -7.81458557e-01 -7.82875180e-01 -5.26346087e-01 -3.21193278e-01 1.14977050e+00 -8.55600297e-01 -5.88861525e-01 8.70455503e-01 -7.81003833e-01 -3.49898590e-03 -5.79244606e-02 3.91472220e-01 -2.39561900e-01 2.92464942e-01 -4.87778425e-01 -1.03398573e+00 -6.44008756e-01 -1.20597422e+00 1.29892433e+00 8.79718363e-01 5.50779402e-02 -1.24865484e+00 -8.27204362e-02 4.56739843e-01 6.16122007e-01 5.90567946e-01 8.56995702e-01 -5.09723961e-01 -1.42675385e-01 -3.46609533e-01 -5.89854836e-01 2.95040965e-01 6.11914814e-01 3.22669968e-02 -1.00578582e+00 -2.72602558e-01 -2.79754519e-01 -3.56627069e-02 1.03470397e+00 5.23823202e-01 1.02006364e+00 6.57331478e-03 -4.91643995e-01 5.74813485e-01 2.13212037e+00 5.85784912e-01 5.16667306e-01 1.47949368e-01 7.65360355e-01 3.76149863e-01 6.25646770e-01 2.81339258e-01 2.27751687e-01 1.55599862e-01 5.65803230e-01 -5.58814585e-01 -4.12844718e-01 2.66668499e-01 3.62868886e-03 5.38107395e-01 -1.79884985e-01 -5.98982424e-02 -8.51622701e-01 7.94766545e-01 -1.60624635e+00 -9.14851725e-01 3.97409908e-02 1.92726266e+00 5.49211681e-01 -1.49531469e-01 -1.91370770e-01 2.42460445e-01 1.02776146e+00 -7.19476417e-02 -6.51044011e-01 -6.13525987e-01 -1.83451205e-01 3.57735097e-01 3.94105762e-01 4.82003450e-01 -1.31482196e+00 5.31119525e-01 5.77156019e+00 1.10398495e+00 -1.30393052e+00 2.53310680e-01 8.10520768e-01 1.06290586e-01 -1.04341283e-01 -1.68962866e-01 -6.59108996e-01 4.94270325e-01 7.85014212e-01 1.02770694e-01 -1.98186934e-01 4.98071581e-01 1.35205775e-01 -6.36283338e-01 -6.76710725e-01 9.04788256e-01 4.23709601e-01 -1.33943331e+00 2.81302631e-01 1.60882756e-01 7.47268677e-01 -3.16209614e-01 2.15235248e-01 -1.90046560e-02 -6.58456534e-02 -1.01127565e+00 -1.31617993e-01 1.37642288e+00 9.05902624e-01 -1.10805213e+00 1.38773739e+00 -9.58727449e-02 -1.15890265e+00 -3.47881764e-01 5.45540117e-02 3.48982215e-01 -1.49973050e-01 3.74020606e-01 -9.27243829e-01 1.15086901e+00 4.96427655e-01 7.87845075e-01 -1.13401294e+00 1.17013192e+00 3.54051799e-01 4.11636561e-01 4.75623459e-02 -8.90106261e-02 4.93054003e-01 3.04553598e-01 3.84294838e-01 1.16518915e+00 2.78500408e-01 -1.08701870e-01 -3.28712128e-02 2.80530483e-01 3.96084934e-01 3.45718503e-01 -3.03959340e-01 1.35455057e-01 2.16463301e-02 1.49662352e+00 -3.66649896e-01 -2.67363220e-01 -5.15138507e-01 1.22401869e+00 5.48750116e-03 1.20742202e-01 -5.96779764e-01 -6.21651113e-01 4.00301576e-01 -9.47469473e-02 8.65013748e-02 4.44773525e-01 -8.27238560e-02 -6.72553241e-01 -2.43517697e-01 -6.90312922e-01 6.78990424e-01 -7.32471108e-01 -1.41057062e+00 7.42160141e-01 -3.71722639e-01 -1.10465753e+00 4.38207239e-02 -5.34855306e-01 -1.06013274e+00 1.16807067e+00 -1.69780934e+00 -1.73252141e+00 -7.56535470e-01 9.08548176e-01 7.83672869e-01 -2.52553612e-01 1.00683796e+00 -2.36599836e-02 -8.64166617e-01 7.17640758e-01 -8.42410773e-02 -6.67403713e-02 7.87448525e-01 -1.44283020e+00 -2.95004487e-01 9.08812404e-01 -4.89195704e-01 4.85215724e-01 4.61069569e-02 -8.05884659e-01 -1.27449846e+00 -1.41500068e+00 4.44096506e-01 -4.82280441e-02 3.63402575e-01 4.91553664e-01 -5.87663472e-01 3.54258679e-02 6.85559452e-01 1.18758902e-01 1.15524793e+00 -2.02664554e-01 5.79322912e-02 -3.42015237e-01 -1.42026830e+00 4.59272981e-01 4.76186901e-01 -5.12671888e-01 -2.61899263e-01 1.18401460e-02 5.42460144e-01 1.19930292e-02 -1.12442482e+00 6.95861816e-01 5.29217958e-01 -9.02220368e-01 6.14817321e-01 -3.99089903e-01 4.46885586e-01 -4.37825799e-01 -4.24629092e-01 -1.23187041e+00 -4.12617892e-01 -3.58618833e-02 1.49684146e-01 1.19138193e+00 3.23573023e-01 -4.23225850e-01 6.41692340e-01 2.47346506e-01 -2.58187532e-01 -1.27308333e+00 -8.22690606e-01 -2.29753196e-01 -1.76908657e-01 -2.14559004e-01 4.66159642e-01 7.30095446e-01 -1.92414075e-01 4.87066768e-02 -1.20924719e-01 1.55808970e-01 5.01824915e-01 -2.21629605e-01 2.39287794e-01 -8.71367574e-01 -2.09983319e-01 -4.13249493e-01 -8.27034891e-01 1.22383490e-01 -3.82652313e-01 -8.89472604e-01 -2.42804900e-01 -1.79894209e+00 6.91797793e-01 -2.19908893e-01 -1.05113542e+00 4.12342995e-01 -7.47312725e-01 5.79770803e-01 1.72875345e-01 -2.22894683e-01 -6.03625178e-01 2.28419960e-01 1.22554719e+00 -3.65896493e-01 7.90392142e-03 -2.67645568e-01 -1.03745127e+00 5.13608456e-01 7.01701581e-01 7.72686973e-02 -3.15777332e-01 -1.65117145e-01 -3.78604800e-01 1.35369316e-01 5.13740778e-01 -1.50772202e+00 5.51550090e-01 -1.28220215e-01 9.67519701e-01 -8.92819047e-01 3.81415784e-01 -7.20242262e-01 1.76170412e-02 7.25335777e-01 -2.41975561e-01 3.00326273e-02 3.48252296e-01 7.72402704e-01 -4.52547401e-01 -8.96623731e-02 7.51388848e-01 -2.21531555e-01 -1.13788533e+00 3.89656663e-01 -3.95470142e-01 -5.47990918e-01 1.50103593e+00 -4.44932699e-01 -4.43800151e-01 1.90442905e-01 -8.32611084e-01 7.71819428e-02 1.61376417e-01 1.21483229e-01 8.28713596e-01 -1.40672028e+00 -7.78869152e-01 6.43035099e-02 4.06683505e-01 -6.07089475e-02 1.05580056e+00 1.14811361e+00 -4.61026102e-01 3.91532302e-01 -4.63494837e-01 -6.81900859e-01 -1.55863452e+00 5.66811085e-01 3.84052485e-01 -4.30895835e-01 -2.03678817e-01 1.02498543e+00 -1.13377320e-02 1.62398182e-02 -9.74346548e-02 -9.73712057e-02 -4.68713582e-01 6.65976331e-02 9.26406085e-01 2.38799617e-01 1.60950631e-01 -6.26593292e-01 -4.06416208e-01 8.23126614e-01 -5.24391115e-01 6.68382570e-02 1.09951353e+00 -1.01089470e-01 -7.17716068e-02 9.64931473e-02 1.22226286e+00 -7.01537654e-02 -1.15706563e+00 -1.54412672e-01 -1.96663931e-01 -3.42112958e-01 3.41149658e-01 -1.44756866e+00 -1.06295061e+00 7.60053754e-01 1.21424663e+00 -3.38038325e-01 1.84903049e+00 -3.18467915e-01 6.76888406e-01 -1.56322181e-01 -5.22248931e-02 -9.52679813e-01 1.00370377e-01 -8.93866792e-02 7.89804041e-01 -1.44600451e+00 -1.15754202e-01 -2.86102921e-01 -9.08772886e-01 9.95967090e-01 8.44989836e-01 -2.14170411e-01 6.45939112e-01 3.49214107e-01 2.43672922e-01 -1.12490796e-01 -9.44969654e-01 -5.72518408e-01 4.88396615e-01 7.32798576e-01 2.12650120e-01 7.67011866e-02 -4.21072781e-01 5.63449502e-01 5.94965041e-01 -2.40837168e-02 2.58362323e-01 1.08530676e+00 -4.09873009e-01 -1.12897718e+00 -2.62567043e-01 8.11769366e-01 -6.34542108e-01 -2.69986302e-01 -4.22987729e-01 8.09853315e-01 6.58211768e-01 8.39013875e-01 4.13757227e-02 -6.49989486e-01 -5.75081669e-02 3.42342742e-02 4.12729710e-01 -4.31477278e-01 -9.75255430e-01 2.21461207e-01 -1.56380117e-01 -4.50990796e-01 -7.11568594e-01 -5.24619102e-01 -1.07466161e+00 2.58418590e-01 -3.60174865e-01 -9.34375599e-02 8.82192910e-01 1.01954591e+00 3.64321142e-01 1.00142908e+00 7.77937531e-01 -4.89177227e-01 5.90000302e-02 -1.02295864e+00 -4.12277311e-01 3.44616652e-01 4.10961539e-01 -5.04437566e-01 -9.22194123e-02 -4.20427173e-02]
[15.610678672790527, -2.9315478801727295]
63cb838e-a5af-448f-9d45-602fe65adc78
pretrained-encyclopedia-weakly-supervised-1
1912.09637
null
https://arxiv.org/abs/1912.09637v1
https://arxiv.org/pdf/1912.09637v1.pdf
Pretrained Encyclopedia: Weakly Supervised Knowledge-Pretrained Language Model
Recent breakthroughs of pretrained language models have shown the effectiveness of self-supervised learning for a wide range of natural language processing (NLP) tasks. In addition to standard syntactic and semantic NLP tasks, pretrained models achieve strong improvements on tasks that involve real-world knowledge, suggesting that large-scale language modeling could be an implicit method to capture knowledge. In this work, we further investigate the extent to which pretrained models such as BERT capture knowledge using a zero-shot fact completion task. Moreover, we propose a simple yet effective weakly supervised pretraining objective, which explicitly forces the model to incorporate knowledge about real-world entities. Models trained with our new objective yield significant improvements on the fact completion task. When applied to downstream tasks, our model consistently outperforms BERT on four entity-related question answering datasets (i.e., WebQuestions, TriviaQA, SearchQA and Quasar-T) with an average 2.7 F1 improvements and a standard fine-grained entity typing dataset (i.e., FIGER) with 5.7 accuracy gains.
['William Yang Wang', 'Wenhan Xiong', 'Veselin Stoyanov', 'Jingfei Du']
2019-12-20
null
https://openreview.net/forum?id=BJlzm64tDH
https://openreview.net/pdf?id=BJlzm64tDH
iclr-2020-1
['triviaqa']
['miscellaneous']
[-2.74800271e-01 5.26930630e-01 -4.08250004e-01 -5.20176649e-01 -1.18872619e+00 -8.07966352e-01 7.50613391e-01 4.79434162e-01 -9.26725626e-01 9.43096399e-01 4.19034749e-01 -3.88947636e-01 -9.27355662e-02 -8.97006750e-01 -1.24333704e+00 1.19863905e-01 5.74206226e-02 8.33294153e-01 4.26154912e-01 -6.04344666e-01 1.49813546e-02 -1.50962304e-02 -1.07836449e+00 5.75636208e-01 1.16689801e+00 7.84420550e-01 -5.59249781e-02 5.44040501e-01 -5.58069289e-01 1.28164637e+00 -5.16362846e-01 -9.54638600e-01 -1.78745344e-01 2.82552689e-01 -1.36395943e+00 -5.27828395e-01 6.98606312e-01 -2.37195119e-01 -3.03925931e-01 5.95377803e-01 3.00123096e-01 3.05369914e-01 4.91356850e-01 -9.14573252e-01 -1.02338243e+00 9.37127471e-01 -1.90356970e-01 3.77258718e-01 4.64466244e-01 -1.04704574e-01 1.58364296e+00 -1.04751027e+00 8.61057043e-01 1.20970011e+00 7.31676519e-01 3.35113525e-01 -1.20324767e+00 -4.89938349e-01 2.85423398e-01 3.34170848e-01 -1.22919250e+00 -4.28539276e-01 3.47833723e-01 -2.02807069e-01 1.42887807e+00 7.39801452e-02 -1.09900497e-01 9.26010251e-01 2.57497486e-02 1.21116960e+00 9.55063164e-01 -4.89366353e-01 1.13982903e-02 3.52687299e-01 5.55725574e-01 8.26360464e-01 3.90610039e-01 -2.31147110e-01 -5.35976171e-01 -2.69778997e-01 3.18849593e-01 -3.63099724e-01 -1.38119131e-01 -3.19650173e-01 -1.18583035e+00 8.95309329e-01 5.21313250e-01 3.40907127e-01 -3.65114033e-01 3.32113832e-01 4.12877530e-01 2.49355942e-01 5.73320210e-01 1.11899137e+00 -1.21444535e+00 -1.53566182e-01 -5.48049390e-01 2.70660043e-01 1.23759651e+00 1.00508046e+00 9.01745021e-01 -3.68320495e-01 -2.18458936e-01 9.63989556e-01 8.38523880e-02 4.98847485e-01 3.50870222e-01 -9.53065038e-01 6.93346977e-01 6.78577542e-01 4.29205000e-01 -7.07733095e-01 -4.85191286e-01 -4.02566373e-01 -2.33463287e-01 -6.82155192e-01 5.49654901e-01 -2.62559384e-01 -9.57643747e-01 1.82923412e+00 3.20035785e-01 9.72102210e-02 3.71639073e-01 6.50828540e-01 1.21298277e+00 7.22818196e-01 6.26936197e-01 2.86736220e-01 1.55156314e+00 -1.17933285e+00 -6.12664282e-01 -4.93797958e-01 1.02989864e+00 -5.51800728e-01 1.38258862e+00 8.16024914e-02 -1.03267145e+00 -4.02235210e-01 -5.22576272e-01 -5.49010396e-01 -9.31385756e-01 4.34471816e-02 1.14180493e+00 3.44226152e-01 -8.80431950e-01 1.99053407e-01 -5.05796850e-01 -4.73639250e-01 2.80196279e-01 2.11216897e-01 -3.21121365e-01 -3.38687003e-01 -1.66119528e+00 1.06950283e+00 6.45023942e-01 -1.97176278e-01 -7.05294311e-01 -1.09311068e+00 -9.54900682e-01 3.87190402e-01 8.42536151e-01 -8.41875017e-01 1.48796296e+00 -5.85864723e-01 -1.10586679e+00 1.08984947e+00 -3.30136627e-01 -7.42719471e-01 4.70141135e-02 -6.96626842e-01 -4.55427021e-01 1.20255128e-01 3.25854510e-01 8.48206937e-01 3.73026371e-01 -1.17625690e+00 -6.27679586e-01 1.30654499e-02 6.86945617e-01 1.63767427e-01 -3.19652289e-01 2.54692249e-02 -4.78219807e-01 -4.53404963e-01 -4.50046718e-01 -6.13017380e-01 -2.18795482e-02 -3.69740486e-01 -2.85457879e-01 -8.48270833e-01 2.66488582e-01 -7.09758043e-01 1.16309524e+00 -1.72140956e+00 -1.09713256e-01 -2.03153312e-01 3.16431612e-01 4.54141140e-01 -4.23050404e-01 5.68746090e-01 7.50500411e-02 3.23201090e-01 -6.84811249e-02 -1.32556126e-01 3.05183649e-01 3.69029731e-01 -5.77999771e-01 -2.22836509e-01 3.66712391e-01 1.63682878e+00 -1.09265018e+00 -4.46451455e-01 -2.69626677e-01 3.89248244e-02 -6.31808579e-01 2.34160915e-01 -8.72355461e-01 -1.49719492e-01 -5.68157971e-01 5.63948333e-01 2.60989159e-01 -8.50242972e-01 -1.26656406e-02 -1.30409196e-01 2.22808689e-01 7.87931502e-01 -5.59660196e-01 1.90683544e+00 -6.02623105e-01 4.38053221e-01 -1.03568975e-02 -8.41420650e-01 4.99633402e-01 2.65626371e-01 2.19525248e-01 -8.07297707e-01 -2.66418248e-01 8.03663582e-02 -1.62617847e-01 -7.68894494e-01 7.31142342e-01 -1.44069463e-01 -3.41748983e-01 4.37674880e-01 4.56555486e-01 8.88700932e-02 2.34276056e-01 7.84032345e-01 1.27275813e+00 4.39775661e-02 2.63799310e-01 -2.17340693e-01 3.51512045e-01 4.60684627e-01 3.69367093e-01 1.02670479e+00 -1.10861942e-01 4.05381471e-02 3.19941282e-01 -3.65560234e-01 -7.18209267e-01 -9.83944893e-01 -1.06910981e-01 1.80415857e+00 1.59434050e-01 -5.96899331e-01 -3.70853901e-01 -1.13346481e+00 3.59279335e-01 1.03320110e+00 -5.67777693e-01 -2.06862986e-01 -5.77788532e-01 -6.12547576e-01 7.29823887e-01 6.21918082e-01 6.76405668e-01 -1.03289843e+00 -9.60176513e-02 3.54446769e-01 -5.49253583e-01 -1.46538103e+00 -3.20802152e-01 4.20742109e-02 -5.91375947e-01 -1.01403248e+00 -4.77559119e-01 -9.63294804e-01 2.79567629e-01 1.02012984e-01 1.82044768e+00 7.66718090e-02 -4.56584170e-02 8.01047862e-01 -4.19018596e-01 -4.02796894e-01 -2.73313206e-02 4.13659453e-01 -2.13143662e-01 -3.47970217e-01 8.45295548e-01 -1.84224218e-01 -4.32567000e-01 1.99731633e-01 -7.06220329e-01 -7.46642426e-02 6.52496755e-01 9.46229935e-01 3.58452886e-01 -1.99568644e-01 1.00035262e+00 -1.35066032e+00 7.63379872e-01 -8.44785392e-01 -3.61198783e-01 8.30792248e-01 -5.79668045e-01 3.70863140e-01 5.33150792e-01 -3.14721942e-01 -1.52929485e+00 -3.87397498e-01 -2.79657960e-01 4.20998596e-02 -1.62880018e-01 1.01345098e+00 -8.34892597e-03 5.13838045e-02 8.90371084e-01 3.71083170e-02 -6.39881372e-01 -5.72269797e-01 8.79197598e-01 3.92945498e-01 5.19516587e-01 -9.64691043e-01 8.39700222e-01 3.12779874e-01 -5.78321517e-01 -5.76311946e-01 -1.75497079e+00 -8.00230086e-01 -4.68547851e-01 3.93817455e-01 9.47959125e-01 -1.17870665e+00 -7.46190190e-01 1.25217810e-01 -1.30251420e+00 -5.38006842e-01 -3.10118049e-01 2.12887049e-01 -2.48012692e-01 1.65767863e-01 -8.52055728e-01 -5.05819738e-01 -3.82206202e-01 -2.65683413e-01 1.23924017e+00 1.29334018e-01 -1.35387197e-01 -1.34691834e+00 1.82087660e-01 9.17425752e-01 4.32273448e-01 -1.89763695e-01 1.20469701e+00 -1.07788575e+00 -8.50852847e-01 -2.07672510e-02 -4.58540976e-01 1.34606108e-01 -1.57354161e-01 -5.01752615e-01 -8.11857224e-01 -6.91647455e-02 -2.78679043e-01 -9.91271019e-01 1.04186130e+00 -7.86110312e-02 9.76238430e-01 -4.75136071e-01 -4.29010600e-01 2.43572131e-01 1.28530061e+00 -2.07633108e-01 4.08185333e-01 3.69458079e-01 5.81916094e-01 6.14711165e-01 6.41751647e-01 -1.33055761e-01 1.03213084e+00 4.34965402e-01 -3.94974127e-02 -5.31026907e-02 -2.21172079e-01 -6.52997613e-01 3.03182658e-02 6.91825926e-01 2.54398733e-01 -4.50813353e-01 -1.26338911e+00 8.63764644e-01 -1.90893698e+00 -8.21946204e-01 2.39751991e-02 1.80079257e+00 1.16072917e+00 2.12238178e-01 -2.87698925e-01 -6.87720895e-01 3.57821554e-01 2.13660330e-01 -7.35408068e-01 -1.33497104e-01 -2.95361102e-01 3.74662280e-01 3.67418706e-01 5.73426068e-01 -1.14105916e+00 1.50152349e+00 6.46625900e+00 7.75722444e-01 -6.94930673e-01 2.87814528e-01 4.27857697e-01 1.88545942e-01 -5.52873015e-01 5.77075668e-02 -9.15431917e-01 2.56428778e-01 1.08416057e+00 -3.20288867e-01 1.14231840e-01 7.62563646e-01 -2.99393207e-01 -1.29166424e-01 -1.09532499e+00 5.48204184e-01 2.20237285e-01 -1.59092832e+00 1.68175325e-01 -2.22522780e-01 8.53477538e-01 3.46611232e-01 -2.51436561e-01 1.18467879e+00 7.69710898e-01 -1.12152886e+00 2.59197265e-01 5.07315934e-01 6.58529520e-01 -4.13293928e-01 8.54820311e-01 6.41169012e-01 -8.81272793e-01 -7.60834441e-02 -4.25795913e-01 8.95548463e-02 3.52493286e-01 4.13416743e-01 -1.05936742e+00 5.38182616e-01 5.69653511e-01 5.60084701e-01 -6.60901606e-01 8.28010201e-01 -6.04701400e-01 1.07114398e+00 -3.06047559e-01 -2.11354241e-01 3.89932513e-01 4.46126074e-01 3.10327262e-01 1.40078533e+00 -3.97436798e-01 5.01323760e-01 8.85572210e-02 8.64406526e-01 -8.22266459e-01 1.21627435e-01 -3.19745600e-01 -3.61334592e-01 5.53408325e-01 1.13928068e+00 -3.23115513e-02 -6.14186883e-01 -5.83341658e-01 8.78649354e-01 9.79773760e-01 6.10090375e-01 -7.71575511e-01 -5.33840001e-01 3.54402393e-01 -1.83646269e-02 3.61009866e-01 -2.72313118e-01 -8.47161263e-02 -1.62286127e+00 7.14107603e-02 -7.32111037e-01 8.31822574e-01 -9.36098337e-01 -1.64117265e+00 3.13034058e-01 -1.04306892e-01 -3.65342796e-01 -2.75544494e-01 -7.39185631e-01 -4.98779565e-01 6.74446583e-01 -1.98385727e+00 -1.31797981e+00 -1.20749570e-01 5.28599858e-01 4.64736462e-01 -8.12997818e-02 9.24368083e-01 3.82411987e-01 -2.76081800e-01 6.69113278e-01 2.74574347e-02 5.29550493e-01 1.04262102e+00 -1.47136629e+00 5.09604096e-01 6.34887159e-01 5.89887738e-01 9.85072911e-01 4.06820029e-01 -6.57651305e-01 -1.44640684e+00 -9.83104527e-01 1.50689185e+00 -1.11285830e+00 9.91105080e-01 -5.60310662e-01 -1.18000150e+00 1.12924802e+00 3.20357978e-01 -2.19629668e-02 8.39822948e-01 9.20144856e-01 -8.92210066e-01 6.28771782e-02 -1.00295329e+00 2.65383780e-01 1.08757448e+00 -8.53487730e-01 -1.43207848e+00 6.72016680e-01 1.26483703e+00 -4.36646342e-01 -1.00906992e+00 4.50747639e-01 2.13062868e-01 -2.78530419e-01 1.16853678e+00 -1.38017333e+00 4.94343609e-01 3.96768637e-02 -3.55795701e-03 -1.21067166e+00 -3.41715425e-01 -4.60868448e-01 -5.88146031e-01 1.22506738e+00 9.00498271e-01 -7.36352444e-01 7.80747771e-01 9.72318947e-01 6.56835288e-02 -7.06230223e-01 -6.53222740e-01 -7.88507402e-01 3.67792159e-01 -2.49713361e-01 3.88026267e-01 1.22340071e+00 1.96839854e-01 9.02978361e-01 -2.97483131e-02 1.90383464e-01 4.09928173e-01 2.44936436e-01 5.81887484e-01 -1.09349751e+00 -3.06353450e-01 -1.34730995e-01 2.82982811e-02 -1.62999415e+00 6.40121043e-01 -1.06971323e+00 2.41098013e-02 -1.81116712e+00 3.31254214e-01 -8.17379177e-01 -3.98433983e-01 6.68714523e-01 -7.01394141e-01 -1.63356647e-01 -7.75341541e-02 8.15302804e-02 -1.34930146e+00 5.56045115e-01 9.65409815e-01 -8.45936164e-02 1.16281494e-01 -3.04889292e-01 -9.80440319e-01 5.67441821e-01 6.68757379e-01 -3.22304845e-01 -3.86269301e-01 -1.04202104e+00 6.55046880e-01 -1.29869491e-01 3.43344480e-01 -3.86717588e-01 5.77417195e-01 -4.83586006e-02 9.11897868e-02 -4.18841034e-01 4.09219235e-01 -4.93859231e-01 -6.67237580e-01 -1.68595836e-02 -7.43121564e-01 -1.47803158e-01 3.86879057e-01 7.57521391e-01 -3.92897934e-01 -1.90078154e-01 1.56246588e-01 -2.95527488e-01 -1.12692571e+00 2.40369707e-01 -2.57088058e-02 8.84423733e-01 6.37890935e-01 3.38885158e-01 -8.78123879e-01 -5.24116874e-01 -5.65478981e-01 7.69506633e-01 -7.23385885e-02 6.24788165e-01 2.40752771e-01 -9.72612798e-01 -8.32426548e-01 -4.56654191e-01 4.08978939e-01 7.37560987e-02 1.19823322e-01 5.23560286e-01 -1.37797207e-01 1.12089264e+00 2.74599820e-01 -2.36183271e-01 -7.52083659e-01 6.57573104e-01 2.48042166e-01 -7.67481029e-01 -2.39187926e-01 1.28002787e+00 2.47662500e-01 -9.89626586e-01 2.10565925e-01 -3.31285894e-01 -8.11736807e-02 -1.52562007e-01 4.03839916e-01 1.32490285e-02 1.17311373e-01 -2.11998403e-01 -4.13371861e-01 1.48671910e-01 -5.23188889e-01 8.40658769e-02 1.35884929e+00 -6.86045662e-02 -8.58273357e-02 2.68413931e-01 1.17313969e+00 2.11233914e-01 -8.63817692e-01 -7.21558809e-01 6.13792956e-01 -6.50198311e-02 -1.55205324e-01 -1.57305896e+00 -4.70941961e-01 7.37889349e-01 -1.62066311e-01 -1.42056635e-02 6.28373265e-01 5.54495096e-01 1.11413503e+00 1.15829277e+00 6.62694216e-01 -1.00105691e+00 1.23579986e-01 9.32773650e-01 7.29734421e-01 -1.66350484e+00 -2.27983773e-01 -4.59033310e-01 -6.09180093e-01 6.41065300e-01 9.18343544e-01 1.30848408e-01 4.74200219e-01 4.11905199e-02 -1.04308620e-01 -4.72617716e-01 -1.09436870e+00 -3.91468287e-01 5.56001961e-01 2.30604172e-01 5.73583484e-01 -6.45577349e-03 -2.71049976e-01 8.35596740e-01 -2.48024508e-01 7.07073733e-02 7.69515261e-02 8.52721453e-01 -6.28441811e-01 -8.64825547e-01 8.17345306e-02 5.58098257e-01 -4.99238163e-01 -4.99768049e-01 -5.06348073e-01 8.74084771e-01 -4.58203703e-02 9.27304327e-01 -1.24289714e-01 -4.22258191e-02 5.00912488e-01 5.07175148e-01 2.85313129e-01 -9.54421878e-01 -7.40762651e-01 -6.69773519e-01 7.12401152e-01 -5.93166232e-01 -2.41591051e-01 -2.76373088e-01 -1.36674309e+00 -1.55405745e-01 -3.73046726e-01 5.29822588e-01 3.22782010e-01 1.08820963e+00 6.80325091e-01 2.39906996e-01 -8.98214728e-02 2.84707099e-01 -6.42983675e-01 -9.01682496e-01 -6.10791184e-02 6.35682523e-01 2.17946813e-01 -4.82360005e-01 -2.09798604e-01 -2.98628975e-02]
[10.582815170288086, 8.071761131286621]
fc09ff9d-a2c8-488c-8527-42a25575687e
illiterate-dall-cdot-e-learns-to-compose-1
2110.11405
null
https://arxiv.org/abs/2110.11405v3
https://arxiv.org/pdf/2110.11405v3.pdf
Illiterate DALL-E Learns to Compose
Although DALL-E has shown an impressive ability of composition-based systematic generalization in image generation, it requires the dataset of text-image pairs and the compositionality is provided by the text. In contrast, object-centric representation models like the Slot Attention model learn composable representations without the text prompt. However, unlike DALL-E its ability to systematically generalize for zero-shot generation is significantly limited. In this paper, we propose a simple but novel slot-based autoencoding architecture, called SLATE, for combining the best of both worlds: learning object-centric representations that allows systematic generalization in zero-shot image generation without text. As such, this model can also be seen as an illiterate DALL-E model. Unlike the pixel-mixture decoders of existing object-centric representation models, we propose to use the Image GPT decoder conditioned on the slots for capturing complex interactions among the slots and pixels. In experiments, we show that this simple and easy-to-implement architecture not requiring a text prompt achieves significant improvement in in-distribution and out-of-distribution (zero-shot) image generation and qualitatively comparable or better slot-attention structure than the models based on mixture decoders.
['Sungjin Ahn', 'Fei Deng', 'Gautam Singh']
2021-10-17
null
null
null
null
['systematic-generalization']
['reasoning']
[ 4.51045513e-01 4.52522218e-01 2.84289774e-02 -2.30102479e-01 -8.82677078e-01 -2.12222695e-01 1.09703767e+00 -4.29950058e-01 -1.38347670e-01 6.54182553e-01 1.96861118e-01 -3.62839758e-01 5.77584803e-02 -9.25601959e-01 -1.06954396e+00 -7.67388940e-01 4.93627310e-01 7.32940316e-01 2.13159040e-01 -3.00375432e-01 -5.08675426e-02 -4.02136380e-03 -1.88913798e+00 4.62360501e-01 7.52905250e-01 9.26356316e-01 7.48397529e-01 7.02665925e-01 -5.55286348e-01 7.67671466e-01 -4.82458860e-01 -2.65915841e-01 3.40686619e-01 -1.07911277e+00 -6.06237829e-01 4.71841425e-01 7.89197683e-01 -3.61584425e-01 -4.57608998e-01 9.32415724e-01 5.46712756e-01 2.65539557e-01 8.78066301e-01 -1.23274529e+00 -1.40123188e+00 8.90657723e-01 -4.05031025e-01 -3.95716093e-02 -2.15511434e-02 4.98852491e-01 9.52165723e-01 -1.00816321e+00 9.06261265e-01 1.39872539e+00 4.11394626e-01 1.09826159e+00 -1.57398963e+00 -4.55754936e-01 1.73901260e-01 -8.90267119e-02 -1.19783270e+00 -4.39186424e-01 4.63385820e-01 -3.94194096e-01 1.15476692e+00 7.05154017e-02 6.14497244e-01 1.32696390e+00 -1.93163250e-02 1.13790631e+00 1.14149261e+00 -5.28632343e-01 3.41370553e-01 2.73385048e-01 -1.51779860e-01 7.04969227e-01 8.71022493e-02 1.91478748e-02 -5.18125951e-01 2.54179507e-01 1.05264747e+00 9.01849940e-03 -2.11048514e-01 -3.65122825e-01 -1.25610948e+00 7.45626867e-01 5.66200078e-01 1.83607757e-01 -4.16034937e-01 8.07541430e-01 2.06532612e-01 1.97794959e-01 5.92442036e-01 6.21876001e-01 -3.02618414e-01 7.00507089e-02 -1.17750037e+00 3.90904367e-01 3.12419623e-01 1.38120615e+00 9.32617903e-01 6.28856838e-01 -6.18217945e-01 7.14093804e-01 1.01434357e-01 4.88817692e-01 8.40828836e-01 -9.35613334e-01 1.29304618e-01 2.74765760e-01 1.72588900e-02 -1.46838680e-01 -1.40211865e-01 -5.25101602e-01 -6.77449584e-01 2.44953737e-01 1.61204398e-01 -1.71005502e-01 -1.66966617e+00 2.13791752e+00 -1.09040458e-02 2.61681139e-01 2.64788121e-01 8.08036923e-01 8.02064717e-01 9.09609675e-01 3.21735084e-01 -2.96842679e-02 1.33559680e+00 -1.25982416e+00 -6.84488714e-01 -2.18901321e-01 3.74064505e-01 -5.64640284e-01 1.35599172e+00 -9.54204798e-02 -1.47702253e+00 -7.31745541e-01 -8.77477825e-01 -4.20856774e-01 -5.27912438e-01 -7.20901340e-02 7.01235116e-01 5.91370106e-01 -1.46500337e+00 3.96620989e-01 -5.03398180e-01 -6.88212097e-01 5.90394855e-01 6.94452971e-02 -1.06084324e-01 -1.83335934e-02 -1.04922330e+00 7.88943410e-01 3.74838293e-01 -4.41479474e-01 -1.30568576e+00 -9.77997482e-01 -9.58610237e-01 2.84787565e-01 3.06132078e-01 -1.29447758e+00 1.55246603e+00 -1.25913417e+00 -1.48575437e+00 7.87364960e-01 -2.05108285e-01 -6.46172047e-01 2.37038836e-01 2.86870301e-02 1.23804688e-01 8.87573734e-02 1.72256127e-01 1.60647595e+00 9.97991741e-01 -1.48103249e+00 -4.50616539e-01 -1.51631176e-01 2.77123719e-01 3.43690097e-01 -1.99099675e-01 -4.09242332e-01 -4.50355619e-01 -8.19697320e-01 -1.92883089e-01 -8.11143637e-01 -2.68400997e-01 3.07049990e-01 -2.01859236e-01 -1.70269206e-01 7.15295851e-01 -2.73750216e-01 8.28888595e-01 -2.15209913e+00 2.17063323e-01 -4.93967831e-01 3.75842787e-02 7.22081214e-02 -5.53547740e-01 5.68975747e-01 -1.03834882e-01 2.46369362e-01 -3.35000455e-01 -7.38672376e-01 2.33851686e-01 3.45613569e-01 -6.21724486e-01 -2.96258386e-02 5.23763835e-01 1.32709801e+00 -8.59724045e-01 -2.39069328e-01 2.55547017e-01 3.06838810e-01 -8.36344361e-01 5.18646576e-02 -9.17178035e-01 2.95372549e-02 -5.78140467e-02 3.27938318e-01 4.15063411e-01 -5.63706040e-01 -1.30391479e-01 7.07961842e-02 1.05088837e-01 1.58307925e-01 -6.55030489e-01 2.22710991e+00 -5.57790220e-01 7.28493154e-01 -2.40319535e-01 -8.59722018e-01 6.38907373e-01 4.20166254e-01 1.13255389e-01 -9.03773546e-01 -2.98406165e-02 2.09186107e-01 -4.72725928e-02 -1.75135657e-01 8.22620332e-01 -4.88291383e-01 -1.32189363e-01 8.18398774e-01 7.33405173e-01 -5.21128297e-01 3.76775503e-01 6.01058602e-01 7.66626000e-01 3.68689030e-01 1.63807776e-02 -3.20685804e-01 -5.17087989e-02 1.40035916e-02 1.08038507e-01 1.13103306e+00 2.33669564e-01 9.44850683e-01 2.48249099e-01 -1.39218077e-01 -1.46252739e+00 -1.33244812e+00 -7.51909018e-02 1.06365967e+00 1.21385574e-01 -4.27341849e-01 -1.02336013e+00 -4.12201315e-01 -2.92997897e-01 1.12683511e+00 -8.15526426e-01 -2.59173751e-01 -7.11169615e-02 -5.36274910e-01 4.01932776e-01 6.39467657e-01 5.98205566e-01 -1.11619341e+00 -6.83867931e-01 1.34805456e-01 -1.03101291e-01 -8.36293876e-01 -5.42824209e-01 3.53431225e-01 -7.09408879e-01 -6.34526134e-01 -1.26545000e+00 -7.11933672e-01 7.46110141e-01 3.56751680e-01 1.15086925e+00 -1.82156920e-01 -5.67684650e-01 7.68534601e-01 -3.46716195e-01 -5.42300344e-01 -2.96134382e-01 -2.30237879e-02 -3.23911190e-01 -1.54121369e-01 1.40952766e-01 -4.97526169e-01 -5.23060799e-01 -6.72266856e-02 -1.35481238e+00 5.41293025e-01 5.83291113e-01 1.21652627e+00 4.70447958e-01 -1.40759423e-01 6.01447344e-01 -8.76815975e-01 6.05128884e-01 -5.86623847e-01 -2.74832010e-01 2.13672549e-01 -4.71108437e-01 3.82864088e-01 5.10699630e-01 -5.98983467e-01 -1.35657728e+00 -3.76993529e-02 6.18622005e-02 -6.54247284e-01 -1.91752285e-01 1.64339438e-01 8.20995569e-02 3.49009961e-01 9.31203306e-01 7.53627896e-01 -6.53323159e-02 -2.65484333e-01 9.39369380e-01 4.08711880e-01 4.50778782e-01 -6.73022926e-01 6.66093647e-01 4.06346679e-01 -1.77459970e-01 -8.13874602e-01 -6.31160617e-01 -1.45948917e-01 -1.67429551e-01 -3.82010117e-02 1.15181744e+00 -1.12058818e+00 -1.15230717e-01 5.14820099e-01 -1.21636522e+00 -8.56465816e-01 -1.06303704e+00 2.31661238e-02 -1.10291004e+00 5.86847141e-02 -5.96023500e-01 -9.49187517e-01 -2.09075972e-01 -1.15954387e+00 1.40762484e+00 1.88483849e-01 -1.77453473e-01 -9.99456227e-01 -2.99222261e-01 -7.68591010e-04 7.40545154e-01 -2.15707228e-01 1.06195426e+00 -3.18198889e-01 -1.05521739e+00 2.53726453e-01 -3.51402193e-01 2.10683092e-01 -1.01121210e-01 -3.86242390e-01 -1.10110521e+00 -2.87567023e-02 -1.46489471e-01 -6.52961135e-01 1.32214558e+00 5.39237857e-01 1.20419180e+00 -2.24418059e-01 -8.81960541e-02 6.49127483e-01 1.45910656e+00 1.68476090e-01 9.19850647e-01 -3.88944186e-02 3.53880107e-01 4.98630136e-01 2.08858877e-01 4.40466523e-01 4.59635049e-01 5.99917948e-01 2.08514407e-01 -3.57706726e-01 -8.80039275e-01 -6.55795097e-01 3.56638074e-01 3.05393845e-01 8.24378952e-02 -5.82705557e-01 -4.83319730e-01 7.35974491e-01 -1.96336973e+00 -1.12988651e+00 1.13246657e-01 1.79729855e+00 8.70076776e-01 -2.31407717e-01 -3.84099446e-02 -2.85512090e-01 5.20797491e-01 2.41721794e-01 -5.46112001e-01 -3.32409859e-01 -2.77405828e-01 3.13229650e-01 3.54468524e-01 4.29695845e-01 -7.54991055e-01 1.38726687e+00 7.29312801e+00 1.20055604e+00 -9.34524715e-01 4.25981790e-01 7.10233688e-01 -2.94731319e-01 -8.59664202e-01 -2.29341015e-02 -7.08555043e-01 4.13496494e-01 9.37931895e-01 -1.50209248e-01 5.64851165e-01 8.13208878e-01 -1.76290259e-01 -2.18337104e-01 -1.12941098e+00 9.68844414e-01 3.28064829e-01 -1.71018684e+00 5.75731337e-01 1.90912820e-02 1.16691387e+00 -1.68097168e-01 5.84564090e-01 4.97587562e-01 5.06329894e-01 -9.61922169e-01 1.14263034e+00 5.73640168e-01 1.10677445e+00 -2.81909376e-01 1.47772595e-01 3.38060528e-01 -8.08686197e-01 -1.50994033e-01 -5.55689514e-01 -4.01221737e-02 2.69824624e-01 3.02449077e-01 -5.78922391e-01 2.17782795e-01 3.01831841e-01 3.83198053e-01 -3.23499799e-01 7.53726721e-01 -1.75874949e-01 3.30789834e-01 -9.72003862e-02 1.27636513e-03 5.35205424e-01 2.25447997e-01 3.67568165e-01 1.14414704e+00 6.38153553e-01 1.33462980e-01 4.76737320e-02 1.46228325e+00 -1.89590141e-01 -1.74737483e-01 -7.08923578e-01 -2.70187885e-01 1.16142400e-01 8.89457405e-01 -7.61017978e-01 -8.00222158e-01 -3.92838120e-01 1.38008344e+00 1.26218632e-01 8.01890492e-01 -7.68719912e-01 -1.01298347e-01 5.59260309e-01 2.61298358e-01 6.62249863e-01 -5.14431819e-02 -3.01051974e-01 -1.23266137e+00 -3.95634711e-01 -6.20344877e-01 -1.05762154e-01 -1.39137030e+00 -1.23315036e+00 5.75729430e-01 2.62350172e-01 -7.90300548e-01 -4.60364282e-01 -7.54083872e-01 -4.88768548e-01 1.01755071e+00 -1.36796522e+00 -1.47249663e+00 -1.38127074e-01 8.29014838e-01 1.16506481e+00 -3.36024106e-01 9.31039035e-01 6.94692656e-02 -2.23401949e-01 5.24069846e-01 -1.56057522e-01 -3.44292641e-01 4.11524445e-01 -1.37747300e+00 6.72542453e-01 7.43892133e-01 4.38143581e-01 6.75230086e-01 7.42139041e-01 -6.12118125e-01 -1.30556846e+00 -1.14328039e+00 7.21063197e-01 -3.46991330e-01 4.66885149e-01 -5.99831045e-01 -5.38953900e-01 9.37003553e-01 5.43925762e-01 -1.82792082e-01 5.34842253e-01 -2.70768572e-02 -5.10153294e-01 2.38815144e-01 -7.77316630e-01 9.83169675e-01 1.25298870e+00 -6.57328725e-01 -5.41121185e-01 4.05194312e-01 9.86462772e-01 -1.96690604e-01 -2.59460092e-01 -2.88957488e-02 2.28081495e-01 -9.15404677e-01 8.48755598e-01 -5.09636819e-01 8.19726586e-01 -1.44845575e-01 -3.86101842e-01 -1.28989732e+00 -5.54844797e-01 -5.10901332e-01 8.83248523e-02 1.02020693e+00 5.97705543e-01 -4.45788234e-01 7.24745989e-01 4.95076597e-01 -5.63422084e-01 -5.40285289e-01 -6.95927441e-01 -8.69437158e-01 2.10206687e-01 -5.03649294e-01 5.32612443e-01 4.87253964e-01 -7.58105591e-02 5.88990629e-01 -5.12819767e-01 -3.84987116e-01 4.78073210e-01 2.54850704e-02 8.26093972e-01 -7.28871942e-01 -6.19438887e-01 -5.98809659e-01 -2.23277524e-01 -1.50813401e+00 1.53305456e-01 -1.17677319e+00 4.09620583e-01 -1.81432533e+00 5.09056926e-01 -3.62087995e-01 -1.43746749e-01 4.81792867e-01 -1.92210451e-01 3.95095021e-01 4.82056290e-01 1.20109960e-01 -6.47718668e-01 8.36054742e-01 1.43862426e+00 -3.01911771e-01 1.15221053e-01 -4.77821559e-01 -9.25313711e-01 3.71855885e-01 4.49045002e-01 -2.33947888e-01 -9.36234176e-01 -5.83586276e-01 -3.86513993e-02 3.68349701e-02 6.08262181e-01 -9.94030058e-01 2.27093488e-01 2.56422181e-02 4.16367829e-01 -4.32810187e-01 7.04615414e-01 -5.37396014e-01 2.57913053e-01 2.06231490e-01 -5.91075718e-01 -3.09526384e-01 3.39149326e-01 6.14702940e-01 -4.41806801e-02 -2.81417340e-01 6.12318754e-01 -6.36179745e-01 -8.66742253e-01 4.18732435e-01 -5.87434471e-01 -9.32390019e-02 9.62821662e-01 -3.67566764e-01 -5.09497941e-01 -4.73682702e-01 -6.87273026e-01 -1.53076187e-01 6.65667057e-01 4.39398199e-01 6.61875606e-01 -1.33850050e+00 -6.00095570e-01 3.31997156e-01 2.22700641e-01 -1.93928719e-01 6.38353765e-01 4.64292824e-01 -1.26214460e-01 4.40479606e-01 -2.66642094e-01 -6.08105063e-01 -6.02544725e-01 7.74014175e-01 2.52939254e-01 -4.79444154e-02 -5.58506608e-01 1.12305415e+00 1.03657389e+00 -2.83532292e-01 -4.86643612e-02 -2.58018911e-01 4.87173051e-01 -1.07699871e-01 5.42694092e-01 -1.38485625e-01 -4.68479484e-01 -2.85780311e-01 2.12342843e-01 3.21087480e-01 -3.60943340e-02 -6.63189411e-01 1.10917783e+00 -1.28453985e-01 1.19615331e-01 7.50570834e-01 8.03847075e-01 -5.71733594e-01 -1.59140575e+00 -8.12279880e-02 -5.10226607e-01 -3.85630995e-01 5.87002747e-02 -9.60796475e-01 -8.62351537e-01 1.34209955e+00 3.90711457e-01 1.18967667e-02 1.12888253e+00 1.88904643e-01 8.09578121e-01 2.51344919e-01 3.60434294e-01 -9.33113635e-01 4.95077431e-01 4.08624262e-01 1.00108445e+00 -1.07898927e+00 -5.12021303e-01 -1.10703118e-01 -7.51693785e-01 8.47411513e-01 7.14130104e-01 -2.02406481e-01 3.95413727e-01 1.71854615e-01 -1.48201793e-01 -1.59549817e-01 -1.16951156e+00 -5.87174356e-01 8.56140405e-02 9.42878008e-01 3.40667099e-01 -2.56687477e-02 -1.58986021e-02 3.30598474e-01 -1.20938137e-01 7.71530624e-03 4.28728849e-01 5.86300611e-01 -6.30806565e-01 -9.50218141e-01 -1.03116289e-01 3.56965333e-01 -1.51646063e-01 -5.16470015e-01 -1.32307708e-01 7.72928596e-01 2.16225326e-01 5.58442652e-01 4.28607464e-01 -1.09904692e-01 -4.28308137e-02 4.32161838e-01 7.97738910e-01 -1.00509965e+00 -3.71547490e-01 2.06745863e-01 -1.35586634e-01 -3.33496869e-01 -2.84653574e-01 -4.20947134e-01 -1.06068528e+00 -8.03997442e-02 -1.51123866e-01 -2.01462835e-01 7.28294671e-01 7.15032399e-01 4.72733766e-01 8.84938121e-01 7.12193176e-02 -1.04094672e+00 -3.59905928e-01 -9.81579423e-01 -7.63934314e-01 5.28800547e-01 2.95152307e-01 -5.64192772e-01 -1.08827583e-01 3.08211148e-01]
[11.146866798400879, -0.04014776274561882]
3500633d-9178-482c-ae6f-9928b792e32a
c-nn-fd-a-deep-learning-framework-for
2306.05889
null
https://arxiv.org/abs/2306.05889v1
https://arxiv.org/pdf/2306.05889v1.pdf
C(NN)FD -- a deep learning framework for turbomachinery CFD analysis
Deep Learning methods have seen a wide range of successful applications across different industries. Up until now, applications to physical simulations such as CFD (Computational Fluid Dynamics), have been limited to simple test-cases of minor industrial relevance. This paper demonstrates the development of a novel deep learning framework for real-time predictions of the impact of manufacturing and build variations on the overall performance of axial compressors in gas turbines, with a focus on tip clearance variations. The associated scatter in efficiency can significantly increase the $CO_2$ emissions, thus being of great industrial and environmental relevance. The proposed \textit{C(NN)FD} architecture achieves in real-time accuracy comparable to the CFD benchmark. Predicting the flow field and using it to calculate the corresponding overall performance renders the methodology generalisable, while filtering only relevant parts of the CFD solution makes the methodology scalable to industrial applications.
['Senthil K. Krishnababu', 'Sepehr Maleki', 'Giuseppe Bruni']
2023-06-09
null
null
null
null
['physical-simulations']
['miscellaneous']
[-3.26296866e-01 -3.11805725e-01 2.94323117e-01 -1.33295832e-02 -9.83656198e-02 -5.79054415e-01 5.41034877e-01 3.07705551e-01 -3.07130590e-02 6.88215733e-01 -4.27359760e-01 -7.21014738e-01 -6.97412550e-01 -9.22270000e-01 -4.91770506e-01 -9.36230242e-01 -4.17542607e-01 5.04288733e-01 -3.96243334e-01 -1.35687605e-01 1.91658765e-01 1.07497227e+00 -1.57445657e+00 7.06096813e-02 3.56553376e-01 9.84263897e-01 2.53292114e-01 4.61503565e-01 6.72764108e-02 6.51400208e-01 -6.44057274e-01 1.14316970e-01 5.66590309e-01 -2.06276149e-01 -7.11269438e-01 -2.34675974e-01 1.11122668e-01 -2.88965613e-01 -3.09693664e-01 6.89594924e-01 7.91800022e-01 7.71451116e-01 6.03109717e-01 -7.11925566e-01 2.83007234e-01 4.22900110e-01 -2.89699733e-01 3.54959637e-01 -4.08866525e-01 3.41483772e-01 6.31869435e-01 -6.29747272e-01 3.17675546e-02 9.67849135e-01 7.90319622e-01 -2.88968403e-02 -7.63638794e-01 -4.39199001e-01 -3.13009083e-01 1.20288014e-01 -9.00849402e-01 8.72808397e-02 5.19158602e-01 -7.54002929e-01 1.28706908e+00 3.44782293e-01 6.65052772e-01 3.33901256e-01 7.50929415e-01 1.84680045e-01 1.11454606e+00 -4.09712568e-02 4.48001057e-01 -2.14471802e-01 -4.25280541e-01 3.05720747e-01 2.57930577e-01 7.74335444e-01 1.49403676e-01 5.32341540e-01 8.65136743e-01 -9.70675424e-02 -1.28606129e-02 -2.11827531e-01 -6.82878554e-01 9.59526777e-01 6.83968306e-01 5.65826654e-01 -1.78877309e-01 2.69395262e-01 9.44457650e-01 3.95976335e-01 3.84056777e-01 1.05020404e+00 -9.57497060e-01 -4.35858876e-01 -8.53905320e-01 6.51213408e-01 7.71803081e-01 5.31679332e-01 2.54000694e-01 6.13091946e-01 4.15509157e-02 4.87525433e-01 -9.40879136e-02 3.75874549e-01 3.95958066e-01 -9.64978516e-01 1.75121650e-01 2.48638630e-01 1.26185820e-01 -9.52938139e-01 -7.48033345e-01 -8.20681453e-01 -1.00274777e+00 6.25051439e-01 2.97822237e-01 -7.94508755e-01 -5.04838943e-01 7.83432722e-01 2.98087955e-01 -1.08197458e-01 -2.18452737e-01 9.43064511e-01 5.74587345e-01 8.26540172e-01 3.28859948e-02 -1.91775307e-01 8.18983078e-01 -8.76484931e-01 -7.10309565e-01 3.93928029e-03 7.24174261e-01 -8.52655888e-01 5.31048298e-01 7.46620953e-01 -1.10331869e+00 -1.02034295e+00 -9.81124878e-01 6.65068403e-02 -8.57183158e-01 2.14480475e-01 6.44194543e-01 7.14990377e-01 -6.26842797e-01 1.63342130e+00 -8.52092624e-01 1.37741625e-01 3.98122668e-01 5.87486625e-01 1.52756467e-01 2.35736936e-01 -1.11315167e+00 1.33438826e+00 3.99389207e-01 5.63762605e-01 -9.21562910e-01 -1.14918363e+00 -6.20685339e-01 3.77046347e-01 3.26570332e-01 -5.21091759e-01 1.45066953e+00 -4.17096943e-01 -1.78426027e+00 1.46629261e-02 3.02934676e-01 -5.44768155e-01 8.16015840e-01 -6.04371846e-01 -3.11017334e-01 -2.52341926e-01 -4.41012621e-01 8.21430385e-02 6.34628832e-01 -1.11348248e+00 -8.71030509e-01 -8.40633661e-02 1.17087163e-01 1.70144544e-03 -5.89379370e-02 -1.78507939e-01 3.95512521e-01 -4.23454314e-01 -6.44424617e-01 -7.90346622e-01 -6.04100227e-01 -4.73847628e-01 -2.32376873e-01 -1.06964380e-01 1.12994349e+00 -6.85344338e-01 1.02109623e+00 -1.72855556e+00 1.47588745e-01 7.44905695e-02 -1.94231018e-01 7.58546293e-01 3.47564906e-01 6.32027388e-01 -4.53908354e-01 5.49238734e-02 -1.35715395e-01 9.26664993e-02 1.44316424e-02 1.79238945e-01 -1.90576747e-01 6.48310006e-01 4.82667595e-01 5.58180392e-01 -1.03726935e+00 1.49162829e-01 1.26416337e+00 4.54840034e-01 -2.96195328e-01 2.94399440e-01 -3.36858518e-02 5.90173244e-01 -5.69958091e-01 -3.96586098e-02 6.96509540e-01 4.61926907e-01 4.88684140e-02 -2.42912427e-01 -6.04715765e-01 1.02097616e-01 -9.36858952e-01 1.20253074e+00 -1.47484016e+00 7.93390155e-01 3.54694724e-01 -1.43019485e+00 1.05076134e+00 1.81513190e-01 8.27160358e-01 -6.07442439e-01 5.91841578e-01 2.62339950e-01 4.93253171e-01 -6.32553816e-01 3.70642751e-01 -5.45640469e-01 2.39376813e-01 -5.27835041e-02 8.21597576e-02 -8.68978322e-01 2.40525410e-01 -5.72975099e-01 8.75524104e-01 1.81488663e-01 8.14267695e-02 -7.75705397e-01 5.67859888e-01 9.41544101e-02 9.15440470e-02 1.67452872e-01 1.76592410e-01 1.50135189e-01 4.01157677e-01 -7.11743057e-01 -1.35168314e+00 -7.43103862e-01 -4.27394658e-01 5.44983387e-01 7.04858676e-02 -3.92339565e-02 -6.78786516e-01 -2.86593795e-01 3.58830184e-01 9.32613909e-01 -3.93811166e-01 -2.23330721e-01 -8.08310151e-01 -6.48350656e-01 -8.38372950e-03 7.76508391e-01 3.72359663e-01 -9.04986084e-01 -1.11865747e+00 3.62535745e-01 7.35813618e-01 -9.52041566e-01 3.73791218e-01 6.53682411e-01 -1.06028044e+00 -1.13363540e+00 -3.18330973e-01 -4.52653319e-01 1.51246160e-01 -3.10635298e-01 1.31597972e+00 1.27308257e-02 -7.58505821e-01 -2.43977383e-01 -2.19435915e-01 -8.44954073e-01 -5.21104991e-01 1.00673534e-01 1.38373803e-02 -7.30027854e-01 5.01882890e-03 -2.05648616e-01 -6.49832666e-01 3.52351129e-01 -6.00709319e-01 -3.52901489e-01 3.78343135e-01 8.35836530e-01 2.58154780e-01 9.55266118e-01 6.21591151e-01 -8.97696495e-01 6.33953631e-01 -3.82918894e-01 -9.03380990e-01 -5.51112413e-01 -8.21081519e-01 -3.20634812e-01 1.36171198e+00 3.55144963e-02 -1.22776449e+00 -7.91279376e-02 -3.81448209e-01 -7.59376764e-01 -4.72802579e-01 6.56992733e-01 5.02134264e-02 5.19805774e-02 2.87652642e-01 -4.86403912e-01 -1.03974920e-02 -7.21404850e-01 1.83252729e-02 2.71357685e-01 3.09932202e-01 -5.17268836e-01 9.55161691e-01 -3.74828503e-02 7.98704624e-01 -1.10568166e+00 -4.19967681e-01 -4.08318341e-01 -8.27883780e-01 -3.22888285e-01 4.89493161e-01 -7.24883199e-01 -7.64255464e-01 3.50907892e-01 -8.34780335e-01 -6.55380487e-01 -8.34227145e-01 5.84989369e-01 -5.27122140e-01 6.66084364e-02 -6.47081196e-01 -7.75784075e-01 -3.17204326e-01 -1.43528020e+00 6.90044463e-01 2.09378213e-01 -1.54964581e-01 -1.40007126e+00 -2.20835358e-01 5.01777790e-02 6.33385658e-01 7.62758553e-01 1.00837457e+00 -3.46817613e-01 -7.50941560e-02 -2.77401239e-01 -6.46564923e-03 7.35269248e-01 2.40997702e-01 3.66806686e-01 -1.20018137e+00 -3.54126453e-01 1.57776207e-01 9.80259199e-03 6.26446128e-01 6.36858344e-01 1.67108774e+00 3.38283218e-02 -4.36831713e-01 4.94012862e-01 1.50666690e+00 4.37975675e-01 2.66375482e-01 1.81173608e-01 6.95991099e-01 5.50710201e-01 9.56047654e-01 7.81415284e-01 -7.81600893e-01 4.12058532e-01 7.70432830e-01 -3.52688283e-01 1.74728166e-02 3.78254235e-01 2.71205921e-02 7.06711650e-01 -5.46036243e-01 -3.20210844e-01 -8.79742384e-01 6.63758755e-01 -1.26937366e+00 -8.54124188e-01 -5.16096234e-01 2.06401014e+00 2.35539868e-01 1.17084607e-01 -1.61958084e-01 4.82532978e-01 4.31767493e-01 9.32402909e-02 -5.14374614e-01 -1.30691683e+00 5.27664721e-01 9.20835495e-01 9.43552434e-01 1.06997311e-01 -1.20141721e+00 1.79310739e-01 6.14098167e+00 8.88621628e-01 -1.35392880e+00 -5.14696054e-02 8.14983070e-01 -2.14174509e-01 2.04493925e-01 -2.81366229e-01 -4.70620841e-01 2.17016771e-01 1.40731156e+00 -3.44246238e-01 3.65919471e-01 8.06126118e-01 7.11840034e-01 1.55117642e-02 -1.18958879e+00 4.86030757e-01 -6.36777878e-01 -1.28461766e+00 -2.58550882e-01 2.58475095e-01 8.45433295e-01 1.81321561e-01 8.15122649e-02 5.19441605e-01 9.70235467e-02 -1.21580017e+00 4.25139785e-01 1.81571603e-01 8.53190541e-01 -1.26048076e+00 1.32975626e+00 1.52487442e-01 -1.10669148e+00 -7.84573495e-01 -4.60204691e-01 -4.68201548e-01 2.22288519e-01 8.99464667e-01 -1.10722184e+00 1.02933133e+00 8.58048856e-01 5.26855052e-01 4.55620512e-02 9.51981127e-01 8.54993537e-02 6.16001368e-01 -3.42730373e-01 -2.13801451e-02 6.02438688e-01 -3.38624448e-01 2.58019775e-01 1.28050244e+00 7.19406545e-01 -3.78132284e-01 6.35313615e-02 8.55347216e-01 1.16988912e-01 1.29215226e-01 -6.42841101e-01 -1.42262205e-01 1.16999730e-01 1.39287341e+00 -5.78725040e-01 -4.73530143e-02 -1.46799088e-01 1.16952106e-01 -4.52644944e-01 7.55564049e-02 -8.87033343e-01 -5.50207496e-01 9.56508338e-01 5.79655886e-01 5.18249273e-01 -3.11488152e-01 -5.83487928e-01 -1.64054438e-01 -4.19506073e-01 -4.68067735e-01 2.26813350e-02 -4.07795846e-01 -1.13076127e+00 1.84908837e-01 2.46881679e-01 -1.24919200e+00 -4.35714722e-01 -1.40760446e+00 -9.40298617e-01 1.26103020e+00 -1.57663107e+00 -6.09655201e-01 -2.49994382e-01 2.30992250e-02 1.03346860e+00 -2.21068636e-01 7.43468523e-01 4.66725796e-01 -7.17774749e-01 -7.45575735e-03 7.66997099e-01 -1.84897631e-01 1.91860139e-01 -1.55708158e+00 5.53633451e-01 5.25098979e-01 -5.97820997e-01 3.22701365e-01 1.05545127e+00 -4.34347868e-01 -1.48403931e+00 -1.54527509e+00 2.86498755e-01 -5.50126806e-02 8.24601054e-01 1.24896519e-01 -6.11548960e-01 2.00675726e-01 4.14427370e-01 1.10501796e-01 1.58942625e-01 1.84572637e-02 5.65949023e-01 -2.99272567e-01 -1.08176863e+00 6.04890473e-02 5.23393631e-01 -2.04021201e-01 -1.16895311e-01 4.50342417e-01 2.92432606e-01 -7.78582275e-01 -1.37703836e+00 6.76380634e-01 3.38583916e-01 -9.52975869e-01 8.51653337e-01 -6.10689342e-01 9.30881739e-01 2.34373063e-02 6.13447726e-01 -1.79415381e+00 -3.88078749e-01 -6.12824321e-01 -2.71555156e-01 1.01064312e+00 5.39400242e-02 -3.66381317e-01 4.85836059e-01 2.73926258e-01 -6.94936872e-01 -1.17456281e+00 -1.08159375e+00 -8.60878766e-01 7.49504149e-01 -2.63904512e-01 7.75402606e-01 1.01372349e+00 -2.25620016e-01 -1.21588029e-01 1.82530612e-01 2.34255224e-01 4.03013779e-04 2.61853874e-01 6.61639810e-01 -1.38836586e+00 -4.92900200e-02 -6.18960977e-01 -2.77871966e-01 -3.45695496e-01 1.88837349e-01 -5.98580539e-01 7.72806779e-02 -1.41585755e+00 -5.88666916e-01 -6.06381595e-01 -3.77225637e-01 -3.35567035e-02 1.58140361e-01 -2.58325338e-01 1.13270812e-01 -3.66523653e-01 4.21645015e-01 3.78765881e-01 1.60304415e+00 -2.26052538e-01 5.14572449e-02 3.04748237e-01 -5.81662636e-04 6.34381890e-01 1.02481675e+00 -6.14981689e-02 -3.25656384e-01 -3.90884668e-01 7.63609335e-02 -1.24321818e-01 2.35439122e-01 -1.48264098e+00 -3.36009532e-01 5.44672948e-04 6.47002041e-01 -4.67821360e-01 -7.85212070e-02 -1.16236663e+00 2.93199480e-01 7.14831948e-01 -1.21854790e-01 2.21537307e-01 8.61729741e-01 2.92837262e-01 -3.64157826e-01 -3.42906475e-01 8.35101783e-01 -2.19647869e-01 -5.57626069e-01 3.16512704e-01 -4.47583616e-01 -3.04969668e-01 1.21418750e+00 -1.12738617e-01 -1.69119947e-02 7.37980008e-02 -5.14558256e-01 1.40377849e-01 3.87069844e-02 3.64078403e-01 2.74813399e-02 -7.49901652e-01 -4.59902823e-01 -9.44611430e-03 -6.63366914e-01 6.72694385e-01 6.25886321e-01 4.83761728e-01 -1.01123643e+00 6.85659528e-01 -4.57585782e-01 -3.61200660e-01 -7.85123944e-01 8.73468280e-01 8.97930861e-01 -4.68604952e-01 -4.78785008e-01 8.31270397e-01 2.12019105e-02 -3.83788019e-01 -3.97908807e-01 -7.72255123e-01 -2.14267731e-01 1.55622125e-01 3.52964811e-02 9.87652838e-01 9.47239459e-01 -4.47144538e-01 -5.69231361e-02 4.85705674e-01 5.80110788e-01 7.13443935e-01 1.61311758e+00 1.86561450e-01 1.85350746e-01 3.76887202e-01 1.31442213e+00 -1.57995433e-01 -1.32766378e+00 5.59992433e-01 -3.17703098e-01 -4.95420605e-01 7.97940910e-01 -9.45895791e-01 -1.75456870e+00 1.40822172e+00 6.92849994e-01 3.98530304e-01 1.14078832e+00 -6.75357759e-01 8.48279119e-01 3.78905296e-01 1.76797032e-01 -1.22055101e+00 -5.40185928e-01 6.02825046e-01 7.25985050e-01 -8.90428007e-01 3.80087018e-01 -2.06126526e-01 -2.03346685e-01 1.62909687e+00 5.65607250e-01 -1.91918463e-01 7.36297369e-01 7.55012929e-01 -5.85444011e-02 -2.61806995e-01 -4.21116441e-01 2.61856109e-01 -1.32093951e-01 4.58889812e-01 8.39207768e-01 2.16230839e-01 -1.20718494e-01 2.78728098e-01 -2.29164898e-01 4.62345080e-03 2.45939314e-01 8.45831513e-01 -1.48917452e-01 -8.73559475e-01 -2.72551924e-01 6.72843993e-01 -7.19155133e-01 2.11646393e-01 1.64111152e-01 1.21808338e+00 6.38962507e-01 7.78792143e-01 4.88815904e-01 -2.40481108e-01 7.58029461e-01 -1.87654212e-01 3.26562136e-01 -3.63739103e-01 -1.22929657e+00 -1.20872691e-01 2.30786949e-01 -4.38338816e-01 -1.75615057e-01 -7.85742581e-01 -1.31522226e+00 -4.82119262e-01 -4.98371333e-01 3.17558229e-01 9.00571167e-01 1.01125562e+00 1.49423242e-01 1.38518953e+00 8.82800758e-01 -1.25565588e+00 -8.17349732e-01 -1.04210055e+00 -9.13888633e-01 1.63619474e-01 3.41457069e-01 -1.03294444e+00 -4.57273751e-01 -2.10023180e-01]
[6.355125427246094, 3.174762010574341]
a3450dbe-d390-46ad-b3f0-28a5d38cf5fe
190409352
1904.09352
null
http://arxiv.org/abs/1904.09352v1
http://arxiv.org/pdf/1904.09352v1.pdf
Donkey and Smuggler Optimization Algorithm: A Collaborative Working Approach to Path Finding
Swarm Intelligence is a metaheuristic optimization approach that has become very predominant over the last few decades. These algorithms are inspired by animals' physical behaviors and their evolutionary perceptions. The simplicity of these algorithms allows researchers to simulate different natural phenomena to solve various real-world problems. This paper suggests a novel algorithm called Donkey and Smuggler Optimization Algorithm (DSO). The DSO is inspired by the searching behavior of donkeys. The algorithm imitates transportation behavior such as searching and selecting routes for movement by donkeys in the actual world. Two modes are established for implementing the search behavior and route-selection in this algorithm. These are the Smuggler and Donkeys. In the Smuggler mode, all the possible paths are discovered and the shortest path is then found. In the Donkeys mode, several donkey behaviors are utilized such as Run, Face & Suicide, and Face & Support. Real world data and applications are used to test the algorithm. The experimental results consisted of two parts, firstly, we used the standard benchmark test functions to evaluate the performance of the algorithm in respect to the most popular and the state of the art algorithms. Secondly, the DSO is adapted and implemented on three real-world applications namely; traveling salesman problem, packet routing, and ambulance routing. The experimental results of DSO on these real-world problems are very promising. The results exhibit that the suggested DSO is appropriate to tackle other unfamiliar search spaces and complex problems.
['Nawzad K. Al-Salihi', 'Rawan A. Al-Rashid Agha', 'Mokhtar Mohammadi', 'Ahmed S. Shamsaldin', 'Tarik A. Rashid']
2019-04-19
null
null
null
null
['metaheuristic-optimization']
['methodology']
[-3.29868972e-01 -5.44373989e-01 -1.97125375e-01 6.33206889e-02 4.77567464e-01 -2.86695629e-01 5.96436739e-01 1.36517406e-01 -7.52901793e-01 9.66810346e-01 -2.72185504e-01 -7.82461166e-02 -7.01425374e-01 -1.06157517e+00 -4.25791554e-02 -8.56943309e-01 -5.90596557e-01 7.08424509e-01 3.83187771e-01 -9.19178486e-01 8.52849185e-01 6.52022958e-01 -1.77916360e+00 -1.57812908e-01 1.17652130e+00 7.75453210e-01 4.97713000e-01 3.38103086e-01 -5.06627679e-01 4.25664663e-01 -8.99598181e-01 1.08983517e-01 3.77542198e-01 -5.63853800e-01 -6.03826523e-01 -8.75533223e-02 -1.16452873e+00 4.72376019e-01 1.45272002e-01 9.38155174e-01 5.45276403e-01 7.57007957e-01 4.40064341e-01 -1.75295520e+00 -4.97819453e-01 3.38871688e-01 -6.08354568e-01 6.03116274e-01 6.19229138e-01 1.10185020e-01 2.93855906e-01 -2.58695126e-01 6.05803132e-01 1.28414273e+00 5.18358290e-01 3.64798307e-01 -3.87322426e-01 -5.16746044e-01 -3.12753469e-02 5.74579358e-01 -1.39858520e+00 -8.09906572e-02 7.80343294e-01 1.41536951e-01 1.03760207e+00 5.37417352e-01 8.89139295e-01 4.34402615e-01 5.50728142e-01 6.80436373e-01 1.11438906e+00 -3.26206744e-01 5.62452555e-01 4.40894991e-01 4.91884984e-02 5.79424500e-01 4.14751947e-01 2.09419787e-01 -2.69079298e-01 -2.97209293e-01 1.72289625e-01 -1.41101539e-01 -2.09806725e-01 -1.06926702e-01 -1.13224423e+00 9.02941227e-01 4.95627820e-01 5.17820954e-01 -5.79529941e-01 -3.76375884e-01 4.22891498e-01 2.94361711e-01 1.34934485e-01 6.06566548e-01 -3.10231656e-01 -3.39277655e-01 -4.95188147e-01 2.71357179e-01 1.03179407e+00 6.05613112e-01 2.73499012e-01 -1.51121430e-02 4.87441957e-01 9.08744335e-01 5.29168427e-01 4.21729267e-01 8.51329327e-01 -2.88391620e-01 3.82470071e-01 7.70614445e-01 3.28064352e-01 -1.39119220e+00 -8.25029254e-01 -1.55059561e-01 -5.88559508e-01 3.33098501e-01 1.91200897e-01 -2.74766445e-01 -7.65407443e-01 1.05309999e+00 7.70691752e-01 1.29789367e-01 4.76675749e-01 7.87025154e-01 9.66950059e-01 1.14234996e+00 -1.84921384e-01 -5.28872967e-01 1.25084662e+00 -1.20681190e+00 -7.79207885e-01 1.48511186e-01 4.03545707e-01 -9.17901456e-01 6.33496642e-01 4.46982414e-01 -8.73431742e-01 -3.02428752e-01 -9.14860189e-01 8.33982825e-01 -9.19536293e-01 -2.93962091e-01 8.08519781e-01 8.26108634e-01 -9.88932133e-01 4.82036918e-01 -5.92096984e-01 -8.86868894e-01 1.51095450e-01 5.09954691e-01 -1.99333295e-01 2.41083726e-01 -1.13282299e+00 1.01742542e+00 6.68621361e-01 3.15058976e-01 -3.35820407e-01 2.10188702e-02 -4.84493941e-01 -1.19494647e-01 3.33152652e-01 -4.05030251e-01 6.05901241e-01 -9.37274635e-01 -1.56411004e+00 6.66880965e-01 -1.12402387e-01 -2.83029050e-01 4.91834521e-01 6.35048568e-01 -7.63453960e-01 3.85453030e-02 1.35413989e-01 7.60679767e-02 3.39732409e-01 -1.17234826e+00 -9.21741962e-01 -1.91682920e-01 -6.39606491e-02 4.90310609e-01 -3.52598689e-02 2.97893345e-01 -1.19688228e-01 -5.59618950e-01 1.87441140e-01 -8.86445403e-01 -3.59997869e-01 -3.84138256e-01 -3.17879260e-01 -3.82278144e-01 9.92859066e-01 -2.11902782e-02 1.29357529e+00 -1.83047783e+00 1.84953108e-01 6.62278235e-01 -4.46326226e-01 4.14972842e-01 -1.07577391e-01 1.01127565e+00 2.47248650e-01 1.93442732e-01 -2.60023415e-01 1.96183562e-01 -1.44647375e-01 3.57483327e-01 4.67578679e-01 4.75960881e-01 -4.09237027e-01 3.89106721e-01 -1.07315981e+00 -4.74360049e-01 1.64454162e-01 1.35781497e-01 -4.38972563e-01 5.73464446e-02 7.40569383e-02 5.62887430e-01 -9.62765396e-01 1.01864374e+00 9.09104645e-01 1.19638510e-01 -3.55117917e-02 1.50511891e-01 -6.18792713e-01 -1.97328508e-01 -1.48274183e+00 1.04090106e+00 -3.61959130e-01 3.73499781e-01 1.04134493e-01 -1.20036912e+00 9.83584881e-01 -4.18040343e-03 6.28184378e-01 -9.60821092e-01 5.31668425e-01 6.26625240e-01 2.94633567e-01 -1.21656489e+00 4.66449916e-01 9.03798267e-02 1.07701473e-01 5.78409195e-01 -5.69772243e-01 -1.57313496e-01 6.41125441e-01 -2.55014211e-01 8.19175065e-01 1.61466878e-02 4.85258728e-01 -6.57049537e-01 8.65157127e-01 6.21213078e-01 7.39481151e-01 4.34365958e-01 -3.67874742e-01 -9.28744823e-02 8.78325850e-02 -7.63245881e-01 -4.00062174e-01 -8.75959098e-01 4.54278488e-05 7.02491760e-01 9.70581055e-01 3.65540087e-02 -5.86395979e-01 -4.30127829e-01 -9.54639614e-02 8.32740128e-01 -6.14487171e-01 -1.66483819e-02 -6.94527268e-01 -1.27331161e+00 2.42675558e-01 -2.34010354e-01 1.23953795e+00 -1.68746364e+00 -1.03014028e+00 4.25801426e-01 9.43470630e-04 -6.50057256e-01 -6.64361268e-02 -2.03563899e-01 -7.65768230e-01 -1.22042418e+00 -3.98265004e-01 -9.15670991e-01 8.20336938e-01 5.26371896e-01 7.10937798e-01 7.89959550e-01 -3.84600759e-01 1.26974866e-01 -9.71400380e-01 -6.19236410e-01 -3.38029265e-01 -5.72305024e-02 2.20732585e-01 5.40305041e-02 4.33512628e-01 -6.50272548e-01 -5.31607091e-01 8.66947711e-01 -7.35530496e-01 -4.32317197e-01 3.74759555e-01 7.43019342e-01 2.19413117e-01 9.38257754e-01 5.80855966e-01 -4.62312311e-01 1.06311822e+00 -9.81188118e-01 -5.78266144e-01 3.22284728e-01 -4.01907027e-01 -3.45746763e-02 8.79735172e-01 -4.77849603e-01 -8.24501932e-01 -5.19703507e-01 -9.80015025e-02 3.30565602e-01 -5.94204366e-02 3.85488451e-01 -3.91025767e-02 -4.01166350e-01 2.67089576e-01 4.68363017e-01 2.05409303e-01 -3.44497234e-01 -4.03884321e-01 9.66872692e-01 -8.70911404e-02 -6.10170603e-01 7.05516040e-01 4.85041201e-01 5.79295307e-02 -1.02702963e+00 1.00850657e-01 -3.84347796e-01 1.55928219e-02 -3.80251408e-01 8.19803357e-01 1.22012421e-01 -1.30775547e+00 7.48515904e-01 -9.30589616e-01 6.35098070e-02 7.88817927e-02 5.69536388e-01 -2.79815167e-01 2.92127207e-02 1.84174627e-02 -1.11576235e+00 -3.14548373e-01 -1.46180630e+00 3.05321515e-01 9.07683194e-01 1.00326262e-01 -1.12147450e+00 -7.79534876e-02 1.00145534e-01 7.51597583e-01 5.15118599e-01 8.34818423e-01 -8.07124019e-01 -4.26700234e-01 -5.62686101e-02 2.75224715e-01 -3.11643332e-01 4.24196452e-01 2.68700331e-01 -1.57141075e-01 -1.37793750e-01 1.81187674e-01 2.88653255e-01 1.25109896e-01 2.80875534e-01 5.40205061e-01 -3.12377155e-01 -7.61118233e-01 5.94257772e-01 1.47865891e+00 1.41720164e+00 6.27151728e-01 1.26482081e+00 -1.63272738e-01 8.33621383e-01 1.09320998e+00 4.52184886e-01 4.83608812e-01 5.71412027e-01 6.77250981e-01 3.03057861e-02 5.41373849e-01 2.45410055e-02 8.03080499e-02 7.25750744e-01 -5.26734948e-01 -7.43059814e-01 -1.02495730e+00 4.21768546e-01 -1.78266418e+00 -1.20900798e+00 -1.75023869e-01 1.96788394e+00 -1.66815184e-02 1.08120635e-01 3.11099976e-01 3.10720414e-01 9.21421409e-01 -6.00622371e-02 -3.26823562e-01 -9.92056429e-01 -2.27682590e-02 8.67745653e-02 5.38877487e-01 2.50841975e-01 -7.83317149e-01 7.49723494e-01 5.70974827e+00 7.93939948e-01 -1.26019669e+00 2.32817158e-02 2.70317554e-01 1.23371191e-01 -8.42766091e-02 -4.35491763e-02 -5.20295322e-01 9.84778523e-01 5.41551352e-01 -4.33503807e-01 8.68026793e-01 4.83581334e-01 6.71140790e-01 -6.35938048e-01 -2.33085766e-01 9.78877187e-01 -3.22775096e-02 -1.23382866e+00 -1.87794611e-01 -1.28792137e-01 6.45875216e-01 -1.77852467e-01 -1.82499051e-01 4.08874340e-02 1.26436427e-01 -9.48369324e-01 5.41603744e-01 3.55045855e-01 -1.91949159e-01 -1.18451905e+00 8.96384716e-01 5.90140283e-01 -1.32734311e+00 -5.28902531e-01 -3.27385426e-01 -4.43291888e-02 4.65411723e-01 -5.89599237e-02 -5.44614136e-01 9.61516023e-01 1.07883036e+00 3.41325939e-01 -2.70688236e-01 1.68761611e+00 1.68776929e-01 5.13706617e-02 -7.93684244e-01 -1.10044658e+00 7.60732472e-01 -8.73991489e-01 1.05536401e+00 6.53428793e-01 4.64323670e-01 2.79976666e-01 -2.78258659e-02 4.26450402e-01 4.88187909e-01 5.28025091e-01 -6.04577243e-01 1.33592933e-01 8.61231208e-01 1.07390225e+00 -1.48087800e+00 -4.07031924e-02 1.66450426e-01 6.23227239e-01 -3.43472749e-01 1.72349319e-01 -1.14796782e+00 -8.47309649e-01 5.76330960e-01 -1.76203866e-02 8.87638852e-02 -9.46988016e-02 -2.77443696e-02 -6.33701682e-01 -1.97907731e-01 -9.14603114e-01 4.59074616e-01 -5.10190666e-01 -9.06841278e-01 1.01731980e+00 2.91253239e-01 -1.46676767e+00 2.55770832e-01 -4.21771705e-01 -1.16464472e+00 4.05876786e-01 -1.48710597e+00 -5.93109846e-01 -3.37863535e-01 5.18068969e-01 8.48400533e-01 -6.68268740e-01 4.70500916e-01 2.95228213e-01 -7.92197406e-01 1.86858490e-01 3.75478417e-01 -3.99691522e-01 -9.32979360e-02 -5.73010266e-01 7.85748959e-02 5.38922310e-01 -4.36023653e-01 7.89109111e-01 8.84622276e-01 -5.30621052e-01 -1.47952342e+00 -3.04387033e-01 5.38367629e-01 1.67927429e-01 4.86065388e-01 3.87911238e-02 -2.97829658e-01 8.83950144e-02 3.40318948e-01 -2.26618946e-01 4.74440962e-01 -4.55560207e-01 7.79172242e-01 -8.60306770e-02 -1.87900281e+00 5.57635248e-01 9.61249590e-01 6.87529981e-01 -5.50172746e-01 3.96006972e-01 1.37846947e-01 -3.41005325e-01 -3.40126514e-01 2.22553268e-01 4.33254808e-01 -1.16783333e+00 9.00863469e-01 -5.60403287e-01 -1.33463740e-01 -6.23290420e-01 1.59354404e-01 -1.52578151e+00 8.36358368e-02 -9.54902828e-01 5.44416904e-01 1.13457966e+00 4.04422194e-01 -1.41260254e+00 5.70714056e-01 2.32922927e-01 -5.07155471e-02 -1.10390651e+00 -1.09564579e+00 -9.43151712e-01 -3.56084883e-01 1.45927042e-01 1.30231118e+00 1.04886711e+00 -1.59811407e-01 -2.87936032e-01 -1.23861291e-01 5.09233139e-02 6.10149920e-01 1.12449750e-01 6.26240551e-01 -1.08695102e+00 -7.81645179e-02 -4.31315988e-01 -5.90621412e-01 -5.82363904e-01 -1.77514568e-01 -5.83382607e-01 -1.35703206e-01 -1.67278206e+00 -4.87376153e-01 -1.07528317e+00 -7.80487061e-02 6.97150379e-02 1.58497021e-01 -1.09170891e-01 2.26506487e-01 1.51267350e-01 -3.15783381e-01 5.48860848e-01 1.34267688e+00 -4.67292555e-02 -7.27577448e-01 4.10720944e-01 -4.39093053e-01 8.87565613e-01 1.12205780e+00 -6.90414786e-01 -5.26730299e-01 -3.48190963e-01 9.64436978e-02 1.94338746e-02 -2.26485908e-01 -9.68549252e-01 5.12983441e-01 -6.19036019e-01 -1.15794972e-01 -4.93005693e-01 2.26871043e-01 -1.14313030e+00 3.48197192e-01 9.75971222e-01 4.31670636e-01 7.28906155e-01 1.15333028e-01 3.76168132e-01 -1.14076480e-01 -5.66138446e-01 8.88280749e-01 -3.62249345e-01 -1.10306144e+00 -9.50315371e-02 -5.19504964e-01 3.32676023e-02 1.88086367e+00 -9.49304342e-01 -3.24862748e-01 -1.56642899e-01 -4.96989191e-01 7.55835176e-01 4.29541200e-01 4.97169256e-01 7.64275312e-01 -1.10356081e+00 -4.47542876e-01 1.48770958e-01 -2.24456936e-01 -3.41139793e-01 1.14499368e-01 1.03911245e+00 -1.39736807e+00 8.08603223e-03 -7.23990500e-01 -3.19600254e-01 -1.28167415e+00 5.60110807e-01 5.59247196e-01 -2.57820189e-02 -3.82031739e-01 7.95265198e-01 -6.35867000e-01 -6.25615001e-01 -9.08073932e-02 4.28974628e-03 -8.07177782e-01 3.47030982e-02 3.65580708e-01 9.01748002e-01 -2.53289163e-01 -8.28312516e-01 -8.75213623e-01 8.10724020e-01 2.91869015e-01 7.96495080e-02 1.64845145e+00 -1.47829324e-01 -5.57751834e-01 -8.94669667e-02 9.46776927e-01 9.02588740e-02 -2.30034232e-01 4.61824805e-01 5.52746654e-02 -7.41580606e-01 -3.61575991e-01 -6.83678210e-01 -1.07456195e+00 1.92384377e-01 5.81957161e-01 7.12563574e-01 1.37285709e+00 -5.87739527e-01 9.88651514e-01 5.07222950e-01 9.59438205e-01 -1.35694778e+00 -3.67767334e-01 3.77565295e-01 6.08800411e-01 -1.11004901e+00 7.47283502e-03 -3.48754197e-01 -6.86529934e-01 1.14649892e+00 5.95124483e-01 -2.85390109e-01 7.45338082e-01 2.89760262e-01 -1.50857652e-02 -2.95891404e-01 -5.07107437e-01 -9.61549580e-02 -2.19166651e-01 7.15399444e-01 -1.52263686e-01 -1.62271783e-01 -1.08354461e+00 2.12464601e-01 -2.98724532e-01 -5.86414523e-02 6.08435512e-01 1.34493327e+00 -7.39523411e-01 -1.12516487e+00 -8.55453968e-01 2.22789526e-01 -2.47075528e-01 2.74001211e-01 -2.08680928e-01 1.19510877e+00 4.49588001e-01 1.48679042e+00 -6.79873899e-02 -3.10443163e-01 4.55207497e-01 -6.19364381e-01 9.49299037e-02 -2.26737589e-01 -9.72481251e-01 -5.57195187e-01 1.22934982e-01 -4.91416335e-01 -7.79878974e-01 -4.19133455e-01 -1.64774990e+00 -6.66463256e-01 -3.91333699e-01 1.16366458e+00 8.82762492e-01 9.01937723e-01 1.54910266e-01 3.49131614e-01 1.03426266e+00 -8.60136390e-01 -2.57815961e-02 -4.45571840e-01 -4.90124881e-01 8.59764740e-02 5.43578789e-02 -1.13447642e+00 -4.41923738e-01 -5.62413335e-01]
[5.653913974761963, 3.498018741607666]
7cb2bd0d-b6d9-4478-bb02-165b72ade36a
robust-point-set-registration-using-gaussian
null
null
https://ieeexplore.ieee.org/document/5674050
https://github.com/bing-jian/gmmreg/blob/master/gmmreg_PAMI_preprint.pdf
Robust Point Set Registration Using Gaussian Mixture Models
In this paper, we present a unified framework for the rigid and nonrigid point set registration problem in the presence of significant amounts of noise and outliers. The key idea of this registration framework is to represent the input point sets using Gaussian mixture models. Then, the problem of point set registration is reformulated as the problem of aligning two Gaussian mixtures such that a statistical discrepancy measure between the two corresponding mixtures is minimized. We show that the popular iterative closest point (ICP) method and several existing point set registration methods in the field are closely related and can be reinterpreted meaningfully in our general framework. Our instantiation of this general framework is based on the the L2 distance between two Gaussian mixtures, which has the closed-form expression and in turn leads to a computationally efficient registration algorithm. The resulting registration algorithm exhibits inherent statistical robustness, has an intuitive interpretation, and is simple to implement. We also provide theoretical and experimental comparisons with other robust methods for point set registration.
['Bing Jian', 'Baba C. Vemuri']
2010-12-23
null
null
null
ieee-transactions-on-pattern-analysis-and-9
['3d-point-cloud-matching']
['computer-vision']
[ 4.29710038e-02 5.19781597e-02 -5.10462262e-02 -2.86512017e-01 -8.58023465e-01 -5.51678300e-01 8.06750298e-01 4.43890505e-02 -3.09657156e-01 1.97207525e-01 -6.98224008e-02 8.43890235e-02 -4.52254087e-01 -4.50573772e-01 -4.23130453e-01 -8.76404047e-01 4.03701924e-02 8.65442455e-01 1.65586218e-01 -2.23044813e-01 3.00155938e-01 9.31885302e-01 -1.21767652e+00 -8.18887293e-01 9.20006931e-01 5.55535495e-01 1.39319412e-02 2.42553666e-01 2.44597778e-01 -1.50036156e-01 -1.68950617e-01 -2.73703903e-01 5.72156250e-01 -1.71496078e-01 -1.14242542e+00 3.03903013e-01 6.12026393e-01 9.01953876e-02 -1.43894508e-01 1.15905249e+00 4.75737602e-01 6.89634860e-01 8.70541096e-01 -1.37383783e+00 -3.22307199e-01 3.86081007e-03 -7.56806791e-01 -2.16823652e-01 4.43072855e-01 -5.23539066e-01 5.39963245e-01 -9.03984785e-01 4.37955618e-01 1.24406493e+00 8.65428865e-01 3.79705906e-01 -1.55533147e+00 -1.89783573e-01 -2.77188897e-01 -2.41488561e-01 -1.75025403e+00 -4.90344524e-01 6.95694685e-01 -5.27091265e-01 4.09496039e-01 6.26544118e-01 4.11764890e-01 3.86134595e-01 1.43175244e-01 1.99248433e-01 8.82574201e-01 -6.23940825e-01 2.11732849e-01 -2.83840001e-01 2.94662446e-01 4.47026998e-01 4.54226166e-01 6.56069815e-02 -1.26733925e-04 -8.34030211e-01 1.15853274e+00 3.48810822e-01 -3.32341343e-01 -9.29621994e-01 -1.44038332e+00 8.03181946e-01 2.54298270e-01 6.42220497e-01 -3.20872903e-01 8.90985355e-02 -8.48473758e-02 -1.40368313e-01 4.67168242e-01 1.14884369e-01 1.30844265e-01 1.72645763e-01 -9.00406122e-01 3.58314395e-01 7.90239871e-01 8.97674918e-01 7.40310133e-01 -3.49846691e-01 4.26119655e-01 8.75064552e-01 9.59366202e-01 6.55313134e-01 3.45143855e-01 -1.35411429e+00 6.14323504e-02 1.96872175e-01 9.76981744e-02 -1.15645373e+00 -3.15379143e-01 1.14086673e-01 -9.15443361e-01 2.66477078e-01 3.25028092e-01 3.43580544e-01 -5.05734682e-01 1.61302555e+00 6.07538879e-01 5.93312442e-01 -5.50142527e-02 8.11116219e-01 5.69690824e-01 3.42228711e-01 -2.09599718e-01 -3.82847637e-01 1.35920382e+00 -4.13895339e-01 -7.88705468e-01 1.72893643e-01 2.06770480e-01 -1.15513992e+00 5.35287559e-01 -5.18248342e-02 -1.17499101e+00 -4.80724812e-01 -1.02702403e+00 -1.51898965e-01 2.47148469e-01 -4.29591507e-01 3.31804454e-01 5.16886473e-01 -1.36022902e+00 8.21576536e-01 -1.26309586e+00 -6.88881457e-01 -9.20324773e-02 7.44099021e-01 -5.99719644e-01 1.86109915e-01 -4.31052148e-01 1.03258550e+00 -2.45818938e-03 2.62990355e-01 2.78890878e-02 -4.32335287e-01 -9.94385242e-01 -4.63399261e-01 -2.61021405e-02 -9.24090922e-01 1.20895433e+00 -4.69260007e-01 -1.56538403e+00 1.21027100e+00 -5.54503679e-01 2.49799285e-02 5.37423790e-01 -8.58572945e-02 -1.72000825e-01 -3.86026502e-02 2.08461046e-01 3.21472615e-01 7.99241126e-01 -1.58670366e+00 -1.03251137e-01 -6.65895998e-01 -5.43260157e-01 2.94903308e-01 1.35882586e-01 4.38008636e-01 -5.91657639e-01 -7.35709906e-01 7.89621055e-01 -1.31194913e+00 -6.80181623e-01 7.06828292e-03 -3.35900754e-01 -2.47144774e-01 5.95083237e-01 -4.67419237e-01 8.45383823e-01 -2.12257743e+00 3.61097157e-01 8.44298244e-01 3.79337877e-01 -2.91942596e-01 1.84966266e-01 2.69759297e-01 -4.35899377e-01 -2.71183718e-02 -6.25428498e-01 -5.64962089e-01 -1.41338864e-02 3.05098742e-01 -1.80082396e-01 1.36640871e+00 -4.71299618e-01 5.66262603e-01 -8.15577805e-01 -5.81343412e-01 4.42390323e-01 6.90264642e-01 -2.52511293e-01 4.78167385e-02 7.34856486e-01 6.67386115e-01 -3.63945931e-01 3.12341452e-01 9.70670938e-01 1.14047959e-01 -7.12130172e-03 -3.95392686e-01 -1.38015538e-01 -1.27014443e-01 -1.64048994e+00 1.70013738e+00 1.19302072e-01 2.66432345e-01 2.89784849e-01 -8.76614749e-01 1.11873126e+00 5.77959359e-01 1.23412347e+00 1.69315875e-01 2.31418759e-01 3.97908598e-01 -2.68917263e-01 8.66096765e-02 5.74579597e-01 -5.10649860e-01 -1.58268720e-01 5.14660239e-01 -5.64544797e-02 -4.63620901e-01 -2.08257660e-01 -1.87312901e-01 5.65488517e-01 -4.12072651e-02 7.62901843e-01 -5.82744956e-01 5.79193532e-01 -1.90018728e-01 5.17775178e-01 4.72945392e-01 -1.99818790e-01 8.64889264e-01 -4.02779788e-01 -2.25701421e-01 -8.77196789e-01 -1.33038425e+00 -5.39921045e-01 3.28426629e-01 5.27364016e-01 -2.49874830e-01 -9.19818163e-01 -1.14788145e-01 4.73657995e-02 1.91152200e-01 -2.91070789e-01 -3.55965458e-02 -9.02281344e-01 -6.71517909e-01 2.13230342e-01 3.17430526e-01 7.66151398e-02 -5.41082621e-01 -1.39617056e-01 8.31186324e-02 -2.71730930e-01 -9.16215062e-01 -8.61651242e-01 -3.53767931e-01 -1.43779290e+00 -1.12451160e+00 -9.02412117e-01 -9.44998384e-01 9.19545829e-01 6.22035027e-01 9.43724573e-01 1.45139873e-01 9.26383361e-02 8.00030470e-01 4.60222140e-02 -1.62353694e-01 -5.91204643e-01 -3.94172132e-01 8.15498710e-01 1.20097682e-01 3.34184676e-01 -7.59353638e-01 -3.49473387e-01 7.75950730e-01 -9.30339336e-01 -5.31306863e-01 -2.37179454e-02 3.59436035e-01 1.01283514e+00 -8.64811316e-02 2.84100249e-02 -4.06491131e-01 5.62120378e-01 -2.35448718e-01 -7.19553471e-01 6.44264966e-02 -4.10769373e-01 9.08241719e-02 7.75775779e-03 -2.83351481e-01 -7.42152691e-01 2.83472091e-01 -2.52470635e-02 -4.47825462e-01 -3.14836979e-01 1.31832613e-02 -4.87095386e-01 -7.58465230e-01 4.37212318e-01 -9.74728167e-03 5.14501512e-01 -8.16888690e-01 7.02484429e-01 6.38625026e-01 1.01440406e+00 -9.01976168e-01 1.25641060e+00 1.01516795e+00 2.45859399e-01 -9.26897943e-01 -9.20241177e-02 -9.17485714e-01 -1.29251420e+00 -4.14890703e-03 6.94357216e-01 -5.60210049e-01 -8.36545646e-01 3.78400385e-01 -1.26266766e+00 1.48907095e-01 -4.08949107e-01 7.00076163e-01 -1.18630791e+00 1.09510767e+00 -3.45408827e-01 -5.85357726e-01 -2.28265360e-01 -1.48622322e+00 1.12324190e+00 7.81780332e-02 -4.73437369e-01 -1.34964657e+00 4.94964987e-01 9.01024565e-02 -8.23753625e-02 5.82695305e-01 5.41545510e-01 -9.14901137e-01 -1.66174486e-01 -4.72222537e-01 2.31585875e-01 1.25100538e-01 4.30975258e-01 2.73350533e-02 -6.91704750e-01 -3.59868973e-01 7.42288351e-01 7.96405971e-01 1.41804948e-01 6.13427043e-01 5.77752233e-01 -3.09419543e-01 -6.79764032e-01 6.71902359e-01 1.38064063e+00 6.03321642e-02 5.55995345e-01 2.91169375e-01 8.00471842e-01 5.34419954e-01 4.89931911e-01 -4.24776552e-03 3.09948236e-01 1.18195331e+00 1.17397659e-01 -1.30287319e-01 3.11997771e-01 1.63951486e-01 9.33783203e-02 1.15455139e+00 -5.88360012e-01 3.98732990e-01 -9.39595997e-01 3.53504419e-01 -1.97840106e+00 -1.03492630e+00 -5.32395840e-01 2.76564360e+00 6.60171807e-01 -5.67014992e-01 3.37143451e-01 3.70531715e-02 1.08842468e+00 -1.51317403e-01 1.19522117e-01 -8.41344744e-02 -1.51470155e-02 1.88160151e-01 6.21187449e-01 7.39638329e-01 -1.30061901e+00 5.30986309e-01 7.50863743e+00 6.16320610e-01 -6.43340766e-01 2.35985458e-01 3.35463360e-02 4.34837162e-01 -1.02165431e-01 7.79703483e-02 -5.81914783e-01 3.00093204e-01 7.00243413e-01 -4.72218633e-01 2.27812335e-01 6.29748344e-01 2.56141454e-01 1.86723262e-01 -1.32191062e+00 1.24944806e+00 1.58154011e-01 -1.05973446e+00 -2.04224467e-01 5.59866607e-01 6.16096199e-01 3.24961096e-02 6.97534904e-02 -5.28281450e-01 2.91246623e-01 -8.31137896e-01 5.53044617e-01 6.29829347e-01 6.53090954e-01 -5.02037525e-01 5.52961111e-01 2.44951725e-01 -1.28749132e+00 5.33025324e-01 -5.08905649e-01 3.49152923e-01 5.62944889e-01 3.85898232e-01 -3.85449409e-01 8.09856117e-01 4.26954299e-01 6.44903660e-01 -4.59796190e-02 1.57935107e+00 2.84252524e-01 -4.56063561e-02 -7.59535968e-01 6.26490414e-01 -2.38684416e-01 -9.48468447e-01 9.97876406e-01 8.01526308e-01 6.50542617e-01 3.61767560e-01 2.99591988e-01 7.44284868e-01 1.86331972e-01 2.84521997e-01 -6.29408002e-01 7.30780184e-01 5.07619143e-01 1.18117583e+00 -8.37164998e-01 -8.12237412e-02 -4.24802840e-01 9.36803102e-01 -7.21288612e-03 2.54055977e-01 -5.19863725e-01 -5.50834872e-02 9.59542334e-01 1.63091987e-01 -1.67374566e-01 -4.90829110e-01 -3.25552434e-01 -1.13616312e+00 3.86507586e-02 -6.20535016e-01 2.96632886e-01 -4.05724794e-01 -1.50295246e+00 5.98994017e-01 5.32725930e-01 -1.56485045e+00 -2.69776493e-01 -5.31264424e-01 -7.12645710e-01 1.06758392e+00 -9.48383689e-01 -9.36459541e-01 -1.37391299e-01 8.30672145e-01 3.87952477e-02 2.01422587e-01 9.72636044e-01 2.00716719e-01 -2.63565600e-01 3.98772895e-01 4.46247637e-01 1.88306458e-02 7.35668302e-01 -1.36764383e+00 4.83534068e-01 8.85525465e-01 8.80492404e-02 1.11471391e+00 8.26296747e-01 -6.18127823e-01 -1.32840300e+00 -8.42354178e-01 7.95606673e-01 -7.60808468e-01 5.61282933e-01 4.61780094e-02 -1.06527042e+00 8.71907532e-01 -2.03210458e-01 1.02688566e-01 8.84475172e-01 -5.73479533e-02 -8.75254944e-02 1.55619577e-01 -1.40639639e+00 5.48105180e-01 8.80548656e-01 -3.26582372e-01 -1.01401711e+00 5.13253093e-01 3.02773148e-01 -4.83744621e-01 -1.40655982e+00 3.95771921e-01 2.54919618e-01 -5.67423522e-01 1.47759974e+00 -2.63875276e-01 -6.74925923e-01 -5.07278383e-01 -3.41633111e-01 -1.09510052e+00 -4.90947038e-01 -1.20500219e+00 8.49356800e-02 1.18950355e+00 -2.35029921e-01 -9.30521309e-01 4.59944904e-01 9.09956038e-01 -6.57855645e-02 -3.07666630e-01 -1.30608642e+00 -9.87286806e-01 1.86954573e-01 -3.08027864e-01 5.93731940e-01 1.04810762e+00 2.04426795e-01 -1.25544429e-01 -1.25413701e-01 4.83469814e-01 1.15514946e+00 8.11781436e-02 8.25390220e-01 -1.86410093e+00 7.22699538e-02 -5.42314947e-01 -7.71984041e-01 -1.09000540e+00 3.76386017e-01 -9.93950248e-01 2.65859038e-01 -1.38195336e+00 1.47252962e-01 -8.31658006e-01 1.32219687e-01 1.60584018e-01 -2.29976997e-01 2.38355264e-01 1.77760884e-01 1.00984192e+00 -2.31349796e-01 3.10302943e-01 8.31034720e-01 2.68653482e-01 -3.36502165e-01 4.58205342e-01 -6.21184707e-01 1.02793682e+00 4.75328863e-01 -4.21257645e-01 9.90451947e-02 -8.12462717e-02 -3.37136596e-01 -7.24881515e-02 4.41949636e-01 -7.84823954e-01 3.34479600e-01 -1.73126817e-01 -1.02963127e-01 -4.69452977e-01 4.35170442e-01 -1.23556769e+00 7.16706395e-01 1.98006406e-01 3.12512189e-01 2.68184155e-01 -1.14448354e-01 4.51169401e-01 -1.55922323e-01 -3.77877355e-01 9.69119072e-01 1.11605093e-01 -3.32540870e-01 5.41248024e-01 1.21038398e-02 -2.82131493e-01 9.81138587e-01 -6.32017672e-01 1.83605939e-01 -2.64485747e-01 -8.71319115e-01 -2.66129106e-01 1.09868681e+00 2.79754639e-01 5.00539184e-01 -1.83486521e+00 -6.86739683e-01 2.60619730e-01 -5.94099835e-02 2.09529072e-01 -4.67999205e-02 1.30612302e+00 -5.78418314e-01 2.69481838e-01 1.03692576e-01 -1.00619149e+00 -1.42656219e+00 5.69545388e-01 6.51576817e-01 2.03322440e-01 -6.35107577e-01 4.63343918e-01 1.81861490e-01 -6.84348345e-01 -1.63546979e-01 -2.87295103e-01 7.78824314e-02 -4.00486350e-01 4.27305967e-01 6.79570615e-01 6.06200770e-02 -1.64028251e+00 -5.85830629e-01 1.35583961e+00 3.63929749e-01 -2.65240401e-01 1.23272371e+00 -3.07273120e-01 -7.65554249e-01 2.66244203e-01 1.40903437e+00 2.73501039e-01 -8.56660247e-01 -4.18742090e-01 1.76387861e-01 -5.05738437e-01 -2.19732732e-01 3.74193430e-01 -8.29893887e-01 3.02587450e-01 6.10173285e-01 1.56962082e-01 8.17163050e-01 3.92055631e-01 5.38124323e-01 1.20758586e-01 4.89998609e-01 -6.58629000e-01 -6.27207577e-01 3.30436349e-01 8.59865963e-01 -9.31371510e-01 2.61639476e-01 -9.86499965e-01 -6.35200813e-02 1.04107070e+00 -4.92858514e-02 -5.29979289e-01 8.78690004e-01 2.53803164e-01 1.06140330e-01 -2.21642613e-01 2.91545719e-01 7.22572654e-02 8.18832099e-01 9.58743453e-01 4.03953463e-01 5.27558289e-02 -4.38379377e-01 4.99239385e-01 -3.58848006e-01 -3.01613957e-01 2.39547536e-01 8.76808763e-01 -4.72061545e-01 -1.39684331e+00 -1.10097277e+00 -7.19287172e-02 -3.07758749e-01 2.91695863e-01 -1.14490576e-01 7.38120079e-01 -2.74524540e-01 8.23205948e-01 2.63038218e-01 2.70961802e-02 4.45095211e-01 -1.35031909e-01 6.27861500e-01 -5.21294534e-01 -2.48939395e-01 6.55232668e-01 -5.52826464e-01 -6.68361604e-01 -9.12666678e-01 -9.50619161e-01 -1.34242666e+00 -4.51412171e-01 -5.00024438e-01 4.62746441e-01 6.76435888e-01 9.20996606e-01 2.22389221e-01 -2.20458105e-01 6.22321725e-01 -1.44588459e+00 -6.82037652e-01 -7.46578395e-01 -6.80432081e-01 5.68630815e-01 2.64038175e-01 -8.50891113e-01 -2.95591503e-01 2.06125274e-01]
[7.741795063018799, -2.855252742767334]
a5c13b9e-34e1-4a4e-af0f-73604b5b495f
concept-to-text-generation-via-discriminative
null
null
https://aclanthology.org/P12-1039
https://aclanthology.org/P12-1039.pdf
Concept-to-text Generation via Discriminative Reranking
null
['Ioannis Konstas', 'Mirella Lapata']
2012-07-01
null
null
null
acl-2012-7
['concept-to-text-generation']
['natural-language-processing']
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
[-7.408435344696045, 3.8702902793884277]
d5ec9c19-2c7d-48c4-8d24-62d0698ba325
fast-real-time-counterfactual-explanations
2007.05684
null
https://arxiv.org/abs/2007.05684v2
https://arxiv.org/pdf/2007.05684v2.pdf
Fast Real-time Counterfactual Explanations
Counterfactual explanations are considered, which is to answer {\it why the prediction is class A but not B.} Different from previous optimization based methods, an optimization-free Fast ReAl-time Counterfactual Explanation (FRACE) algorithm is proposed benefiting from the development of multi-domain image to image translation algorithms. Built from starGAN, a transformer is trained as a residual generator conditional on a classifier constrained under a proposal perturbation loss which maintains the content information of the query image, but just the class-specific semantic information is changed. The transformer can transfer the query image to any counterfactual class, and during inference, our explanation can be generated by it only within a forward time. It is fast and can satisfy the real-time practical application. Because of the adversarial training of GAN, our explanation is also more realistic compared to other counterparts. The experimental results demonstrate that our proposal is better than the existing state of the art in terms of quality and speed.
['Yunxia Zhao']
2020-07-11
null
null
null
null
['counterfactual-explanation']
['miscellaneous']
[ 7.58771360e-01 8.74888062e-01 -3.49929929e-01 -2.65972435e-01 -7.32671022e-01 -5.52812099e-01 8.48365724e-01 -6.02116704e-01 -2.69528721e-02 1.17442203e+00 1.00896634e-01 -4.51171845e-01 2.27073133e-01 -9.02042687e-01 -1.17868400e+00 -6.87808096e-01 3.69443655e-01 6.07693017e-01 -1.48158208e-01 -6.29199371e-02 1.26232713e-01 4.12061028e-02 -1.45184338e+00 3.72216374e-01 1.03995895e+00 1.14059544e+00 1.30816013e-01 3.26529413e-01 -1.40719954e-02 8.70082617e-01 -5.18201292e-01 -9.74524736e-01 4.15714830e-01 -1.10759842e+00 -7.97007978e-01 1.40627936e-01 3.94180298e-01 -3.41866732e-01 -2.51149654e-01 1.27253652e+00 3.16295356e-01 -8.75633731e-02 5.87221384e-01 -1.70843327e+00 -1.08309710e+00 5.59451818e-01 -2.95305580e-01 -3.86998445e-01 2.26233214e-01 3.88732225e-01 8.31208348e-01 -6.34564817e-01 8.63697767e-01 1.28709376e+00 2.10085571e-01 1.07597852e+00 -1.04684174e+00 -9.19924617e-01 2.81038880e-01 5.10045409e-01 -9.20662642e-01 -2.27519602e-01 9.69727576e-01 -1.13911636e-01 3.42080921e-01 5.71229815e-01 8.14629316e-01 1.37366438e+00 2.64108479e-01 8.96236956e-01 1.38696539e+00 -3.40735137e-01 3.31333727e-01 4.46178585e-01 -7.39168108e-01 7.60181665e-01 4.45616245e-02 4.38509136e-01 -3.62453282e-01 8.80358815e-02 7.04033017e-01 -1.20490091e-02 -5.54928064e-01 -6.11467659e-01 -1.24913633e+00 1.05548084e+00 7.16259122e-01 -4.79476601e-02 -3.91974360e-01 3.68871570e-01 5.27587868e-02 3.64108831e-01 4.23174530e-01 3.87248486e-01 -3.63623619e-01 1.99907660e-01 -8.39550376e-01 3.97442162e-01 5.55437922e-01 1.11735451e+00 6.89753950e-01 3.74949306e-01 -4.25617784e-01 1.65370852e-01 8.50579143e-02 7.47788370e-01 7.48772085e-01 -1.13745129e+00 4.51345235e-01 4.75663543e-01 1.16406269e-01 -7.49611080e-01 2.28014424e-01 -5.74259937e-01 -9.92949843e-01 3.52861345e-01 2.26058170e-01 -1.83785707e-01 -9.18502510e-01 2.02742076e+00 4.44932252e-01 5.67526340e-01 3.04496735e-01 1.07293868e+00 5.61400056e-01 5.67895412e-01 -2.17580840e-01 -5.54154038e-01 1.19171774e+00 -1.03051829e+00 -7.59436667e-01 -3.06068331e-01 1.99466228e-01 -6.19135737e-01 1.13938189e+00 1.44812465e-01 -9.22846675e-01 -5.22367895e-01 -1.11222363e+00 1.10979415e-01 -2.74140626e-01 -2.18139902e-01 8.33503902e-01 8.85980368e-01 -8.25635433e-01 3.62420857e-01 -1.82071224e-01 1.27586097e-01 7.21906126e-01 1.34799436e-01 -3.51704746e-01 -1.13353692e-01 -1.36865616e+00 8.87988985e-01 4.74150419e-01 -1.41702369e-01 -1.15931070e+00 -8.49351585e-01 -9.33398366e-01 1.10261112e-01 4.72456157e-01 -1.23117995e+00 1.22834122e+00 -1.89176977e+00 -1.74502420e+00 8.84210706e-01 -7.87520856e-02 -8.74221861e-01 1.10727465e+00 1.23263106e-01 -3.90919596e-01 9.63896737e-02 3.65652382e-01 8.49301279e-01 1.23320413e+00 -1.36964595e+00 -4.95378435e-01 -3.52098882e-01 1.97684035e-01 2.26939231e-01 1.92382321e-01 -3.80401373e-01 -1.83079362e-01 -6.77428126e-01 -9.53695104e-02 -9.64662194e-01 -1.40765250e-01 1.88179433e-01 -5.78926027e-01 3.12191218e-01 1.06140459e+00 -6.30317092e-01 6.40101433e-01 -2.08061099e+00 -1.39388397e-01 -6.92505948e-03 -9.33821574e-02 5.04877232e-02 -2.39847135e-02 -3.00882421e-02 -4.16194141e-01 2.65074462e-01 -6.85568511e-01 -1.10697992e-01 1.23190552e-01 1.44397289e-01 -8.94973516e-01 4.73274857e-01 1.90645069e-01 1.23834503e+00 -9.00038660e-01 -6.36889696e-01 1.64584368e-01 2.10709706e-01 -6.09109640e-01 5.11291683e-01 -5.37569880e-01 7.10823894e-01 -3.16153437e-01 3.09023052e-01 9.63991642e-01 -2.00029865e-01 1.37918279e-01 4.64350432e-02 2.93787718e-01 9.87835135e-03 -9.30828452e-01 1.85310459e+00 -5.68132341e-01 4.56908792e-01 -3.37276220e-01 -1.28199756e+00 7.66957641e-01 5.04509568e-01 6.47237077e-02 -7.21259177e-01 5.60293347e-03 2.78523445e-01 -6.01565987e-02 -2.81259477e-01 1.50618121e-01 -5.74705780e-01 -1.55787477e-02 3.64366651e-01 -9.11468640e-02 -5.61911643e-01 -4.03956950e-01 9.04923156e-02 6.27405465e-01 4.28631693e-01 4.51839894e-01 -6.49698153e-02 5.79720914e-01 1.27862036e-01 8.44863057e-01 6.84628904e-01 5.07142730e-02 8.54500949e-01 4.61141765e-01 -3.20031881e-01 -1.05789387e+00 -1.11727250e+00 2.28839591e-01 9.59000513e-02 4.02575463e-01 2.57734209e-01 -9.51704562e-01 -1.24222684e+00 -1.92952693e-01 1.30529237e+00 -6.97893143e-01 -4.25837219e-01 -3.93229812e-01 -5.06594896e-01 2.18326047e-01 9.09547284e-02 1.17347944e+00 -1.14098263e+00 -4.99602318e-01 -1.38389890e-03 -4.28590208e-01 -1.02583921e+00 -6.66340590e-01 -4.71043587e-01 -8.64359498e-01 -1.19161499e+00 -6.38208389e-01 -5.77448487e-01 7.35917449e-01 1.25724509e-01 1.03398514e+00 7.23094642e-02 1.46896347e-01 8.32567438e-02 -2.09668666e-01 -7.40198016e-01 -6.60850406e-01 -5.44084370e-01 -2.67091841e-01 5.32818198e-01 -1.97048277e-01 -6.00721061e-01 -8.61232460e-01 1.27933785e-01 -1.01519859e+00 5.33922255e-01 6.84859514e-01 1.18266368e+00 7.52136409e-01 1.29191905e-01 6.76883101e-01 -1.33418250e+00 2.02384457e-01 -4.48219866e-01 -7.03169942e-01 3.74861032e-01 -9.77264404e-01 2.04515502e-01 1.02519715e+00 -4.45838809e-01 -1.58744478e+00 -7.97505211e-03 2.27196619e-01 -5.38005829e-01 -4.28824350e-02 -4.78965230e-02 -6.18581712e-01 2.56987363e-01 3.74312311e-01 5.86081564e-01 1.38329268e-01 -5.24803065e-02 7.46885657e-01 4.78708893e-01 7.32798994e-01 -2.89409816e-01 1.02037072e+00 8.31084907e-01 1.40844196e-01 -1.44287258e-01 -7.40723550e-01 3.79786015e-01 -1.42720059e-01 -1.36218801e-01 8.49598169e-01 -8.43996465e-01 -6.53624713e-01 1.60302177e-01 -1.43704581e+00 -1.33412629e-01 -7.01096416e-01 5.57712436e-01 -1.08099759e+00 1.63928241e-01 6.48676325e-03 -7.59011567e-01 -3.55918705e-01 -1.10956931e+00 8.69313359e-01 1.27789795e-01 1.94898963e-01 -8.81842554e-01 -3.15174073e-01 5.41830540e-01 2.05603704e-01 5.75842857e-01 1.00725985e+00 -4.84301209e-01 -1.08349240e+00 -8.99746493e-02 -1.63705036e-01 4.15339708e-01 -7.38044605e-02 -3.49000067e-01 -9.48726535e-01 -1.69253051e-01 5.29753268e-01 -8.31116736e-02 7.79158413e-01 2.94710696e-01 1.44784713e+00 -1.00846601e+00 -3.18533868e-01 6.65045321e-01 1.61848676e+00 2.24204481e-01 9.74271297e-01 2.09754884e-01 4.12946731e-01 4.45182681e-01 8.80330384e-01 5.75130023e-02 3.22553933e-01 7.52243102e-01 1.08425665e+00 -3.42442423e-01 -3.12539577e-01 -7.07071364e-01 5.51699162e-01 6.87324954e-03 1.57961339e-01 -3.77700120e-01 -2.06525952e-01 6.24352396e-01 -1.97889376e+00 -1.39947176e+00 5.23318350e-02 2.25812602e+00 6.95842266e-01 -1.26187652e-01 -3.83823156e-01 -1.57137495e-02 8.69543195e-01 2.35110685e-01 -8.98017585e-01 -4.57613617e-01 -2.08198100e-01 1.94597751e-01 5.15478134e-01 4.02547687e-01 -7.90495813e-01 9.36150253e-01 5.70927334e+00 1.02657974e+00 -1.19961381e+00 4.58061785e-01 9.05826569e-01 2.23196931e-02 -9.11001146e-01 3.64217639e-01 -1.06891409e-01 6.24639452e-01 6.39598429e-01 -4.15025920e-01 4.84468251e-01 9.81213808e-01 1.39721379e-01 1.60924703e-01 -1.05097234e+00 8.43197167e-01 1.86768726e-01 -1.65679181e+00 4.41796064e-01 -2.41146255e-02 9.10794437e-01 -5.36117196e-01 3.90273273e-01 3.17803741e-01 3.61796707e-01 -1.06819212e+00 1.11335766e+00 4.83782500e-01 1.08543229e+00 -7.64216423e-01 8.18076253e-01 6.51336670e-01 -6.48338437e-01 -9.61156264e-02 -3.16044837e-01 -2.94690654e-02 1.51808513e-02 5.67600548e-01 -1.00375640e+00 9.15827155e-01 2.59426504e-01 4.29617465e-01 -9.23221931e-02 4.58383113e-01 -8.99163783e-01 5.89240372e-01 2.06837595e-01 1.34250298e-01 3.03366482e-02 -2.86757231e-01 6.28259838e-01 6.40919983e-01 6.81542933e-01 1.65241167e-01 -1.88903764e-01 1.31810057e+00 -2.38305971e-01 -1.67530812e-02 -9.33542907e-01 4.34980005e-01 3.36299479e-01 8.64031136e-01 -3.12365115e-01 -6.13082170e-01 -2.63208508e-01 1.50406384e+00 2.51993686e-02 3.39237034e-01 -1.35433972e+00 -6.48378283e-02 5.02749205e-01 -1.40368808e-02 3.58159602e-01 6.17428601e-01 -3.55513841e-01 -1.47056091e+00 3.10530037e-01 -1.04800427e+00 4.38110739e-01 -1.12658203e+00 -1.08480930e+00 5.31002045e-01 -1.52582422e-01 -1.52453554e+00 -4.95173246e-01 -3.10494035e-01 -8.47582936e-01 7.08574474e-01 -1.66888392e+00 -1.55605793e+00 2.36898623e-02 6.94092035e-01 5.80393016e-01 -3.42998296e-01 1.00725698e+00 -2.70043403e-01 -2.52008706e-01 6.86267614e-01 -1.80630282e-01 -6.24568723e-02 3.97483945e-01 -1.28247070e+00 1.66257203e-01 1.24312532e+00 2.84223884e-01 2.38785401e-01 9.35943365e-01 -4.75232840e-01 -1.24072015e+00 -1.24394190e+00 8.83508146e-01 -2.57454813e-01 4.03877079e-01 -1.83263391e-01 -4.59813803e-01 8.25118184e-01 4.71567363e-01 -8.79470631e-02 4.23620641e-01 -5.76272428e-01 -3.78092051e-01 -3.29843193e-01 -1.69016027e+00 7.59692073e-01 1.09845996e+00 -3.96254182e-01 -6.84310555e-01 3.85074556e-01 1.13081014e+00 -4.42630261e-01 -3.51106316e-01 2.62215227e-01 2.85364449e-01 -1.15987694e+00 9.50408041e-01 -7.48897433e-01 7.91693926e-01 -4.56005782e-01 -8.09205100e-02 -1.39551091e+00 6.15931861e-02 -9.72600758e-01 -4.30521555e-02 1.10942781e+00 4.22737777e-01 -9.37635958e-01 8.67430389e-01 5.43806255e-01 -1.48801422e-02 -4.13905233e-01 -1.37278342e+00 -8.97798479e-01 4.34820503e-02 -3.70836616e-01 1.29097509e+00 9.10610199e-01 -2.77322710e-01 2.80914068e-01 -6.76273346e-01 3.46678913e-01 8.21754873e-01 9.21359479e-01 9.46012795e-01 -7.03795791e-01 -3.95708174e-01 -1.70920938e-01 -3.81555498e-01 -9.15885448e-01 5.82824886e-01 -1.03481650e+00 -6.21958971e-02 -1.09994757e+00 3.15298766e-01 -7.63973445e-02 -9.70470607e-02 3.24078113e-01 -3.26593101e-01 1.57744363e-01 4.41553324e-01 4.89784479e-02 -1.30473375e-01 8.95386577e-01 1.72343898e+00 -4.34510350e-01 3.79697025e-01 1.25921413e-01 -9.40139830e-01 5.62271476e-01 7.34033704e-01 -6.73992753e-01 -6.95522368e-01 -1.73212707e-01 -1.32106915e-01 4.32330132e-01 8.30836833e-01 -6.37927353e-01 -1.80409610e-01 -4.05730575e-01 1.82127982e-01 -1.45518333e-01 3.13199043e-01 -1.07407165e+00 6.03727579e-01 7.90175676e-01 -2.89592415e-01 -1.46490097e-01 -2.71726936e-01 7.46915936e-01 -3.89498919e-01 -9.94523466e-02 8.62706482e-01 -3.92530650e-01 -6.48686528e-01 5.11769712e-01 2.22978875e-01 2.05122270e-02 1.09697485e+00 -2.03758642e-01 -3.81298423e-01 -7.36932397e-01 -3.53107244e-01 -1.79896038e-02 5.35945356e-01 3.27095658e-01 7.34129250e-01 -1.63246751e+00 -7.70083427e-01 1.16083875e-01 1.15318991e-01 -3.10844481e-02 3.46887976e-01 4.89877015e-01 -2.01437950e-01 3.63619715e-01 -2.08205253e-01 -2.41450652e-01 -9.01719272e-01 9.50818539e-01 4.29357558e-01 -3.87336701e-01 -5.17277181e-01 5.46732545e-01 6.69839501e-01 -4.44656223e-01 -3.83127868e-01 1.57608420e-01 7.07824305e-02 -5.65840662e-01 3.55251223e-01 -2.60506198e-02 -4.16383982e-01 -4.47203547e-01 -1.33663595e-01 3.14764172e-01 3.25470358e-01 -3.21719617e-01 1.20822918e+00 -9.93169099e-02 -8.46867934e-02 1.28244922e-01 1.10312700e+00 -6.33481219e-02 -1.34386492e+00 -3.34659708e-03 -4.71871465e-01 -8.56399894e-01 -3.28425795e-01 -9.73776400e-01 -1.22020710e+00 7.74428546e-01 6.46682918e-01 -1.35800466e-02 1.39994562e+00 -1.29288092e-01 6.87350333e-01 -1.51086777e-01 5.03300369e-01 -9.21959817e-01 -6.54497147e-02 -1.31079733e-01 1.42044854e+00 -1.39491308e+00 -1.00012921e-01 -5.47855675e-01 -9.48328376e-01 7.45173752e-01 5.28810441e-01 -1.54679894e-01 2.47063935e-01 -2.75269419e-01 1.01387180e-01 -1.12011610e-02 -6.90602541e-01 -6.45525753e-03 1.63241208e-01 8.71273398e-01 -1.55301586e-01 3.15443188e-01 -3.83760005e-01 5.65499902e-01 -5.64944386e-01 3.24899405e-02 5.26099920e-01 1.07984148e-01 8.73751268e-02 -1.14753044e+00 -2.71818459e-01 1.54666930e-01 -3.90120476e-01 -1.56597227e-01 -2.93998092e-01 1.02444422e+00 4.39372808e-01 9.00878310e-01 -9.08654183e-02 -9.20091569e-02 2.04185829e-01 -4.46629636e-02 3.75876009e-01 -4.52855557e-01 -3.24960321e-01 -3.54442954e-01 -8.02447647e-02 -6.97407246e-01 -5.24937510e-01 -5.96769691e-01 -1.22560298e+00 -2.41331205e-01 -3.18558991e-01 3.38654220e-01 6.48754954e-01 1.01808369e+00 3.80371213e-01 3.32503974e-01 9.62558389e-01 -2.59559989e-01 -6.97870076e-01 -7.05447376e-01 -4.52196747e-01 7.34903336e-01 3.15508425e-01 -4.56930608e-01 -3.47133070e-01 2.41035014e-01]
[11.622353553771973, -0.32373806834220886]
3142a695-8c56-4362-910d-5932c30766d2
dynamic-scenario-representation-learning-for
2303.04364
null
https://arxiv.org/abs/2303.04364v1
https://arxiv.org/pdf/2303.04364v1.pdf
Dynamic Scenario Representation Learning for Motion Forecasting with Heterogeneous Graph Convolutional Recurrent Networks
Due to the complex and changing interactions in dynamic scenarios, motion forecasting is a challenging problem in autonomous driving. Most existing works exploit static road graphs to characterize scenarios and are limited in modeling evolving spatio-temporal dependencies in dynamic scenarios. In this paper, we resort to dynamic heterogeneous graphs to model the scenario. Various scenario components including vehicles (agents) and lanes, multi-type interactions, and their changes over time are jointly encoded. Furthermore, we design a novel heterogeneous graph convolutional recurrent network, aggregating diverse interaction information and capturing their evolution, to learn to exploit intrinsic spatio-temporal dependencies in dynamic graphs and obtain effective representations of dynamic scenarios. Finally, with a motion forecasting decoder, our model predicts realistic and multi-modal future trajectories of agents and outperforms state-of-the-art published works on several motion forecasting benchmarks.
['Hongkai Xiong', 'Yikang Li', 'Xiaogang Jia', 'Xing Gao']
2023-03-08
null
null
null
null
['trajectory-prediction', 'motion-forecasting']
['computer-vision', 'computer-vision']
[-1.94057047e-01 -2.02573299e-01 -1.94159284e-01 -4.34312969e-01 -4.95571457e-02 -4.98336136e-01 9.64250863e-01 -2.58450300e-01 -1.64500177e-01 7.38675296e-01 7.30406821e-01 -3.60587239e-01 -2.16916669e-02 -9.35600460e-01 -8.34800601e-01 -6.33332789e-01 -5.45422375e-01 4.33301389e-01 8.14064682e-01 -6.89771712e-01 -9.21115726e-02 5.91909409e-01 -1.71360648e+00 2.03833103e-01 8.15686345e-01 3.09868723e-01 3.12574059e-01 1.08612573e+00 -1.67674884e-01 1.21563947e+00 -2.28556827e-01 -4.83709306e-01 -8.33190009e-02 -1.87318400e-01 -3.43525827e-01 2.79339799e-03 5.30465953e-02 -3.98270816e-01 -1.40081310e+00 5.25252163e-01 2.29623541e-01 5.23906648e-01 6.80272281e-01 -1.57107174e+00 -4.09868211e-01 8.14969897e-01 -4.44486499e-01 5.39819419e-01 1.77213207e-01 6.01330876e-01 4.88574147e-01 -2.60185301e-01 7.94052780e-01 1.40971673e+00 4.81072664e-01 4.86244768e-01 -6.70997322e-01 -5.05335271e-01 9.45896924e-01 8.42999220e-01 -8.92866910e-01 -3.79414678e-01 9.98683155e-01 -6.97204530e-01 1.17259753e+00 -8.32768604e-02 7.90887415e-01 1.58528805e+00 7.78044939e-01 9.38121736e-01 3.62343609e-01 4.61600482e-01 -2.27845460e-02 -5.54745317e-01 2.63670325e-01 6.55817032e-01 -1.23082630e-01 2.19377354e-01 -3.46271366e-01 2.43149221e-01 3.11677605e-01 2.26576060e-01 6.50887713e-02 -2.89857149e-01 -1.24872029e+00 5.13871908e-01 5.82037330e-01 -1.36485314e-02 -6.30441308e-01 7.31710017e-01 5.02353907e-01 -6.54656366e-02 4.88644600e-01 -3.34421635e-01 -4.92174886e-02 -4.09222156e-01 -5.60024500e-01 6.32849813e-01 6.07154071e-01 1.04670095e+00 8.16461205e-01 4.77422506e-01 -3.84123772e-01 4.69225228e-01 2.50785977e-01 8.41384172e-01 3.21458369e-01 -7.54927218e-01 7.56660640e-01 5.50368071e-01 1.28871262e-01 -1.39910030e+00 -8.60280335e-01 -4.34802145e-01 -9.48492229e-01 -2.22266465e-01 1.85877025e-01 -4.17538553e-01 -1.13709044e+00 1.82491314e+00 2.72240520e-01 1.06023693e+00 1.63291067e-01 7.44566262e-01 9.20192599e-01 1.01764357e+00 2.57540107e-01 1.27188280e-01 7.89556503e-01 -1.31575131e+00 -8.44753504e-01 -4.05212849e-01 6.05316460e-01 -9.01208445e-02 5.47718287e-01 -3.10761899e-01 -8.56495619e-01 -6.77406490e-01 -8.05514157e-01 1.46106020e-01 -6.40985489e-01 -3.03974003e-01 7.13670194e-01 6.14457652e-02 -1.15195966e+00 1.86642662e-01 -1.02896285e+00 -2.18332365e-01 2.80522257e-01 2.06221864e-01 -1.25103980e-01 -2.87166834e-01 -1.46199250e+00 8.21241438e-01 1.98745087e-01 5.12714148e-01 -1.36767030e+00 -6.04793787e-01 -1.17436302e+00 -1.19745843e-01 3.29205662e-01 -8.47694039e-01 9.79583025e-01 -2.98193097e-01 -1.31741905e+00 3.75267267e-02 -4.87526953e-01 -6.57970965e-01 7.44816601e-01 1.86096609e-01 -1.02221584e+00 -2.89347053e-01 -5.80108538e-02 5.65000713e-01 5.64141214e-01 -1.16188955e+00 -9.14843738e-01 -1.12241514e-01 3.83853167e-01 3.66495878e-01 6.32688254e-02 -4.24472570e-01 -7.72920191e-01 -3.92679870e-01 -4.76465434e-01 -1.24435115e+00 -6.54217064e-01 -7.02061951e-01 -3.67073506e-01 -1.95850745e-01 1.18521261e+00 -5.69846451e-01 1.43450165e+00 -2.03068662e+00 4.83555526e-01 -1.15040317e-01 4.99207228e-01 2.19526365e-01 -3.01611900e-01 6.52666271e-01 6.38248205e-01 -9.92456675e-02 -1.44488975e-01 -3.95787984e-01 3.08982104e-01 5.57216883e-01 -3.17419112e-01 2.97692388e-01 -4.12234142e-02 1.45662916e+00 -1.17436659e+00 -3.22914809e-01 6.21952415e-01 8.22772086e-01 -2.61447966e-01 1.52258813e-01 -3.68260741e-01 7.67103612e-01 -8.00124943e-01 2.75170445e-01 6.66055620e-01 -2.29763195e-01 5.38302846e-02 1.21622749e-01 -2.25787893e-01 -1.53306365e-01 -7.66938984e-01 1.39971197e+00 -5.43854058e-01 1.03366780e+00 -4.16435808e-01 -6.91955924e-01 7.84390926e-01 9.21926834e-03 5.08687377e-01 -8.62537086e-01 3.06365229e-02 -1.07257210e-01 -4.89469841e-02 -8.31734061e-01 8.41048002e-01 5.66394687e-01 -2.08745703e-01 6.88226447e-02 -2.46835232e-01 2.09575012e-01 3.31638187e-01 3.18669587e-01 1.34605432e+00 1.21679470e-01 -3.50665480e-01 1.03537709e-01 4.72120255e-01 -5.30011393e-03 6.08157218e-01 5.49267173e-01 -3.24622780e-01 3.04962277e-01 5.17084062e-01 -9.34707642e-01 -8.53211761e-01 -8.93253744e-01 4.69866872e-01 9.87661719e-01 6.13874853e-01 -2.44000629e-01 -3.02069098e-01 -5.78337967e-01 2.10634157e-01 7.58729935e-01 -8.21074307e-01 -3.12694192e-01 -1.09484184e+00 -8.75614464e-01 3.03614706e-01 6.02853298e-01 5.75180590e-01 -1.07055688e+00 -5.11223674e-01 6.50185049e-01 -2.71430075e-01 -1.39219892e+00 -4.07869905e-01 -5.13840497e-01 -2.96584487e-01 -8.89783084e-01 -5.14444292e-01 -4.34131771e-01 1.59761190e-01 5.37318707e-01 1.14484262e+00 -1.59357578e-01 6.62453845e-02 3.36598217e-01 -2.72197813e-01 -2.33633712e-01 -2.56318897e-01 2.72794396e-01 -3.07321787e-01 3.04248452e-01 1.17517814e-01 -6.51365459e-01 -8.27353418e-01 4.44513470e-01 -7.50107348e-01 3.90811861e-01 2.98475087e-01 5.15389979e-01 2.68097699e-01 3.97529230e-02 5.67962527e-01 -9.51634765e-01 3.05839896e-01 -1.07855821e+00 -4.80514705e-01 2.50570595e-01 -1.71453759e-01 -5.30956686e-02 7.85099983e-01 -4.94565755e-01 -1.26986659e+00 9.45431143e-02 3.10287066e-02 -5.37423730e-01 -2.30661735e-01 5.86276233e-01 -1.23599716e-01 9.31803808e-02 2.80645818e-01 2.14017794e-01 -4.54855025e-01 1.48454115e-01 6.25191033e-01 1.59187183e-01 6.51490271e-01 -7.56855235e-02 7.34864831e-01 6.95199728e-01 1.45500302e-01 -6.80809677e-01 -3.83167833e-01 -3.63579601e-01 -5.23060620e-01 -8.74490023e-01 9.25387263e-01 -1.10144651e+00 -6.54166818e-01 9.78259563e-01 -1.07189775e+00 -9.99406099e-01 1.28248185e-01 6.20441675e-01 -6.12804830e-01 1.04948059e-01 -5.24909317e-01 -8.21252465e-01 3.47608417e-01 -1.33741462e+00 1.20058346e+00 3.99371505e-01 2.32839420e-01 -1.42610002e+00 3.39861304e-01 1.13700569e-01 6.74584866e-01 8.37614596e-01 5.93070447e-01 3.51812281e-02 -9.53068495e-01 -2.24650651e-01 -1.08286880e-01 -5.80392659e-01 1.52492568e-01 3.05890739e-01 -5.62672794e-01 -6.30212203e-02 -7.41479754e-01 3.01430970e-01 1.20870745e+00 7.19730198e-01 1.04538822e+00 -2.90049255e-01 -9.15087879e-01 8.33352864e-01 9.34875369e-01 3.14732045e-01 7.16472864e-01 1.07618883e-01 1.33708298e+00 8.21576953e-01 4.47328210e-01 5.10562599e-01 1.34562862e+00 7.91982830e-01 6.86389863e-01 1.31892353e-01 -2.57114410e-01 -3.93571138e-01 4.83341634e-01 7.32998252e-01 -1.74762279e-01 -1.01231313e+00 -1.23997557e+00 1.00422275e+00 -2.51562834e+00 -1.35763967e+00 -5.48257828e-01 1.50327575e+00 -1.59357652e-01 2.76885569e-01 3.33167225e-01 -6.02002978e-01 8.53390872e-01 8.75006318e-01 -1.03925109e+00 -1.69424906e-01 -4.62360799e-01 -8.04589868e-01 8.20104837e-01 6.06307507e-01 -1.12976217e+00 1.27263820e+00 6.55385208e+00 5.43910623e-01 -1.07623827e+00 -9.85472575e-02 6.18328929e-01 -6.17574342e-02 -8.29141259e-01 -2.05801591e-01 -8.91044140e-01 7.74323165e-01 1.29597616e+00 -3.10802639e-01 3.81774932e-01 4.51986670e-01 5.05867541e-01 2.99007028e-01 -6.39829397e-01 7.70937741e-01 -2.91039627e-02 -1.60588038e+00 2.06710815e-01 1.25692813e-02 9.78991568e-01 5.83209515e-01 2.28155613e-01 5.67691088e-01 1.11308193e+00 -9.90381718e-01 5.52806377e-01 1.13404262e+00 2.28997707e-01 -8.66091967e-01 6.78486764e-01 3.62068236e-01 -1.86756718e+00 -3.25931102e-01 -8.84890258e-02 -2.80900858e-02 9.31090951e-01 2.97930002e-01 -4.85644132e-01 7.89200485e-01 3.88428986e-01 1.55456948e+00 -5.73009014e-01 8.25259447e-01 5.65374829e-02 6.61213398e-01 -1.21038437e-01 -5.77894524e-02 7.16300488e-01 -2.86775947e-01 6.84271872e-01 1.28552413e+00 3.58357400e-01 -6.21368438e-02 3.87320131e-01 3.74368310e-01 1.56713277e-01 -4.39772278e-01 -1.09083688e+00 2.20936369e-02 3.14276546e-01 1.03852284e+00 -4.78883594e-01 -3.73659819e-01 -5.32981038e-01 5.79478323e-01 3.90262038e-01 8.01072836e-01 -1.29722977e+00 1.58565715e-01 9.90686595e-01 -1.81019716e-02 2.84781843e-01 -8.15837264e-01 3.55567157e-01 -1.14683294e+00 -8.73552561e-02 -2.38655865e-01 3.49377275e-01 -7.41349578e-01 -1.08193672e+00 9.08521235e-01 2.38690108e-01 -1.23038256e+00 -6.99251592e-01 -1.59288585e-01 -8.57847810e-01 4.65070248e-01 -1.85158753e+00 -1.70063245e+00 -5.65600455e-01 7.70086467e-01 7.64538407e-01 -2.57705331e-01 6.82833344e-02 3.20449382e-01 -7.87805319e-01 1.48442939e-01 6.18882589e-02 -7.84460604e-02 1.71323061e-01 -1.02662528e+00 1.17289841e+00 1.00426733e+00 -8.63424018e-02 -1.34284511e-01 7.25976527e-01 -8.43806982e-01 -1.60228825e+00 -1.74746883e+00 6.61778450e-01 -3.40564191e-01 9.38853621e-01 -3.88437212e-01 -9.07983720e-01 8.85961115e-01 1.20185554e-01 3.56533051e-01 1.71943113e-01 -2.20814511e-01 -1.47975892e-01 -8.08338635e-03 -6.01006031e-01 9.37524796e-01 1.62451577e+00 -4.15371001e-01 1.13598213e-01 2.94194639e-01 9.59868193e-01 -5.47500610e-01 -3.25147480e-01 5.58090746e-01 6.40386581e-01 -7.41127908e-01 7.98726082e-01 -7.61299849e-01 2.66283900e-01 -3.83751035e-01 1.03209037e-02 -1.58051932e+00 -6.51655078e-01 -6.47881269e-01 -4.87051249e-01 9.11572635e-01 4.38229620e-01 -6.68884933e-01 7.30581701e-01 5.28138518e-01 -6.14100516e-01 -4.63338643e-01 -7.91604280e-01 -4.59359765e-01 -1.55943945e-01 -4.83154297e-01 1.14460266e+00 6.23953462e-01 -4.18026030e-01 -5.06852753e-02 -8.85255218e-01 3.05053383e-01 3.52423042e-01 9.37886983e-02 1.07989609e+00 -9.18467879e-01 5.92130423e-02 -5.14781594e-01 -9.02230680e-01 -1.21781123e+00 7.80280590e-01 -6.33703291e-01 2.59271879e-02 -1.77511764e+00 -9.01614651e-02 -2.30993703e-01 -2.55722046e-01 3.12253274e-02 -3.29619348e-01 -1.29474744e-01 7.05200359e-02 4.17897403e-02 -9.58338082e-01 8.49476635e-01 1.29451931e+00 -5.29778242e-01 -2.22994149e-01 4.88108136e-02 -6.34624204e-03 4.49659288e-01 7.72189617e-01 -1.53005701e-02 -9.06440079e-01 -8.37584257e-01 3.48381579e-01 3.41586471e-01 3.99580061e-01 -1.12323689e+00 5.77710152e-01 -5.71441054e-01 -8.43466893e-02 -1.13733816e+00 4.41083163e-01 -5.49396753e-01 7.47872770e-01 2.19168723e-01 -3.37734133e-01 4.82494414e-01 2.43906692e-01 1.16782188e+00 -2.80861080e-01 6.04190588e-01 1.03764974e-01 -4.57667140e-03 -1.49235368e+00 1.14020860e+00 -8.84607494e-01 -3.94603088e-02 1.35571885e+00 -1.67397410e-01 -6.52827322e-01 -7.34059215e-01 -7.54651785e-01 9.97520924e-01 4.09093112e-01 9.33294117e-01 5.24327219e-01 -1.39230180e+00 -9.58619833e-01 -1.54586270e-01 1.69132471e-01 -1.46753574e-02 1.20262134e+00 5.29550314e-01 -5.36858797e-01 2.27766752e-01 -3.41974318e-01 -6.87145352e-01 -8.84232461e-01 6.08582377e-01 4.09666032e-01 -3.59000474e-01 -9.14623201e-01 5.14297843e-01 3.89292717e-01 -3.84353399e-01 -9.87101421e-02 -2.59993941e-01 -7.31454432e-01 2.88034957e-02 4.20364022e-01 5.59193790e-01 -2.62055218e-01 -1.29946482e+00 -3.05882156e-01 5.50828278e-01 1.99834347e-01 9.30416733e-02 1.27906549e+00 -5.44149280e-01 3.75856161e-01 8.42548251e-01 1.14201891e+00 -3.44177425e-01 -1.83616710e+00 -2.39001930e-01 -1.01494595e-01 -3.16495180e-01 -1.57295063e-01 -2.99153984e-01 -1.43946815e+00 8.60746086e-01 2.73100615e-01 2.11824343e-01 7.58039474e-01 -1.22340836e-01 1.17776346e+00 3.18011552e-01 4.92124081e-01 -8.36323380e-01 -2.56991893e-01 8.99676144e-01 8.34214568e-01 -1.20112503e+00 -4.50587720e-01 -4.28004354e-01 -9.76639509e-01 9.72298801e-01 8.00941765e-01 -2.12076619e-01 9.29076433e-01 3.14281374e-01 -6.23480454e-02 -2.33698741e-01 -1.22578764e+00 -4.92759258e-01 2.71810889e-01 7.81527221e-01 -1.01933263e-01 4.88601923e-01 3.00549895e-01 1.66239291e-01 -2.53958434e-01 -2.13373005e-01 6.61347449e-01 5.36209226e-01 -1.30338997e-01 -6.66511834e-01 3.73518854e-01 4.74289209e-01 9.70437154e-02 4.35076505e-01 -1.13235280e-01 7.23453641e-01 2.71256547e-02 1.13958108e+00 3.27324688e-01 -7.81113505e-01 3.62248182e-01 -3.75216246e-01 -8.51570591e-02 -1.46575928e-01 -3.28047872e-01 -3.46936703e-01 3.07807356e-01 -8.15571427e-01 -4.91674304e-01 -8.25326145e-01 -1.23083913e+00 -7.04941511e-01 2.92007148e-01 -2.42282748e-01 5.62428713e-01 9.86604095e-01 7.27065206e-01 1.06511021e+00 6.93015039e-01 -1.16151416e+00 3.20293486e-01 -6.93008542e-01 -3.24575305e-01 5.09436309e-01 7.44574845e-01 -1.07388127e+00 -2.03189984e-01 -1.52868330e-01]
[5.979978561401367, 0.9209823608398438]
12ef1209-2cf6-4f0b-9446-1d5862bf0bf0
the-naked-sun-malicious-cooperation-between
1911.02423
null
https://arxiv.org/abs/1911.02423v1
https://arxiv.org/pdf/1911.02423v1.pdf
The Naked Sun: Malicious Cooperation Between Benign-Looking Processes
Recent progress in machine learning has generated promising results in behavioral malware detection. Behavioral modeling identifies malicious processes via features derived by their runtime behavior. Behavioral features hold great promise as they are intrinsically related to the functioning of each malware, and are therefore considered difficult to evade. Indeed, while a significant amount of results exists on evasion of static malware features, evasion of dynamic features has seen limited work. This paper thoroughly examines the robustness of behavioral malware detectors to evasion, focusing particularly on anti-ransomware evasion. We choose ransomware as its behavior tends to differ significantly from that of benign processes, making it a low-hanging fruit for behavioral detection (and a difficult candidate for evasion). Our analysis identifies a set of novel attacks that distribute the overall malware workload across a small set of cooperating processes to avoid the generation of significant behavioral features. Our most effective attack decreases the accuracy of a state-of-the-art classifier from 98.6% to 0% using only 18 cooperating processes. Furthermore, we show our attacks to be effective against commercial ransomware detectors even in a black-box setting.
['Giulio Pagnotta', 'Luigi V. Mancini', 'Dorjan Hitaj', 'Fabio De Gaspari', 'Lorenzo De Carli']
2019-11-06
null
null
null
null
['behavioral-malware-detection']
['miscellaneous']
[ 1.53783888e-01 -3.95822108e-01 -2.85403341e-01 -6.04013130e-02 -1.47160947e-01 -9.80906129e-01 9.62750256e-01 1.63894653e-01 -1.98706731e-01 3.14734399e-01 -2.84409463e-01 -7.76503444e-01 5.37789017e-02 -5.69766998e-01 -3.45010102e-01 -8.20080876e-01 -4.10381138e-01 5.06128132e-01 3.76053572e-01 -1.40917867e-01 5.68955123e-01 6.34855211e-01 -1.07245803e+00 1.86607406e-01 3.79806757e-01 4.80896771e-01 -3.35684270e-01 1.08626533e+00 1.78741977e-01 1.02338874e+00 -1.01853549e+00 -2.41617352e-01 3.37487161e-01 -4.37840134e-01 -4.59114790e-01 -1.70039788e-01 1.43309593e-01 -4.51521873e-01 -1.75868973e-01 1.06302190e+00 -9.52485427e-02 -4.63205963e-01 6.74930334e-01 -1.61757994e+00 -5.18413782e-01 6.66347265e-01 -7.01087475e-01 6.12062573e-01 1.95512503e-01 6.13700867e-01 7.59126067e-01 -7.21489117e-02 5.51570773e-01 1.30920243e+00 6.61101401e-01 9.75658417e-01 -1.43393314e+00 -9.25903440e-01 -2.14067265e-01 -1.88614756e-01 -9.31064427e-01 -3.26616168e-01 5.73589087e-01 -7.15354204e-01 1.09931707e+00 4.86306548e-01 6.39876008e-01 1.67613137e+00 7.98833668e-01 3.72147292e-01 1.24047065e+00 1.28518298e-01 3.61298919e-01 2.14374110e-01 5.98562479e-01 6.69502676e-01 8.69912922e-01 2.98561662e-01 -1.83022216e-01 -1.07105839e+00 4.14172530e-01 3.08388352e-01 2.85220277e-02 -2.05095068e-01 -8.88333142e-01 1.07326114e+00 -7.52945058e-03 5.52987933e-01 4.96466979e-02 4.03252512e-01 7.29025364e-01 5.03567398e-01 1.41504750e-01 8.89638841e-01 -5.89561045e-01 -4.92423236e-01 -3.92025560e-01 1.06430866e-01 1.13483310e+00 1.04746327e-01 5.24498165e-01 2.62844205e-01 6.61427900e-02 2.00123653e-01 1.23760112e-01 7.06878245e-01 5.01491308e-01 -5.86467087e-01 -5.68379201e-02 6.75691962e-01 -1.06742613e-01 -1.27692974e+00 -1.15390114e-01 -2.26005360e-01 -4.45283949e-01 9.81813371e-02 3.73394102e-01 -1.88332662e-01 -4.17333722e-01 1.46871269e+00 1.12108856e-01 1.73212156e-01 -4.41131979e-01 3.03124666e-01 -1.26802400e-01 5.16983509e-01 2.33610049e-02 -3.37954015e-01 1.24217427e+00 -6.09705448e-01 -2.72305667e-01 -2.52392203e-01 7.47587264e-01 -4.71433371e-01 8.54439676e-01 3.06938916e-01 -4.15676087e-01 -1.27983987e-01 -8.96426618e-01 9.10665095e-01 -3.27762485e-01 -6.68144643e-01 7.93281555e-01 1.33310413e+00 -1.06463373e+00 4.22591269e-01 -1.10782850e+00 -2.92668313e-01 4.95626450e-01 5.61461747e-01 -2.56651342e-01 4.43014681e-01 -5.00397325e-01 8.71381044e-01 -3.10909767e-02 -5.80186903e-01 -1.44704676e+00 -5.07074237e-01 -3.56070131e-01 9.95633379e-02 6.98411286e-01 -2.69797921e-01 1.08569181e+00 -1.03686953e+00 -9.81196702e-01 6.23442829e-01 -1.86180651e-01 -6.98920786e-01 6.33725941e-01 -9.77023318e-03 -3.16562206e-01 1.44014865e-01 -1.41287312e-01 -2.08270624e-02 1.49834204e+00 -1.36119115e+00 -1.19345583e-01 -4.70839262e-01 -1.31304175e-01 -7.55471408e-01 -7.29247391e-01 4.80276555e-01 3.77876639e-01 -4.12588835e-01 -3.99009943e-01 -1.07784879e+00 -6.30270466e-02 -6.94522083e-01 -6.29238844e-01 -9.85933840e-02 1.67373550e+00 -1.51744157e-01 1.03425574e+00 -1.96727514e+00 -3.28295767e-01 2.88910478e-01 7.26527393e-01 4.75862443e-01 -3.92702483e-02 4.23720121e-01 1.60760388e-01 8.68985891e-01 4.50112112e-02 8.66990685e-02 -8.38978812e-02 -8.91449228e-02 -6.49040401e-01 8.99325550e-01 1.66712403e-01 8.84840846e-01 -8.69676590e-01 -8.69319886e-02 6.98979199e-02 3.85733157e-01 -5.18415868e-01 3.44604313e-01 -1.42712425e-02 3.98613244e-01 -5.49296498e-01 9.38004971e-01 2.57918000e-01 -3.84051681e-01 2.83739328e-01 5.88854969e-01 1.28259659e-01 1.83156833e-01 -2.66591668e-01 -1.04224257e-01 -2.56876379e-01 5.98198712e-01 1.68251961e-01 -5.58963776e-01 9.45855975e-01 -6.03413172e-02 5.61689854e-01 -1.71774682e-02 7.10238874e-01 1.53127372e-01 6.63408637e-01 -3.01315039e-01 1.02626376e-01 1.65014625e-01 3.27326767e-02 1.18748736e+00 -3.37911725e-01 5.15472591e-02 2.71220133e-02 7.39878193e-02 2.10204053e+00 -7.35254884e-01 6.48525476e-01 -4.50502336e-01 5.19605041e-01 7.48205185e-02 3.85530293e-01 1.17450094e+00 -8.41686726e-01 -3.20808709e-01 8.63943994e-01 -5.63836992e-01 -9.37674701e-01 -1.01721358e+00 1.36968881e-01 1.35482764e+00 4.42683324e-02 -5.26536942e-01 -1.01996052e+00 -1.21719146e+00 3.01686108e-01 2.99625933e-01 -9.15549576e-01 -5.23479521e-01 -6.88472867e-01 -9.24413443e-01 1.02040768e+00 3.58460844e-01 2.43307605e-01 -8.33469927e-01 -9.41995621e-01 4.31373641e-02 6.29986703e-01 -9.31110144e-01 -3.95908415e-01 2.67519206e-01 -7.94328570e-01 -1.60195768e+00 2.41490573e-01 -1.84831455e-01 6.24447942e-01 5.77908576e-01 6.95748210e-01 5.08328199e-01 -5.09966671e-01 4.06133115e-01 -3.70657802e-01 -4.06643361e-01 -1.15581489e+00 -1.00964820e-02 4.00402337e-01 8.31835046e-02 8.07797253e-01 -2.17492566e-01 -2.74113677e-02 6.34243071e-01 -5.23122907e-01 -8.15069675e-01 4.84406173e-01 7.88982630e-01 -5.73168695e-02 2.87036538e-01 2.88244963e-01 -1.02625537e+00 8.64644408e-01 -5.92410505e-01 -5.50335407e-01 -1.19798101e-01 -5.62863052e-01 -1.27778441e-01 1.14459622e+00 -1.07044327e+00 -6.37497187e-01 -1.08808883e-01 3.56309146e-01 -5.62138200e-01 -3.52065653e-01 -3.82406116e-01 7.72292167e-02 -4.59360778e-01 8.54475617e-01 3.45551282e-01 3.37923825e-01 -1.75595164e-01 -1.82447746e-01 4.75766599e-01 8.42429698e-03 -4.74529386e-01 1.38495505e+00 7.25481093e-01 3.93380746e-02 -1.12055719e+00 -2.88009644e-01 -4.90962744e-01 -1.94223627e-01 -3.08573544e-01 6.29336238e-01 -6.50959164e-02 -1.50846839e+00 5.31876802e-01 -1.01058781e+00 -2.71374971e-01 1.86911851e-01 -3.19351815e-02 -2.12954298e-01 4.50571656e-01 -1.04067016e+00 -1.00885224e+00 -3.58855516e-01 -1.40937829e+00 8.51828754e-01 -1.96843043e-01 -7.53290653e-01 -1.09378183e+00 3.48383278e-01 4.50889409e-01 7.92682528e-01 1.89738587e-01 9.53827798e-01 -1.55895567e+00 -4.17047501e-01 -5.39281249e-01 6.93332590e-03 1.51411280e-01 2.97391534e-01 5.23963690e-01 -1.10658526e+00 -6.22902989e-01 5.50032914e-01 -2.08822981e-01 7.43054271e-01 2.09991261e-02 1.11054099e+00 -4.85052407e-01 -6.94026411e-01 3.30508202e-01 1.08969116e+00 5.57494879e-01 1.03651509e-01 3.89322281e-01 7.90180624e-01 6.89662814e-01 8.71112421e-02 3.87327820e-01 -3.37982446e-01 3.96214992e-01 6.53419316e-01 3.88478041e-01 5.86109042e-01 5.81754884e-03 8.79994094e-01 3.79274905e-01 -9.58131254e-02 5.11723459e-02 -1.26712453e+00 -1.43534064e-01 -1.48884034e+00 -1.24533117e+00 -3.33823442e-01 2.02589178e+00 2.58614779e-01 4.67274666e-01 4.90271538e-01 8.00051391e-02 7.84337580e-01 3.59317124e-01 -4.93990660e-01 -7.93768942e-01 1.61447525e-01 -6.34583738e-03 6.31208897e-01 3.59007895e-01 -1.36859715e+00 7.70710111e-01 6.90160799e+00 5.51918805e-01 -1.29334092e+00 3.51143062e-01 7.61864603e-01 -4.70252289e-03 1.25416011e-01 9.87487659e-03 -9.10300434e-01 5.86442888e-01 1.28126526e+00 -2.94868022e-01 5.53357303e-01 1.27478170e+00 4.34666455e-01 -2.34385561e-02 -1.14198565e+00 7.48496115e-01 4.01139319e-01 -1.07575607e+00 -8.90762880e-02 7.30163097e-01 6.74660087e-01 3.40256765e-02 2.55391121e-01 3.96050900e-01 5.16089678e-01 -1.33872819e+00 2.97681481e-01 -2.53127456e-01 -4.93805446e-02 -8.77263904e-01 8.72881770e-01 4.54613686e-01 -1.00546849e+00 -6.10751212e-01 -9.69492793e-02 -2.44190186e-01 -2.35229626e-01 5.10872006e-01 -1.21386087e+00 -3.34261149e-01 4.73295808e-01 4.56531405e-01 -9.31405425e-01 4.69942689e-01 1.12790056e-01 1.02478039e+00 -2.90868152e-02 -4.65706408e-01 2.52982229e-01 1.49556696e-01 6.72072113e-01 1.19479370e+00 -1.96138948e-01 -4.51809317e-01 4.82804656e-01 8.40796173e-01 4.49117199e-02 -2.90270932e-02 -1.24239647e+00 -6.41313493e-01 4.36531126e-01 1.26472902e+00 -1.01301908e+00 -2.50797629e-01 -6.06248109e-03 5.58887720e-01 2.21433774e-01 -1.70527652e-01 -9.04013395e-01 -3.25715505e-02 1.05597806e+00 3.31450045e-01 -8.04504678e-02 -2.89102525e-01 -4.40268457e-01 -1.07146144e+00 -3.85196626e-01 -1.34676611e+00 1.98977441e-01 2.38064095e-01 -1.67744958e+00 6.12583280e-01 -2.32813209e-01 -8.66697371e-01 -3.80962878e-01 -7.81278431e-01 -1.03567469e+00 2.41868332e-01 -4.32834238e-01 -8.46139848e-01 -8.35572630e-02 2.79816657e-01 4.66401994e-01 -4.25159395e-01 4.91554648e-01 -5.16238809e-01 -6.86896682e-01 3.62453520e-01 2.28285789e-02 2.87846029e-01 5.08983195e-01 -9.04794216e-01 4.77010280e-01 8.20558131e-01 -1.44256219e-01 1.23862982e+00 8.99517953e-01 -1.02840221e+00 -2.03864002e+00 -1.05913222e+00 2.17183545e-01 -1.06751895e+00 1.37369013e+00 -5.40429950e-01 -9.33619142e-01 7.34965742e-01 -4.75503094e-02 -7.72315189e-02 6.73998892e-01 9.08317566e-02 -8.95005167e-01 2.86057383e-01 -1.29787719e+00 7.11825252e-01 7.40291238e-01 -6.43626690e-01 -3.05960715e-01 1.14432737e-01 2.40161449e-01 3.95866662e-01 -4.02483165e-01 2.40260988e-01 4.65425283e-01 -1.32942283e+00 7.54643917e-01 -7.26984203e-01 2.10630193e-01 -2.22203806e-02 -6.29240572e-02 -9.07136977e-01 -1.60215020e-01 -9.04868305e-01 -5.59002042e-01 8.56144726e-01 -5.43754473e-02 -1.03037345e+00 8.70783746e-01 7.25001618e-02 4.94716257e-01 -8.52796495e-01 -5.70191920e-01 -1.35348690e+00 1.02970358e-02 -2.87591159e-01 2.77543902e-01 1.29512048e+00 2.52175778e-01 2.75981635e-01 -2.72700548e-01 -1.28512248e-01 7.76028335e-01 -1.81067929e-01 1.01713431e+00 -1.22303987e+00 -7.82634854e-01 -8.42590868e-01 -5.65466940e-01 -3.38784039e-01 5.15126586e-01 -6.70986712e-01 -1.49740830e-01 -3.56888950e-01 6.94979310e-01 -8.22541788e-02 1.59054786e-01 4.77130681e-01 -8.15881565e-02 3.40389311e-01 2.07206309e-01 4.17788327e-01 -3.31748247e-01 -9.65816230e-02 7.43772864e-01 -3.37463051e-01 -1.37522087e-01 -6.00934848e-02 -5.05273998e-01 9.04263318e-01 1.09763050e+00 -6.44522727e-01 -3.34691256e-01 3.27601343e-01 -1.80115253e-01 -4.29653227e-01 3.33944499e-01 -6.86950207e-01 -2.54194379e-01 -5.11787474e-01 2.01918036e-01 -1.34653375e-01 2.61210740e-01 -6.44703627e-01 -2.11689323e-02 1.31223536e+00 -1.13783516e-01 4.53120619e-01 -5.86604141e-02 7.76064277e-01 3.16141844e-01 -1.17996834e-01 1.08141828e+00 -2.43805736e-01 -3.09031904e-01 1.36923641e-01 -1.04677343e+00 7.51061663e-02 1.69555807e+00 -2.53091753e-01 -9.73528028e-01 -2.09307894e-01 -1.39138743e-01 -2.88020045e-01 1.02639270e+00 4.10178542e-01 2.66240239e-01 -4.14267033e-01 -3.81887704e-01 3.76321852e-01 -1.03733920e-01 -1.21972740e+00 -2.24441484e-01 1.02029026e+00 -5.56684673e-01 4.47091967e-01 -3.58722150e-01 -6.94431841e-01 -1.84890592e+00 9.20830905e-01 3.47654551e-01 -3.51769656e-01 -4.28694546e-01 5.14092147e-01 4.03391831e-02 -3.91496792e-02 -1.77492760e-02 2.72368103e-01 1.52361050e-01 -2.06080079e-01 8.84979486e-01 7.61811614e-01 -1.59383461e-01 -6.49487615e-01 -6.72750950e-01 3.44017856e-02 -3.57247978e-01 2.06076235e-01 9.19544339e-01 6.11324370e-01 -6.04251862e-01 3.65177542e-01 1.11029923e+00 3.71896684e-01 -8.14320683e-01 5.98325729e-01 3.74203861e-01 -9.25414383e-01 -4.47155058e-01 -4.76978421e-01 -6.21063173e-01 6.55748725e-01 2.38656074e-01 1.09270835e+00 5.32498419e-01 4.37263353e-03 6.26094341e-01 7.05085218e-01 6.70418441e-01 -4.03323859e-01 7.24494934e-01 8.61880362e-01 8.90207812e-02 -1.18718266e+00 -2.68360317e-01 -5.20438910e-01 -3.55499327e-01 9.71711993e-01 1.01123810e+00 -4.06899750e-01 5.53223491e-01 9.25321341e-01 -1.24782631e-02 -1.78381637e-01 -1.07608747e+00 5.58427095e-01 -3.49836290e-01 8.08775604e-01 8.56387466e-02 4.25713927e-01 -2.09900096e-01 1.74098447e-01 -8.70694295e-02 -8.23025644e-01 9.53170121e-01 1.10473943e+00 -8.29075992e-01 -1.04638541e+00 -8.57672751e-01 9.01279569e-01 -7.40159929e-01 1.30497545e-01 -1.38543689e+00 8.24396133e-01 -3.55907232e-01 1.01258183e+00 -3.50990333e-02 -9.72265303e-01 -3.81645918e-01 5.63075766e-02 2.46006534e-01 -8.07127178e-01 -1.16730464e+00 -1.48200393e-01 2.33600587e-02 -6.34126902e-01 1.94943562e-01 -5.53946316e-01 -8.65052462e-01 -1.07814324e+00 -1.09804712e-01 6.84242509e-03 5.34426928e-01 9.59438980e-01 2.73134917e-01 1.92702979e-01 8.87445927e-01 -8.46986532e-01 -1.10092890e+00 -7.37383902e-01 -3.71571362e-01 3.59158248e-01 3.07552993e-01 -7.33024836e-01 -8.59130979e-01 -1.76332176e-01]
[14.388594627380371, 9.666810035705566]
f775b21c-9f64-4041-8f82-e81b1aa126b0
commonsense-knowledge-aware-concept-selection
2102.02963
null
https://arxiv.org/abs/2102.02963v1
https://arxiv.org/pdf/2102.02963v1.pdf
Commonsense Knowledge Aware Concept Selection For Diverse and Informative Visual Storytelling
Visual storytelling is a task of generating relevant and interesting stories for given image sequences. In this work we aim at increasing the diversity of the generated stories while preserving the informative content from the images. We propose to foster the diversity and informativeness of a generated story by using a concept selection module that suggests a set of concept candidates. Then, we utilize a large scale pre-trained model to convert concepts and images into full stories. To enrich the candidate concepts, a commonsense knowledge graph is created for each image sequence from which the concept candidates are proposed. To obtain appropriate concepts from the graph, we propose two novel modules that consider the correlation among candidate concepts and the image-concept correlation. Extensive automatic and human evaluation results demonstrate that our model can produce reasonable concepts. This enables our model to outperform the previous models by a large margin on the diversity and informativeness of the story, while retaining the relevance of the story to the image sequence.
['Hideki Nakayama', 'Hiroya Takamura', 'Yifei HUANG', 'Hong Chen']
2021-02-05
null
null
null
null
['visual-storytelling']
['natural-language-processing']
[ 3.61881196e-01 3.65279436e-01 -3.74505579e-01 -2.25174665e-01 -3.73944074e-01 -5.03473222e-01 9.07968342e-01 2.40347579e-01 -7.92248175e-02 6.90176308e-01 6.40309215e-01 2.33264685e-01 5.32066748e-02 -7.33356118e-01 -6.04255140e-01 -4.36927915e-01 6.66856617e-02 2.20834747e-01 3.25562716e-01 -1.93726957e-01 6.61778867e-01 1.44280687e-01 -2.03086996e+00 6.16710663e-01 1.04068851e+00 6.51720285e-01 6.96630001e-01 4.24571007e-01 -3.66449714e-01 1.28737950e+00 -6.83768570e-01 -2.85075754e-01 -3.79711129e-02 -1.05992663e+00 -8.77246022e-01 7.36493349e-01 1.74549520e-01 -2.54115522e-01 -1.40993401e-01 1.10196316e+00 8.74435976e-02 3.23652267e-01 5.82126141e-01 -1.26828790e+00 -8.44107568e-01 9.52251196e-01 -6.15960419e-01 -7.54000386e-03 5.72439969e-01 -2.71967500e-02 1.24115729e+00 -1.17393804e+00 1.21345448e+00 1.26040888e+00 8.21033269e-02 6.22279882e-01 -1.00917673e+00 -5.58644533e-01 3.10110450e-01 4.60083246e-01 -1.37554002e+00 -2.32479841e-01 1.05429578e+00 -3.93223614e-01 4.78106081e-01 2.92778254e-01 1.07087088e+00 8.56513381e-01 -1.21895205e-02 1.00866890e+00 6.37101650e-01 -4.46130484e-01 4.76394087e-01 3.98605675e-01 -1.24015965e-01 5.57315528e-01 1.20961912e-01 -3.15565228e-01 -9.26643908e-01 -5.03676049e-02 7.23085582e-01 -4.93815057e-02 -3.33427310e-01 -6.66713953e-01 -1.37412345e+00 8.14119339e-01 4.49325323e-01 4.87747937e-01 -5.29523015e-01 9.97614264e-02 2.25896984e-01 -9.17510167e-02 2.95361429e-01 7.76080847e-01 2.54206866e-01 3.23853195e-01 -1.00169194e+00 2.42003113e-01 4.33919609e-01 1.18296063e+00 8.91588330e-01 -9.75325257e-02 -6.14648163e-01 7.33722925e-01 1.06837735e-01 2.11602971e-01 4.48389053e-01 -9.06418324e-01 1.84028804e-01 8.84974837e-01 8.25424120e-03 -1.33712053e+00 3.01608115e-01 -3.85049343e-01 -6.53141201e-01 2.34690104e-02 -2.62451410e-01 2.42398068e-01 -7.62403369e-01 1.70255589e+00 2.18123958e-01 1.63196430e-01 -5.04740886e-02 7.95745313e-01 7.74825752e-01 9.80542243e-01 1.64439023e-01 -4.45070714e-01 1.19808769e+00 -9.74905789e-01 -7.27468848e-01 -3.51187915e-01 2.60208488e-01 -7.67865539e-01 1.28733802e+00 1.18036710e-01 -1.15044475e+00 -6.10711396e-01 -1.18701410e+00 9.93557572e-02 2.25940831e-02 1.53806895e-01 2.99940228e-01 2.12613046e-01 -9.22480345e-01 3.87860119e-01 -1.04249835e-01 -5.14313936e-01 5.15906036e-01 -2.72718787e-01 -1.61009118e-01 -3.91671956e-01 -9.45905089e-01 6.57471418e-01 1.02372992e+00 -5.21672130e-01 -1.19454646e+00 -5.37811399e-01 -8.46400738e-01 1.47087932e-01 4.96119022e-01 -8.29340279e-01 8.59799206e-01 -1.50494945e+00 -9.17558670e-01 7.74010777e-01 -1.56231672e-01 -3.52895141e-01 2.08933309e-01 -8.19318369e-02 -2.50908762e-01 6.81136429e-01 3.01616013e-01 1.21194804e+00 1.03264618e+00 -1.76636696e+00 -9.28842247e-01 1.99397668e-01 1.75782531e-01 5.13438582e-01 -6.53211653e-01 -5.67524992e-02 -7.84322202e-01 -8.98888290e-01 1.44303828e-01 -7.16930926e-01 -3.58599931e-01 1.41288172e-02 -4.21749532e-01 7.48998523e-02 7.32725322e-01 -4.72228616e-01 1.25039220e+00 -2.25809050e+00 6.93481788e-02 3.20882231e-01 3.58525813e-01 -1.80861548e-01 -2.89493322e-01 5.17931581e-01 1.69371702e-02 3.73390503e-02 -3.18122983e-01 1.19448878e-01 -3.62386733e-01 1.29443347e-01 -3.35233599e-01 -6.70107752e-02 4.01922256e-01 8.41663480e-01 -1.27921927e+00 -8.17448795e-01 2.67339554e-02 2.14023590e-01 -6.45953476e-01 3.43224317e-01 -5.81146836e-01 2.69650280e-01 -5.71776330e-01 3.90039563e-01 4.33948457e-01 -4.09198344e-01 3.24826986e-01 4.34808768e-02 2.23042995e-01 5.18481508e-02 -1.04123414e+00 1.51278317e+00 -6.28689304e-02 6.92357302e-01 -5.89131892e-01 -6.24015450e-01 1.15774810e+00 1.10998705e-01 4.13217902e-01 -4.34505165e-01 -5.33010922e-02 -1.35456830e-01 -1.84681356e-01 -6.12212479e-01 1.04683149e+00 -3.81726682e-01 1.35636330e-01 5.68358362e-01 3.14903110e-02 -1.95043057e-01 5.59723377e-01 8.45470488e-01 6.85047269e-01 6.26123995e-02 4.71046478e-01 -3.66923183e-01 4.75650728e-01 2.58481711e-01 3.29255968e-01 7.01521456e-01 9.76063590e-03 7.42996454e-01 4.47847784e-01 -2.82765925e-01 -1.28659260e+00 -1.01559460e+00 4.92318213e-01 1.05832279e+00 5.46206176e-01 -5.38008392e-01 -6.08138084e-01 -5.79614460e-01 -3.74261916e-01 1.15077901e+00 -7.48558462e-01 -4.03846502e-01 -4.05112624e-01 -3.47434163e-01 8.14755484e-02 3.49141419e-01 4.19851929e-01 -1.27647889e+00 -6.59858108e-01 1.13181345e-01 -7.87928164e-01 -9.39349592e-01 -6.75563455e-01 -4.15107578e-01 -6.75736964e-01 -1.00647163e+00 -6.64294541e-01 -1.08666658e+00 1.11602426e+00 7.66751587e-01 1.24326313e+00 1.90367967e-01 -4.89744954e-02 3.04969043e-01 -8.06791544e-01 -3.62893462e-01 -6.32705986e-01 -4.76827890e-01 -3.20972770e-01 1.37433544e-01 1.36028584e-02 -4.15230542e-01 -5.00060856e-01 1.08573824e-01 -1.26900530e+00 5.79449952e-01 5.06537557e-01 7.46704757e-01 6.72825992e-01 3.09491009e-01 5.90183854e-01 -5.45270324e-01 7.60996103e-01 -5.68515241e-01 -1.26799911e-01 3.79869461e-01 -4.48468357e-01 2.05649897e-01 4.62487578e-01 -6.22054219e-01 -1.32657158e+00 6.76032528e-02 3.24051619e-01 -4.31909084e-01 2.17446491e-01 6.07363522e-01 -1.01496331e-01 3.99686396e-01 7.80920565e-01 5.24330914e-01 -2.03982055e-01 6.46117553e-02 8.41443419e-01 1.19664639e-01 5.23609877e-01 -3.44364107e-01 6.45901322e-01 6.16560161e-01 -3.35453212e-01 -7.32341886e-01 -7.41070032e-01 -4.99561816e-01 -4.74803507e-01 -6.02286041e-01 6.88875854e-01 -9.25449550e-01 3.81126292e-02 -1.53935820e-01 -1.19507504e+00 3.10611516e-01 -7.20725179e-01 3.97873133e-01 -6.35574162e-01 5.12114704e-01 -2.06863359e-01 -8.19636464e-01 -2.25310758e-01 -8.40282202e-01 6.77133918e-01 2.29999885e-01 -4.35323387e-01 -5.47271073e-01 -1.72768936e-01 1.85154185e-01 3.74885462e-02 3.69072646e-01 9.85368311e-01 -4.24903989e-01 -8.99752975e-01 -1.85474902e-01 -2.67784506e-01 -1.45547709e-03 4.72127497e-02 2.13352256e-02 -7.05217600e-01 -5.72766997e-02 1.81174278e-02 -4.32636112e-01 1.13465250e+00 4.11005706e-01 9.73705173e-01 -5.55135906e-01 -3.13502461e-01 3.87342647e-02 1.36662185e+00 3.96975905e-01 8.77943516e-01 3.72900367e-01 5.74156821e-01 8.28089833e-01 8.72734189e-01 7.31437802e-01 3.37160170e-01 1.85732439e-01 3.74208540e-01 2.66549494e-02 -1.33437425e-01 -7.92600453e-01 3.25682223e-01 6.62556052e-01 1.17853455e-01 -4.52569008e-01 -6.28310263e-01 1.02724075e+00 -1.92486668e+00 -1.43882883e+00 1.66911572e-01 2.04468155e+00 8.14528406e-01 1.04863159e-01 1.04149535e-01 -7.60074379e-03 1.02017343e+00 2.15508386e-01 -3.04389626e-01 -9.21070874e-02 -3.62614781e-01 -3.80385399e-01 -2.32107624e-01 2.73453623e-01 -7.14073122e-01 1.10142398e+00 6.48784542e+00 1.06329548e+00 -6.15225255e-01 -2.36515060e-01 7.96432555e-01 -3.42693269e-01 -8.89999628e-01 2.40381747e-01 -3.80139500e-01 1.63129136e-01 1.71977878e-01 -6.37791514e-01 1.17556028e-01 9.81166184e-01 1.71846926e-01 -2.68758714e-01 -9.39006090e-01 7.69916415e-01 5.74063778e-01 -1.78229308e+00 7.82902896e-01 -2.55866379e-01 1.14704406e+00 -6.92129195e-01 -9.98674631e-02 1.86218135e-02 2.98862159e-01 -7.58973598e-01 9.68802094e-01 5.71991205e-01 6.24841511e-01 -1.08074594e+00 4.27986860e-01 2.92058259e-01 -1.23337519e+00 -5.60759306e-02 -3.76320511e-01 1.44105688e-01 1.64002046e-01 6.15001440e-01 -9.32001233e-01 4.25309747e-01 4.62267101e-01 7.09370732e-01 -6.60407245e-01 8.74802172e-01 -5.42108774e-01 3.38603079e-01 2.68184125e-01 -2.25184575e-01 3.00929993e-01 -8.25304445e-03 7.43853271e-01 1.19918096e+00 5.33934236e-01 2.57488757e-01 2.22654656e-01 1.20791924e+00 -2.07149759e-01 4.06689793e-01 -6.89778030e-01 -1.84672117e-01 6.32680774e-01 1.18914926e+00 -1.01677096e+00 -7.09541202e-01 -1.00285716e-01 1.06769550e+00 1.73920915e-01 2.91905433e-01 -5.95771730e-01 -3.02058399e-01 1.42361745e-01 1.92884773e-01 3.22366178e-01 7.08654448e-02 -3.29700023e-01 -9.83846068e-01 4.47178893e-02 -8.24176908e-01 5.57422280e-01 -1.27626848e+00 -1.03141904e+00 7.44796574e-01 1.43405929e-01 -1.50043464e+00 -3.03501159e-01 1.16391793e-01 -7.38390386e-01 4.55331206e-01 -1.30068624e+00 -1.27226806e+00 -3.62273932e-01 4.96237069e-01 8.95554960e-01 -2.21372277e-01 2.68120766e-01 -3.84517103e-01 -3.30160782e-02 1.19280182e-01 -6.04556985e-02 -2.64129788e-01 5.42052925e-01 -1.05206406e+00 2.54483253e-01 1.21659505e+00 2.26603761e-01 6.55540168e-01 9.22417700e-01 -9.78680968e-01 -8.06799293e-01 -1.13191378e+00 1.20585299e+00 -5.59693649e-02 5.92593551e-01 -1.91088974e-01 -8.95812988e-01 3.68935376e-01 5.56626081e-01 -5.84696054e-01 7.61828065e-01 -1.66448280e-01 -5.19099295e-01 2.63115019e-01 -9.82772768e-01 1.01915181e+00 9.77701485e-01 -2.59087682e-01 -7.97979712e-01 1.14090137e-01 1.00882065e+00 2.09847003e-01 -2.54792869e-01 -1.84039455e-02 4.76934671e-01 -9.66149509e-01 1.02714849e+00 -4.01347011e-01 1.18611777e+00 -4.98393208e-01 -1.61687955e-02 -1.36862588e+00 -6.23870075e-01 -4.43134636e-01 -2.67817592e-03 1.29322267e+00 4.02323544e-01 1.47992298e-01 7.31796920e-01 4.02669489e-01 -4.51832116e-02 -4.00864571e-01 -4.71987814e-01 -7.94175923e-01 -3.66184413e-01 -2.53968537e-01 7.16402948e-01 1.09903932e+00 5.84137797e-01 5.93605161e-01 -6.79987967e-01 -1.77299663e-01 5.32254100e-01 4.27530259e-01 5.57046950e-01 -8.76312077e-01 -2.21154138e-01 -4.37624782e-01 -2.72182494e-01 -8.91354144e-01 -1.02396302e-01 -8.15150261e-01 3.74069601e-01 -1.73541176e+00 9.97131348e-01 6.56684190e-02 -2.20444530e-01 2.94830322e-01 -6.82105243e-01 1.32833838e-01 4.95430022e-01 4.94968832e-01 -9.10876691e-01 8.06309462e-01 1.56017637e+00 -1.50162026e-01 -4.15779471e-01 -4.42101806e-01 -1.13902879e+00 6.41206205e-01 8.34396899e-01 -3.54233980e-01 -8.85643423e-01 -1.55940190e-01 2.71272093e-01 -9.99592766e-02 3.46339852e-01 -9.69138324e-01 1.25431135e-01 -5.81333995e-01 4.93195266e-01 -8.22265446e-01 2.57681429e-01 -5.32862067e-01 1.56665936e-01 4.04469609e-01 -7.72444189e-01 -1.56608075e-01 5.07582687e-02 7.18264580e-01 -3.35957915e-01 -3.29123080e-01 8.92709434e-01 -3.75615984e-01 -1.05314422e+00 1.24998122e-01 -5.66205025e-01 5.45638315e-02 1.24885345e+00 -5.16248524e-01 -2.10543692e-01 -7.84905553e-01 -4.17069048e-01 3.33325535e-01 6.16737485e-01 6.06454194e-01 1.25638139e+00 -1.69616139e+00 -9.27578509e-01 -3.07438336e-02 7.32874513e-01 -3.18693578e-01 3.20729047e-01 1.41327560e-01 -2.39714071e-01 5.68874367e-02 -5.45591295e-01 -3.20347279e-01 -1.41954541e+00 9.45501387e-01 -2.49441773e-01 -1.42863855e-01 -6.67690635e-01 8.95271480e-01 6.58363521e-01 2.60366231e-01 3.40212993e-02 -3.45274508e-02 -6.24651313e-01 9.54279080e-02 8.98389995e-01 1.68122694e-01 -6.32708430e-01 -7.56389499e-01 -1.79170981e-01 2.93393850e-01 -1.03092104e-01 -2.85768598e-01 1.19501686e+00 -4.00065392e-01 -1.14607617e-01 2.42776617e-01 7.14680672e-01 -7.82827735e-02 -1.22842109e+00 -3.39958698e-01 1.57582194e-01 -7.15061605e-01 -1.96551934e-01 -5.87666988e-01 -8.44457030e-01 5.24737775e-01 5.47604747e-02 7.24965408e-02 1.51239622e+00 2.64036477e-01 5.84552050e-01 1.47579327e-01 1.92069694e-01 -1.11334121e+00 8.62663448e-01 2.83884585e-01 1.28547537e+00 -9.32359636e-01 1.16377562e-01 -7.07069278e-01 -1.32828689e+00 1.03804111e+00 6.36432350e-01 3.67270783e-02 1.56928539e-01 -2.64468551e-01 -1.84411600e-01 -2.61042982e-01 -7.86572278e-01 -3.92629206e-01 4.62782621e-01 7.79073179e-01 3.12660575e-01 -8.64571780e-02 -5.22348642e-01 4.81905550e-01 -5.17460480e-02 8.33454356e-02 7.75946677e-01 8.88916194e-01 -1.09312356e+00 -7.75303543e-01 -4.50616896e-01 1.58086568e-01 6.42944798e-02 -2.11964712e-01 -8.12719882e-01 4.57042366e-01 2.35716090e-01 1.03625810e+00 -1.89693142e-02 -4.14467007e-01 1.82025269e-01 -1.42301098e-01 2.34689757e-01 -6.83723330e-01 -3.17332774e-01 2.95513332e-01 1.43978134e-01 -2.43336901e-01 -5.50928235e-01 -4.84833390e-01 -1.33406222e+00 -1.59878969e-01 -3.86458188e-01 1.38513133e-01 1.75104797e-01 8.91398191e-01 3.42135847e-01 4.19275880e-01 7.25894868e-01 -5.81521332e-01 5.15549406e-02 -6.40995502e-01 -7.32676804e-01 7.09681392e-01 -1.43512309e-01 -4.17291164e-01 -5.34400530e-02 6.96931362e-01]
[11.119112014770508, 0.6785460114479065]
1b9283bf-38cc-4aec-adcc-56514f62308f
improving-perceptual-quality-of-drum
2004.00188
null
https://arxiv.org/abs/2004.00188v5
https://arxiv.org/pdf/2004.00188v5.pdf
Improving Perceptual Quality of Drum Transcription with the Expanded Groove MIDI Dataset
We introduce the Expanded Groove MIDI dataset (E-GMD), an automatic drum transcription (ADT) dataset that contains 444 hours of audio from 43 drum kits, making it an order of magnitude larger than similar datasets, and the first with human-performed velocity annotations. We use E-GMD to optimize classifiers for use in downstream generation by predicting expressive dynamics (velocity) and show with listening tests that they produce outputs with improved perceptual quality, despite similar results on classification metrics. Via the listening tests, we argue that standard classifier metrics, such as accuracy and F-measure score, are insufficient proxies of performance in downstream tasks because they do not fully align with the perceptual quality of generated outputs.
['Jesse Engel', 'Lee Callender', 'Curtis Hawthorne']
2020-04-01
null
null
null
null
['drum-transcription']
['music']
[ 3.93350199e-02 2.01788858e-01 2.32065842e-02 1.18915349e-01 -1.22909498e+00 -9.55032766e-01 5.63752174e-01 7.82195032e-02 -2.02613324e-01 4.61400628e-01 1.03323877e+00 -1.52317628e-01 -3.49975437e-01 -5.46663165e-01 -3.74544442e-01 -3.02976251e-01 -1.61291838e-01 3.36821198e-01 8.37918893e-02 -4.14937913e-01 3.83459151e-01 9.36216768e-03 -1.73935139e+00 7.18832910e-01 4.07838672e-01 1.11536801e+00 -2.65082240e-01 1.33154893e+00 5.18269300e-01 1.00894308e+00 -1.09780455e+00 -3.57281387e-01 9.81165841e-02 -7.66522408e-01 -1.01540506e+00 -4.55258965e-01 9.86737311e-02 -1.95442036e-01 -1.41974196e-01 1.31046742e-01 1.04802072e+00 1.42970458e-01 8.66307974e-01 -1.19377935e+00 -4.41238463e-01 1.16310453e+00 5.48661388e-02 3.27072263e-01 7.32372999e-01 6.00699306e-01 1.55583429e+00 -6.76983476e-01 6.98209167e-01 1.27920425e+00 1.25901604e+00 3.06435525e-01 -1.51613307e+00 -5.56817412e-01 -4.13740844e-01 -8.14201459e-02 -1.08784866e+00 -6.60011351e-01 5.15385509e-01 -8.45725536e-01 1.13038373e+00 4.48975295e-01 9.97875452e-01 1.40300047e+00 -1.02712803e-01 6.18286967e-01 7.00470686e-01 -2.81285405e-01 1.04558572e-01 -2.79476583e-01 -2.13752866e-01 2.20851049e-01 -1.89259350e-01 3.92589390e-01 -1.22000229e+00 -2.45950744e-02 5.61026216e-01 -1.27391243e+00 -3.54906261e-01 3.65691930e-01 -1.60002089e+00 6.65666163e-01 -1.52553385e-02 1.90725267e-01 -1.06813781e-01 3.62681657e-01 7.89964676e-01 4.33885217e-01 4.08084124e-01 1.22387815e+00 -5.67758799e-01 -1.20616233e+00 -8.98628354e-01 5.93037307e-01 7.78500259e-01 6.90046549e-01 2.42613051e-02 4.32335645e-01 -4.79527891e-01 9.47100759e-01 -2.20605321e-02 1.59299299e-01 7.78342426e-01 -1.53611708e+00 5.98869860e-01 9.60263386e-02 -3.83502841e-02 -7.59889841e-01 -7.25729942e-01 -5.72838962e-01 -2.40676820e-01 -1.16688209e-02 7.44386435e-01 -4.00309980e-01 -3.92712533e-01 1.81925261e+00 -2.74159044e-01 -2.06693504e-02 -8.68065655e-03 8.85691702e-01 9.80338395e-01 6.29731417e-01 -7.14060441e-02 -1.75826520e-01 7.76042223e-01 -8.16042900e-01 -4.62863594e-01 -1.07499428e-01 8.03247571e-01 -1.04531693e+00 1.58868933e+00 9.22721624e-01 -1.22097886e+00 -9.01080251e-01 -1.21899116e+00 -1.05329998e-01 1.65808350e-01 3.32047679e-02 7.44592428e-01 7.67504454e-01 -9.89999533e-01 1.11048102e+00 -4.78556842e-01 8.71786475e-03 -3.90790552e-02 1.60820052e-01 -1.92179307e-01 8.18861663e-01 -9.72583115e-01 8.25284600e-01 2.29066208e-01 -1.75007164e-01 -9.63174164e-01 -8.93894851e-01 -6.33968353e-01 -3.18769485e-01 -3.41908574e-01 -5.19681990e-01 1.79155946e+00 -3.47853035e-01 -1.76478469e+00 5.45769930e-01 4.99213159e-01 -2.65432715e-01 6.09473050e-01 -3.94119620e-01 -3.47191125e-01 2.45325640e-02 -3.38936746e-02 7.59131551e-01 6.65145874e-01 -1.00222290e+00 -6.90657079e-01 4.96561974e-02 -2.25086927e-01 1.71133190e-01 -2.39660099e-01 -6.24509715e-02 8.87149200e-02 -1.03142214e+00 -1.22513890e-01 -8.74887347e-01 3.65488857e-01 -7.84717798e-01 -4.73650247e-01 -9.11471397e-02 4.05778050e-01 -9.82045293e-01 1.68628740e+00 -2.06396079e+00 3.54293227e-01 5.19595146e-02 -3.59964743e-02 -1.91332534e-01 -2.61348516e-01 3.77305061e-01 -9.58523229e-02 3.89496356e-01 3.99589352e-02 -3.02773923e-01 1.00217789e-01 -2.45302469e-01 -4.93051171e-01 7.81527311e-02 3.35845739e-01 6.30948246e-01 -8.66739988e-01 -3.30062598e-01 -6.10329658e-02 7.14755803e-02 -1.00156760e+00 7.42483437e-02 -2.98182443e-02 7.29200125e-01 1.99933037e-01 3.88431430e-01 -2.08643898e-01 3.04185629e-01 -2.63994366e-01 -1.42297849e-01 -1.04260907e-01 8.14663589e-01 -1.02483571e+00 1.72392249e+00 -5.82942486e-01 1.03508699e+00 -2.38606721e-01 -4.87577647e-01 8.94009471e-01 5.94130576e-01 4.19540524e-01 -3.81683797e-01 1.94009632e-01 3.84756297e-01 6.47036135e-01 -5.44822752e-01 6.38235629e-01 -9.98143256e-02 -4.80485499e-01 5.83402932e-01 4.58030820e-01 -8.92418325e-01 3.02882552e-01 -1.20330900e-01 1.26256013e+00 4.40693319e-01 -2.40869522e-01 -7.97201395e-02 -1.48158967e-01 1.34171829e-01 5.17851532e-01 6.91328883e-01 4.53128181e-02 1.10238779e+00 6.30170941e-01 1.12305522e-01 -1.29405212e+00 -1.16347384e+00 -1.84069127e-01 1.33866704e+00 -4.72066879e-01 -1.03044951e+00 -8.64790082e-01 1.53850913e-02 3.92717943e-02 7.40729034e-01 -4.76993769e-01 -3.45112622e-01 -4.91621494e-01 -6.15614474e-01 1.38752949e+00 7.97015309e-01 -2.29636505e-01 -1.19665694e+00 -5.62160313e-01 5.16210794e-01 -3.70844245e-01 -6.49644375e-01 -3.90045255e-01 3.22370738e-01 -8.74542773e-01 -9.07354891e-01 -4.22604591e-01 -5.44730365e-01 -5.32693446e-01 -5.13584912e-01 1.32192945e+00 -1.80233404e-01 -2.00688630e-01 2.02515185e-01 -5.53429842e-01 -4.54173028e-01 -8.05504560e-01 3.60712469e-01 4.39433783e-01 -5.84485173e-01 -1.10065520e-01 -9.62597966e-01 -3.59909952e-01 2.68437147e-01 -3.69072199e-01 2.04454392e-01 3.93568367e-01 8.79159272e-01 2.86630362e-01 -2.00744569e-01 9.91641164e-01 -4.15227413e-01 1.20750809e+00 -1.80781603e-01 5.11044860e-02 -4.16315883e-01 -5.18092573e-01 -1.33328140e-01 5.40625095e-01 -6.12559795e-01 -7.38930941e-01 -4.60855633e-01 -5.61761260e-01 -1.96600661e-01 7.06389025e-02 3.90589416e-01 9.40843523e-02 5.19021749e-01 1.17731714e+00 -3.29539835e-01 -1.38361275e-01 -5.84985435e-01 3.77192467e-01 9.00057375e-01 1.12965429e+00 -6.78670704e-01 6.37951434e-01 -2.23277628e-01 -1.54910237e-01 -6.90684855e-01 -6.71424150e-01 -1.11890420e-01 -4.36165333e-01 -3.79544079e-01 7.24180698e-01 -9.09517944e-01 -8.98932040e-01 4.93127078e-01 -8.97933543e-01 -8.22692633e-01 -4.80165064e-01 7.27675676e-01 -1.13595438e+00 -1.86966240e-01 -1.09271598e+00 -9.71079528e-01 -2.30263159e-01 -1.06248474e+00 1.13761055e+00 -1.20059885e-01 -1.11914313e+00 -6.76328361e-01 2.37202272e-01 4.06521767e-01 2.23550007e-01 4.26237166e-01 1.04290009e+00 -5.44494569e-01 8.16754475e-02 5.16463071e-03 3.28285962e-01 3.46859276e-01 -6.92527592e-02 3.93033236e-01 -1.40703440e+00 2.36612111e-02 -3.62712443e-01 -7.01555729e-01 6.77305579e-01 4.18870568e-01 9.59137380e-01 -2.08114401e-01 2.01542884e-01 5.68347812e-01 7.36945033e-01 1.93209693e-01 4.70422417e-01 3.05702209e-01 4.94383216e-01 7.36761928e-01 7.10353255e-01 7.11541474e-01 1.12145744e-01 8.66715074e-01 -3.72021161e-02 3.63066316e-01 -4.94093090e-01 -8.52014244e-01 5.70882916e-01 1.32550073e+00 -4.73040164e-01 -2.34260082e-01 -1.01582718e+00 5.85555494e-01 -1.58890438e+00 -9.79472876e-01 -2.60310680e-01 1.87763679e+00 1.09010935e+00 4.27074581e-01 8.45242918e-01 1.00672889e+00 4.20772076e-01 -1.24501944e-01 -1.72993451e-01 -7.64688849e-01 -1.26501158e-01 5.10352790e-01 4.06702012e-02 3.91458690e-01 -8.22974503e-01 6.53619051e-01 8.28037739e+00 8.24927986e-01 -7.12339818e-01 9.76817980e-02 5.95690072e-01 -5.08623302e-01 -2.38921851e-01 -1.12722188e-01 -5.04386961e-01 5.67423463e-01 1.32531631e+00 -6.64081052e-02 6.93526924e-01 5.18068075e-01 4.36983377e-01 1.25532061e-01 -1.60211575e+00 9.16957438e-01 -2.67335773e-01 -1.25754786e+00 -2.34216586e-01 -2.91486904e-02 7.62889147e-01 -2.71040171e-01 2.53363132e-01 5.64962685e-01 1.65116757e-01 -1.25092387e+00 1.57044137e+00 4.55257088e-01 1.04963279e+00 -8.04539740e-01 6.35429680e-01 2.62898386e-01 -1.10228610e+00 -2.81187057e-01 2.07707062e-01 -4.64918405e-01 2.15499997e-01 2.55259365e-01 -9.16988134e-01 1.44667953e-01 7.81070411e-01 4.61185455e-01 -7.29906380e-01 9.97859657e-01 -1.82427481e-01 1.22332048e+00 -2.54873812e-01 7.97818825e-02 -5.58939427e-02 2.27413923e-01 8.97443116e-01 1.51175010e+00 5.42585135e-01 -3.11535001e-01 -3.16792488e-01 5.85371494e-01 7.10158944e-02 4.48837318e-02 -6.47540927e-01 -2.17547953e-01 7.17730820e-01 8.03719223e-01 -4.16074157e-01 2.67752260e-03 1.80973634e-02 5.64064085e-01 -4.19083145e-03 2.87466254e-02 -7.65642703e-01 -5.98007202e-01 7.65531659e-01 5.27031273e-02 -1.92378014e-01 -9.19448286e-02 -7.88147807e-01 -5.97517848e-01 -2.54999638e-01 -9.83110130e-01 2.39147976e-01 -1.03124917e+00 -1.21934056e+00 4.86259401e-01 -9.42381993e-02 -1.63255072e+00 -9.43834543e-01 -6.06912911e-01 -5.68919182e-01 6.05030060e-01 -3.52461666e-01 -7.91076362e-01 -2.15521395e-01 9.43636373e-02 4.71223205e-01 -2.81850696e-01 8.39505315e-01 1.26276761e-01 -2.06141204e-01 7.33531177e-01 -2.35052824e-01 1.11097507e-01 8.08622241e-01 -1.52618790e+00 5.46028376e-01 4.12304372e-01 4.61199164e-01 2.51530528e-01 1.13020611e+00 -3.05201709e-01 -9.60635066e-01 -8.10746670e-01 3.86703432e-01 -9.67284083e-01 7.23055661e-01 -3.16249281e-01 -4.64279383e-01 4.71337110e-01 4.16402072e-02 -6.61270261e-01 8.49994659e-01 4.12821919e-01 -5.66078275e-02 7.63806254e-02 -6.32050455e-01 5.45418501e-01 1.62644529e+00 -7.20557272e-01 -8.10800612e-01 1.72312796e-01 9.34556305e-01 -6.07446253e-01 -1.45502949e+00 5.57971299e-01 8.50400686e-01 -1.01061237e+00 8.02833140e-01 -5.99132299e-01 8.10182154e-01 -1.28027678e-01 -1.72191799e-01 -1.84409773e+00 -2.45876893e-01 -9.79457855e-01 -1.33002102e-01 1.64670992e+00 9.16221619e-01 -8.38202536e-02 4.45428520e-01 8.74600038e-02 -6.03709579e-01 -6.32671833e-01 -8.33787501e-01 -8.29153776e-01 2.08147541e-01 -1.08986509e+00 5.80053568e-01 7.04749525e-01 5.40347099e-01 7.17051506e-01 -3.49327087e-01 -5.28655350e-01 5.19769415e-02 -1.14867248e-01 8.68830025e-01 -1.30382395e+00 -9.06709969e-01 -8.48230481e-01 -5.43948472e-01 -5.61066866e-01 -1.09575033e-01 -8.02458227e-01 2.54082620e-01 -1.22645891e+00 -2.11015597e-01 -4.45553154e-01 2.28287606e-03 4.32278574e-01 6.39016926e-02 8.07691336e-01 2.41484225e-01 1.46285132e-01 -1.88813969e-01 3.77523184e-01 1.08142900e+00 -6.11794144e-02 -5.90430439e-01 3.73849794e-02 -7.85707772e-01 7.57408857e-01 6.59740746e-01 -2.60993749e-01 -4.50229287e-01 -5.32178164e-01 4.08015311e-01 2.64769167e-01 6.88731521e-02 -1.40934563e+00 -3.64287883e-01 1.51452720e-01 6.29313052e-01 -2.59599090e-01 5.04759908e-01 1.59062803e-01 3.13789994e-01 2.68919587e-01 -7.44816065e-01 2.85361439e-01 4.15823579e-01 2.49686614e-01 -2.65413493e-01 -2.35921457e-01 4.40996706e-01 1.91931203e-01 -2.53612339e-01 -2.62694478e-01 -6.78680122e-01 3.85260940e-01 4.32184637e-01 -2.90048838e-01 -4.60509628e-01 -7.29969621e-01 -9.37827826e-01 -2.29881823e-01 3.22501183e-01 6.19744599e-01 1.32702857e-01 -1.51155961e+00 -8.38333011e-01 -1.10830083e-01 2.63972014e-01 -2.03623593e-01 -1.80058390e-01 7.27729082e-01 -5.92881799e-01 3.61232907e-01 -1.33699596e-01 -5.47173262e-01 -9.50316727e-01 -8.22284371e-02 7.41028488e-02 -9.03743058e-02 -5.63554347e-01 1.10871696e+00 -2.55341262e-01 -3.32255781e-01 2.25304812e-01 -5.58326006e-01 -1.45067543e-01 3.59319061e-01 3.02767605e-01 6.33168519e-01 3.76921177e-01 -6.24453425e-01 -1.80542730e-02 3.12031120e-01 7.14628220e-01 -8.20975184e-01 1.27096057e+00 2.04413310e-01 4.31181788e-01 9.41042364e-01 9.37788248e-01 4.31525707e-01 -1.20787632e+00 5.52533031e-01 4.38827686e-02 -3.95446926e-01 1.15574496e-02 -9.05907094e-01 -5.19376636e-01 8.60781670e-01 4.02646244e-01 3.37490678e-01 1.00881600e+00 -4.86642458e-02 6.64371252e-01 1.40299171e-01 2.30229586e-01 -1.60884893e+00 5.36453664e-01 5.11162877e-01 1.17340910e+00 -4.86992002e-01 -4.01046008e-01 -6.92243129e-02 -8.53412271e-01 8.73818696e-01 7.09351420e-01 8.14558119e-02 1.09104529e-01 5.46196103e-01 -2.78050452e-03 3.12553287e-01 -1.14338660e+00 -1.03726722e-01 3.23421299e-01 8.71733904e-01 8.41776431e-01 1.10322617e-01 -2.50618100e-01 1.15096664e+00 -1.51761544e+00 -1.53290406e-01 3.70757699e-01 4.76058424e-01 -2.95584321e-01 -8.43077004e-01 -5.24528503e-01 7.81586945e-01 -5.56680083e-01 1.78396791e-01 -5.34773588e-01 7.19812155e-01 3.10355127e-01 1.38900673e+00 3.06480885e-01 -1.16118562e+00 4.29955482e-01 2.67810732e-01 6.74006343e-01 -6.34169579e-01 -8.98161232e-01 1.45243466e-01 8.43564987e-01 -3.47866237e-01 -1.73395872e-02 -6.74792349e-01 -1.03075731e+00 -2.88645566e-01 -2.92295247e-01 1.47383049e-01 4.17448789e-01 5.82342267e-01 2.13280022e-02 7.84402192e-01 5.77867508e-01 -1.24514258e+00 -4.75435883e-01 -1.50548744e+00 -5.96474409e-01 6.53121710e-01 8.12163502e-02 -6.77553356e-01 -5.03411591e-01 2.23356664e-01]
[16.041061401367188, 5.580324172973633]
aa8ecece-1bb4-4f59-a357-c24243282b08
graph-based-thermal-inertial-slam-with
2104.07196
null
https://arxiv.org/abs/2104.07196v3
https://arxiv.org/pdf/2104.07196v3.pdf
Graph-based Thermal-Inertial SLAM with Probabilistic Neural Networks
Simultaneous Localization and Mapping (SLAM) system typically employ vision-based sensors to observe the surrounding environment. However, the performance of such systems highly depends on the ambient illumination conditions. In scenarios with adverse visibility or in the presence of airborne particulates (e.g. smoke, dust, etc.), alternative modalities such as those based on thermal imaging and inertial sensors are more promising. In this paper, we propose the first complete thermal-inertial SLAM system which combines neural abstraction in the SLAM front end with robust pose graph optimization in the SLAM back end. We model the sensor abstraction in the front end by employing probabilistic deep learning parameterized by Mixture Density Networks (MDN). Our key strategies to successfully model this encoding from thermal imagery are the usage of normalized 14-bit radiometric data, the incorporation of hallucinated visual (RGB) features, and the inclusion of feature selection to estimate the MDN parameters. To enable a full SLAM system, we also design an efficient global image descriptor which is able to detect loop closures from thermal embedding vectors. We performed extensive experiments and analysis using three datasets, namely self-collected ground robot and handheld data taken in indoor environment, and one public dataset (SubT-tunnel) collected in underground tunnel. Finally, we demonstrate that an accurate thermal-inertial SLAM system can be realized in conditions of both benign and adverse visibility.
['Pedro P. B. de Gusmao', 'Chris Xiaoxuan Lu', 'Niki Trigoni', 'Andrew Markham', 'Bing Wang', 'Muhamad Risqi U. Saputra']
2021-04-15
null
null
null
null
['probabilistic-deep-learning']
['computer-vision']
[ 2.63460696e-01 -4.02663648e-01 2.44057328e-01 -3.63918304e-01 -6.95165932e-01 -4.58249778e-01 6.95525229e-01 -1.16066955e-01 -7.69651294e-01 6.68816805e-01 -1.79110289e-01 -1.19199015e-01 -3.24698240e-01 -7.08266497e-01 -9.15307760e-01 -8.49708557e-01 -2.16240883e-02 6.41700268e-01 -1.27069205e-01 -6.04902394e-02 2.06729040e-01 6.05963647e-01 -1.58776700e+00 -5.39893746e-01 8.26794386e-01 1.09493017e+00 6.69282615e-01 5.95990598e-01 4.10793759e-02 2.43242770e-01 -2.20602989e-01 3.21569204e-01 4.98921245e-01 1.91723421e-01 -2.38220543e-01 1.46320358e-01 4.39133823e-01 -2.31808156e-01 -2.56495535e-01 1.00582421e+00 5.77638030e-01 2.49509975e-01 6.63712025e-01 -1.23022509e+00 -1.74961269e-01 -2.17152283e-01 -3.44130784e-01 -3.21601689e-01 4.86086160e-01 2.77497798e-01 6.46132052e-01 -8.92269969e-01 5.14879882e-01 1.04558778e+00 7.88119078e-01 8.29649270e-02 -1.23965728e+00 -3.58301520e-01 -2.40614101e-01 2.66872525e-01 -1.76786685e+00 -4.22694415e-01 8.07588339e-01 -4.16250378e-01 8.03348303e-01 1.73938021e-01 6.59434676e-01 1.39036870e+00 6.49924695e-01 1.80420741e-01 1.32494414e+00 -3.93455416e-01 5.33866346e-01 2.31164709e-01 -2.79584289e-01 8.68486404e-01 4.29773986e-01 2.19762757e-01 -7.75395095e-01 -2.51368880e-01 4.71805930e-01 2.02317476e-01 -4.07376975e-01 -9.47729349e-01 -1.11738515e+00 6.97576225e-01 7.63705730e-01 3.95048931e-02 -5.69201350e-01 4.16291088e-01 -8.46283883e-02 4.67216745e-02 2.14072347e-01 3.22166592e-01 -1.94305077e-01 -6.38144731e-04 -9.12074089e-01 -1.50484711e-01 6.67619884e-01 8.58558178e-01 1.30272233e+00 -4.05848883e-02 4.92794484e-01 4.51034665e-01 8.46753120e-01 1.07519066e+00 4.72533315e-01 -8.33816707e-01 4.13442016e-01 3.67579669e-01 2.73502707e-01 -1.07543051e+00 -6.38992786e-01 -5.43442070e-01 -8.97564769e-01 2.93093711e-01 -1.59498975e-01 -2.38281060e-02 -1.05847561e+00 1.52336848e+00 4.27080482e-01 2.72902995e-01 1.45252913e-01 9.27741766e-01 2.11257517e-01 3.74267519e-01 -2.91452467e-01 5.26870228e-02 1.18105733e+00 -4.45240319e-01 -5.41089773e-01 -5.45753062e-01 6.51959538e-01 -6.69819176e-01 8.36739659e-01 2.49559507e-01 -1.72934771e-01 -5.74169993e-01 -1.38494658e+00 5.59453182e-02 -6.21255517e-01 2.15821609e-01 4.65707213e-01 6.54604375e-01 -1.22293818e+00 4.09098238e-01 -1.19805968e+00 -7.70909190e-01 -1.02966309e-01 5.10984600e-01 -5.99706173e-01 -3.56360942e-01 -1.06204331e+00 1.10089266e+00 2.85833925e-01 6.62000179e-01 -9.27643776e-01 -6.88875392e-02 -1.21996880e+00 -4.57768023e-01 7.82251805e-02 -1.04239404e+00 4.48280364e-01 -4.00951028e-01 -1.36817002e+00 5.04167974e-01 -2.40472868e-01 -4.90260333e-01 4.70659196e-01 -4.47295874e-01 -1.19365491e-01 5.29629849e-02 1.32164687e-01 6.02077305e-01 1.00860882e+00 -1.40162802e+00 -2.18130797e-01 -7.15814471e-01 -1.38174042e-01 5.73329210e-01 -1.72868386e-01 -6.22871578e-01 -1.21641271e-01 1.91111341e-01 6.84538722e-01 -1.46901226e+00 -3.75764579e-01 -5.53835258e-02 -4.03596431e-01 4.10573363e-01 8.16210985e-01 -5.34958720e-01 4.37459618e-01 -2.16975164e+00 1.71067715e-01 2.88182765e-01 -3.08699887e-02 -3.33838135e-01 7.17892572e-02 3.93174529e-01 3.35751444e-01 -2.51165926e-01 -2.56663412e-01 -9.49683368e-01 1.82768509e-01 5.36673129e-01 -1.22298405e-01 1.13669324e+00 -2.55329221e-01 6.27818286e-01 -6.16294444e-01 -4.58893269e-01 6.90845549e-01 5.53832412e-01 -2.43571684e-01 5.56099527e-02 5.27524315e-02 7.79451013e-01 -2.88038760e-01 5.52384555e-01 7.44995296e-01 2.51383066e-01 -1.91672504e-01 -3.47132742e-01 -3.92218351e-01 1.14621945e-01 -1.25290239e+00 2.24845910e+00 -9.82989430e-01 5.36622405e-01 3.67687881e-01 -4.50092405e-01 1.04760778e+00 4.25786003e-02 2.81416923e-01 -6.49416506e-01 1.50918379e-01 3.75320047e-01 -5.34859896e-01 -3.45053881e-01 8.72504711e-01 -4.40170951e-02 -1.03316747e-01 5.38976975e-02 -5.71410283e-02 -3.82151216e-01 -4.88524258e-01 -2.03199834e-02 1.01008153e+00 3.11826169e-01 9.44795087e-02 -1.30774006e-01 4.41086888e-01 -3.33927609e-02 3.36875647e-01 8.44978869e-01 -1.60067692e-01 5.73311567e-01 -3.45037907e-01 -3.43983442e-01 -8.43130767e-01 -1.16364634e+00 -3.24470736e-02 5.35354555e-01 4.61872041e-01 -1.02240555e-01 -1.56552047e-01 -2.93147773e-01 -7.13765547e-02 4.53010052e-01 -2.95926392e-01 -2.43532389e-01 -2.20999315e-01 -7.66156197e-01 2.47030050e-01 3.00955866e-02 7.35429168e-01 -5.02580583e-01 -7.18835235e-01 1.21124990e-01 -3.25285196e-01 -1.24520409e+00 2.05875397e-01 5.54123282e-01 -7.86565900e-01 -7.72491097e-01 -1.38623536e-01 -2.17436969e-01 5.03149033e-01 4.33206260e-01 4.19121861e-01 -4.39553261e-01 -3.67480993e-01 7.42161870e-01 -1.58537224e-01 -1.87748984e-01 -1.06580190e-01 1.25616193e-01 6.20946288e-01 1.98777676e-01 -3.22515443e-02 -7.86482215e-01 -4.67214555e-01 1.11948341e-01 -6.44802153e-01 -4.40195277e-02 7.31957495e-01 5.60952961e-01 5.10164499e-01 9.18619931e-02 -2.88580626e-01 -1.28869548e-01 1.66691512e-01 -4.17565137e-01 -7.44236767e-01 -2.74610668e-02 -5.28540671e-01 1.55733719e-01 3.52561235e-01 -8.06223508e-03 -7.10889578e-01 6.30268574e-01 -1.47976011e-01 -3.86998296e-01 -3.79867226e-01 5.11176586e-01 -2.62596041e-01 -5.87822556e-01 7.18660057e-01 5.05815685e-01 -8.52089003e-02 -4.90465909e-01 3.51780891e-01 7.25022614e-01 4.75727886e-01 -4.68171895e-01 1.25022507e+00 8.86532068e-01 4.71233964e-01 -1.20336986e+00 -3.31802815e-01 -6.30364120e-01 -7.46449232e-01 -2.55293876e-01 8.28604877e-01 -1.16425276e+00 -5.32344401e-01 3.49098444e-01 -1.05980003e+00 -9.68353823e-02 1.42696604e-01 9.00710940e-01 -6.35207236e-01 6.25433683e-01 -1.39255643e-01 -1.06091225e+00 -1.95365340e-01 -1.24793088e+00 1.33473182e+00 9.71176550e-02 4.92094178e-03 -9.36588466e-01 2.26243943e-01 3.65585983e-01 4.37366754e-01 3.37632686e-01 4.00537252e-01 -1.95199117e-01 -9.97618973e-01 -2.63714343e-01 2.84740850e-02 2.91649878e-01 1.77880928e-01 -3.88203502e-01 -1.10754979e+00 -4.75963116e-01 3.60754162e-01 -1.81714743e-01 8.84939373e-01 3.20666730e-01 3.92280579e-01 4.50636856e-02 -3.80902618e-01 1.03815949e+00 1.75637412e+00 -1.15702808e-01 3.41596693e-01 5.72761536e-01 1.02008080e+00 4.70934927e-01 7.48770058e-01 3.63529950e-01 4.77866828e-01 6.25650644e-01 1.11941540e+00 8.57941583e-02 3.45870852e-01 -1.26298442e-01 5.07204294e-01 6.48222804e-01 1.41427532e-01 -1.36185646e-01 -8.66769493e-01 4.16984886e-01 -1.77681577e+00 -2.61072248e-01 -5.49738780e-02 2.40084410e+00 1.98179912e-02 1.43391564e-01 -6.33877456e-01 3.20305414e-02 5.04717290e-01 3.58833224e-01 -3.45093399e-01 -1.03048675e-01 -1.00350410e-01 -8.35784450e-02 1.03516734e+00 8.55870366e-01 -1.17155123e+00 7.71878362e-01 4.70056295e+00 4.44769144e-01 -1.08728206e+00 1.85849205e-01 -2.64923096e-01 1.66233793e-01 -2.40401164e-01 2.24128678e-01 -6.70784950e-01 2.23704711e-01 1.09235930e+00 5.34564555e-01 4.93283212e-01 8.30122650e-01 3.16249579e-01 -5.22481859e-01 -7.32772470e-01 1.09857607e+00 3.08934078e-02 -8.05304408e-01 -3.37126881e-01 4.67686623e-01 3.17761242e-01 6.61690176e-01 1.99780047e-01 1.97386164e-02 2.28959359e-02 -6.70794606e-01 7.49912798e-01 6.31724298e-01 5.55688083e-01 -5.46962440e-01 7.22303808e-01 5.69739819e-01 -1.16732335e+00 2.27778722e-02 -3.36558372e-01 -1.45701408e-01 1.58166900e-01 8.06288779e-01 -1.33087587e+00 9.78260934e-01 5.39630175e-01 5.40488303e-01 -5.96253395e-01 9.85298753e-01 -3.04992825e-01 9.49501842e-02 -9.05345321e-01 -4.25217748e-02 3.06883693e-01 -3.14177662e-01 7.12257445e-01 8.45697939e-01 7.56255031e-01 -6.49790883e-01 3.24852616e-01 7.96323717e-01 4.32113975e-01 -1.82274505e-01 -1.13183689e+00 4.36245680e-01 4.40483868e-01 1.34486747e+00 -4.97016400e-01 4.89955470e-02 -7.99875408e-02 1.23902774e+00 4.47512455e-02 4.95334625e-01 -8.13593924e-01 -1.06802657e-01 8.19501638e-01 -1.33470237e-01 -3.74112688e-02 -1.12228549e+00 4.56135236e-02 -1.03194273e+00 5.17928861e-02 -2.62252599e-01 -3.72603565e-01 -1.02213740e+00 -7.47968674e-01 5.04704058e-01 -1.84081376e-01 -1.24922824e+00 -2.32390210e-01 -5.96191287e-01 -6.85697868e-02 1.00199175e+00 -1.48751545e+00 -1.30845356e+00 -6.31720483e-01 6.73656642e-01 7.86913410e-02 2.84747511e-01 8.22996378e-01 2.16841146e-01 -3.44103396e-01 -2.09767565e-01 3.17329377e-01 -3.31045568e-01 7.22556233e-01 -1.20646501e+00 2.51004487e-01 9.48449850e-01 4.09686208e-01 6.76438928e-01 9.82948363e-01 -6.65710747e-01 -1.89478493e+00 -1.23411822e+00 6.33580565e-01 -3.45040441e-01 5.93822300e-01 -7.45686293e-01 -4.85693276e-01 6.55533671e-01 -1.67318150e-01 1.02155153e-02 2.32850283e-01 -9.63361710e-02 -4.62552868e-02 -2.89440036e-01 -1.05740499e+00 4.01979774e-01 7.67259240e-01 -9.28164661e-01 -4.09002572e-01 4.80029136e-01 6.37086391e-01 -4.86947775e-01 -5.55152893e-01 5.27617991e-01 5.44049382e-01 -1.02208364e+00 9.71507668e-01 2.80581087e-01 -4.47830915e-01 -6.04775310e-01 -6.69673741e-01 -1.29249525e+00 -1.16925918e-01 -4.00575399e-01 1.18294358e-01 9.95243192e-01 -7.34328404e-02 -9.52300787e-01 7.69406676e-01 2.58206517e-01 -1.50775209e-01 -1.71249405e-01 -1.40120316e+00 -8.92315686e-01 -7.62677073e-01 -7.67666340e-01 2.99391091e-01 5.25951147e-01 -6.39065087e-01 2.45336190e-01 -6.02450490e-01 9.79334414e-01 9.23383176e-01 -1.30163923e-01 8.57295334e-01 -1.23525465e+00 -4.50802706e-02 2.33290642e-01 -7.42669165e-01 -8.53905320e-01 3.03322047e-01 -6.80494368e-01 4.89626676e-01 -1.66691160e+00 -2.31898591e-01 -4.62945968e-01 -2.38496691e-01 8.87164623e-02 3.93440336e-01 2.02184498e-01 -4.98884395e-02 2.98100948e-01 -5.15252590e-01 8.62274289e-01 5.87419391e-01 -2.45595649e-02 -9.58324298e-02 -4.43734601e-02 5.39299697e-02 7.15833187e-01 6.45510972e-01 -5.32133102e-01 -3.04356158e-01 -5.85744858e-01 3.60043019e-01 7.48552289e-03 7.11886108e-01 -1.61246169e+00 5.27020872e-01 -2.16475967e-02 2.15139344e-01 -6.69058144e-01 1.01294005e+00 -1.20899940e+00 3.47083300e-01 6.68168485e-01 2.99576610e-01 1.23362832e-01 -9.93168801e-02 9.99324620e-01 -1.02015257e-01 -1.42989680e-01 5.15851140e-01 -1.09113827e-01 -9.83023763e-01 2.65399516e-01 -3.23882282e-01 -8.40307295e-01 5.90228975e-01 -2.87440240e-01 -9.98796895e-02 -4.20219272e-01 -5.40118694e-01 7.45422542e-02 8.77704203e-01 3.27597439e-01 7.27735460e-01 -1.19130290e+00 -2.47970447e-01 4.96273369e-01 3.68974298e-01 9.82424542e-02 2.64457852e-01 1.09457338e+00 -7.21946239e-01 4.92859751e-01 -7.87962154e-02 -9.97529268e-01 -1.02382135e+00 3.43265414e-01 3.53725135e-01 2.80903727e-02 -2.56224781e-01 6.29550040e-01 -4.33150046e-02 -7.14292705e-01 4.14277762e-02 -4.56099600e-01 2.79439747e-01 -1.69630006e-01 -5.42053021e-02 2.53401995e-01 2.32375070e-01 -9.27956879e-01 -7.77493536e-01 8.51229727e-01 5.32113314e-01 -5.32731116e-01 1.02984786e+00 -7.63174593e-01 -2.54842252e-01 6.58463657e-01 1.41831696e+00 -1.67714828e-03 -1.06246388e+00 -1.54948086e-01 -1.16966300e-01 -2.54376113e-01 2.40387231e-01 -2.41929814e-01 -4.70207125e-01 8.74047756e-01 1.00999570e+00 -1.46963477e-01 9.86078322e-01 -3.30576926e-01 4.74073589e-01 7.23687887e-01 1.00429010e+00 -9.12357628e-01 -1.74576372e-01 6.80664539e-01 6.39716029e-01 -1.28296447e+00 1.03026040e-01 -6.79246634e-02 -2.59303182e-01 8.99076819e-01 2.04213724e-01 -9.63360220e-02 5.29524505e-01 3.09341326e-02 1.45554826e-01 -1.28144339e-01 -3.62312406e-01 -4.13886577e-01 5.80171868e-02 6.38362706e-01 -2.17861444e-01 1.19160190e-01 2.34905422e-01 -1.62520543e-01 -2.31818289e-01 -3.37037265e-01 4.00423944e-01 1.12337303e+00 -6.11615598e-01 -7.65120804e-01 -8.37993085e-01 -3.42089124e-02 1.29392684e-01 -1.71910059e-02 -7.13132918e-02 7.59199679e-01 4.02365655e-01 8.47671986e-01 -2.21457824e-01 -7.76292145e-01 4.59976047e-02 2.36924961e-01 6.34248972e-01 -4.34349507e-01 -2.21050382e-02 -2.82423310e-02 -3.30178924e-02 -7.81429827e-01 -3.74973565e-01 -5.89497089e-01 -9.32472944e-01 6.56371042e-02 -2.97304720e-01 -4.42657173e-02 1.73629272e+00 1.05392325e+00 3.20606917e-01 3.11958492e-01 6.29818499e-01 -1.29359138e+00 -4.26263779e-01 -9.55861568e-01 -9.43314075e-01 -1.57685846e-01 7.49135494e-01 -9.09576416e-01 -6.21670127e-01 -3.94323915e-01]
[7.486794471740723, -2.182952880859375]
06ec1beb-f9f0-4d8d-93b2-471c890d303f
set-supervised-action-learning-in-procedural
null
null
http://openaccess.thecvf.com//content/CVPR2022/html/Lu_Set-Supervised_Action_Learning_in_Procedural_Task_Videos_via_Pairwise_Order_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Lu_Set-Supervised_Action_Learning_in_Procedural_Task_Videos_via_Pairwise_Order_CVPR_2022_paper.pdf
Set-Supervised Action Learning in Procedural Task Videos via Pairwise Order Consistency
We address the problem of set-supervised action learning, whose goal is to learn an action segmentation model using weak supervision in the form of sets of actions occurring in training videos. Our key observation is that videos within the same task have similar ordering of actions, which can be leveraged for effective learning. Therefore, we propose an attention-based method with a new Pairwise Ordering Consistency (POC) loss that encourages that for each common action pair in two videos of the same task, the attentions of actions follow a similar ordering. Unlike existing sequence alignment methods, which misalign actions in videos with different orderings or cannot reliably separate more from less consistent orderings, our POC loss efficiently aligns videos with different action orders and is differentiable, which enables end-to-end training. In addition, it avoids the time-consuming pseudo-label generation of prior works. Our method efficiently learns the actions and their temporal locations, therefore, extends the existing attention-based action localization methods from learning one action per video to multiple actions using our POC loss along with video-level and frame-level losses. By experiments on three datasets, we demonstrate that our method significantly improves the state of the art. We also show that our method, with a small modification, can effectively address the transcript-supervised action learning task, where actions and their ordering are available during training.
['Ehsan Elhamifar', 'Zijia Lu']
2022-01-01
null
null
null
cvpr-2022-1
['action-localization', 'weakly-supervised-action-segmentation-action', 'action-segmentation']
['computer-vision', 'computer-vision', 'computer-vision']
[ 5.74946582e-01 -1.78540558e-01 -5.63403845e-01 -3.65287334e-01 -8.04030716e-01 -5.63092709e-01 4.30100620e-01 -2.54324198e-01 -4.99763727e-01 5.22633433e-01 4.40705955e-01 6.70696124e-02 -8.88342708e-02 -1.46358043e-01 -9.97069180e-01 -7.57022083e-01 -1.83779210e-01 2.98221916e-01 4.20000643e-01 2.49204636e-01 1.91343904e-01 1.47808045e-01 -1.26526153e+00 4.49750781e-01 6.21130288e-01 1.01422811e+00 3.11755180e-01 5.48235297e-01 2.16444373e-01 1.31223607e+00 -2.97858745e-01 -8.56479406e-02 5.84349990e-01 -7.53067791e-01 -1.09487784e+00 6.78440571e-01 8.08862925e-01 -8.26007366e-01 -6.03509068e-01 9.76631403e-01 3.42718139e-02 4.36527491e-01 4.58413601e-01 -1.52402878e+00 -5.16587555e-01 4.59164053e-01 -7.14213669e-01 2.04586267e-01 2.92821705e-01 2.94739157e-01 1.27429676e+00 -4.99651670e-01 5.37210584e-01 1.24071538e+00 4.43660825e-01 4.45557833e-01 -1.04687941e+00 -6.41623080e-01 6.15845799e-01 5.49450159e-01 -1.10207820e+00 -3.43613952e-01 6.96510077e-01 -4.58069682e-01 8.43282580e-01 -1.37578145e-01 6.02728426e-01 1.01662064e+00 -1.38312727e-01 1.18433487e+00 6.87748849e-01 -2.56580889e-01 1.10760689e-01 -5.84014893e-01 -1.59667403e-01 8.12301934e-01 -5.88480569e-02 -2.85066396e-01 -6.31496489e-01 1.35897964e-01 9.94773626e-01 4.86448526e-01 -3.17493141e-01 -7.09193528e-01 -1.55133653e+00 6.67705655e-01 2.73722768e-01 2.12354347e-01 -3.55229110e-01 6.11192703e-01 4.92469072e-01 1.29500479e-01 2.38367587e-01 2.27290377e-01 -6.29758954e-01 -4.11170453e-01 -6.96599066e-01 1.62277669e-01 4.16264772e-01 1.05755341e+00 8.73924673e-01 -3.39093596e-01 -3.51867169e-01 6.05325401e-01 2.50259161e-01 2.79659182e-01 4.35312599e-01 -1.65796721e+00 7.02565074e-01 5.73891461e-01 4.13265109e-01 -8.09274435e-01 3.64263915e-03 9.95615423e-02 -4.20294195e-01 -3.76008600e-02 5.87035120e-01 -6.43261746e-02 -9.43255901e-01 2.00303602e+00 3.64813536e-01 8.06659222e-01 -1.77421972e-01 9.22796011e-01 1.30330533e-01 3.87153983e-01 2.67338008e-02 -2.24328354e-01 1.08084393e+00 -1.56779993e+00 -6.57701790e-01 -3.70618731e-01 8.80530834e-01 -4.71071124e-01 1.11810219e+00 1.37250453e-01 -9.44478095e-01 -5.26174784e-01 -7.87187040e-01 -1.03271008e-01 2.62441754e-01 1.17768347e-01 6.97175145e-01 2.53678467e-02 -8.25266659e-01 8.13087404e-01 -1.19847667e+00 -1.99676901e-01 8.37111294e-01 2.77648956e-01 -4.47085798e-01 -2.64964998e-01 -8.55818927e-01 4.02766734e-01 3.38415533e-01 -1.19885385e-01 -1.14110196e+00 -7.11145580e-01 -1.07239330e+00 -2.64323372e-02 9.45578992e-01 -4.24307644e-01 1.46761024e+00 -1.56098783e+00 -1.58283174e+00 6.09045684e-01 -2.92241573e-01 -6.38270676e-01 4.32007223e-01 -7.34574139e-01 8.43821019e-02 5.20963609e-01 4.36075389e-01 8.83298814e-01 8.91002893e-01 -8.93535078e-01 -1.11877143e+00 -9.36168507e-02 4.63267267e-01 4.04872179e-01 -3.21213067e-01 1.30523399e-01 -9.57971931e-01 -5.27915835e-01 -7.02359676e-02 -1.07016802e+00 -2.39509925e-01 3.65817577e-01 -1.49887830e-01 -2.71276653e-01 7.58923233e-01 -5.87976813e-01 9.68711615e-01 -2.28559828e+00 4.73980188e-01 -1.54328108e-01 1.03599511e-01 1.18742660e-01 -3.36475492e-01 2.13012800e-01 -6.46431372e-03 -5.96540533e-02 -4.83011365e-01 -5.33364296e-01 -1.02637887e-01 5.32860279e-01 -2.82121450e-01 5.71576118e-01 3.15102637e-01 8.95095348e-01 -1.35256958e+00 -6.61225617e-01 2.08725169e-01 2.43527070e-01 -7.35795498e-01 3.92827272e-01 -4.00469661e-01 8.13392580e-01 -4.66797948e-01 5.39950311e-01 2.15829030e-01 -5.70934832e-01 3.50504071e-01 -1.54950693e-01 1.79031536e-01 2.55393714e-01 -1.10647583e+00 1.93207192e+00 -3.02101403e-01 5.66345096e-01 -1.72381952e-01 -1.30406117e+00 2.68323064e-01 2.90707260e-01 1.04858661e+00 -5.54908633e-01 -1.39218524e-01 -1.32293557e-03 3.76560129e-02 -6.31049812e-01 4.73152958e-02 1.38576284e-01 7.70629272e-02 6.91738307e-01 2.31016919e-01 3.53250325e-01 5.13694346e-01 2.19859317e-01 1.34299266e+00 6.52051151e-01 3.49375814e-01 2.88837671e-01 4.06039923e-01 -2.59241939e-01 8.89707088e-01 7.08587289e-01 -5.20995378e-01 5.52749515e-01 6.51194394e-01 -3.83497953e-01 -8.88543606e-01 -7.11268723e-01 1.96581423e-01 1.35102081e+00 3.08042586e-01 -3.85378510e-01 -6.86558068e-01 -1.25306225e+00 -8.00521970e-02 1.48235962e-01 -5.26462078e-01 8.21897164e-02 -8.86585593e-01 -1.52783722e-01 2.46765256e-01 9.29076135e-01 6.41015232e-01 -1.16125047e+00 -6.38175130e-01 1.84639171e-01 -4.24379736e-01 -1.50277781e+00 -1.11371589e+00 -2.11958140e-02 -8.32394123e-01 -1.44363332e+00 -6.67879641e-01 -8.39862227e-01 8.68688405e-01 5.91597080e-01 9.88709867e-01 1.85735583e-01 1.07182696e-01 6.89908147e-01 -6.28367245e-01 6.94682896e-02 -3.61735672e-02 -2.22742021e-01 -9.64569487e-03 3.61095041e-01 3.02687317e-01 -4.82764930e-01 -7.59077072e-01 5.33096969e-01 -8.41833115e-01 1.43167377e-01 6.37307644e-01 8.05984795e-01 8.24721575e-01 -6.67526051e-02 2.24629551e-01 -7.24613070e-01 -1.72671929e-01 -3.38256866e-01 -3.86747628e-01 3.21567208e-01 -5.35074398e-02 5.39583825e-02 7.08533883e-01 -4.81154561e-01 -8.26886237e-01 4.33723152e-01 3.89744759e-01 -9.60005164e-01 -2.38798589e-01 1.85544431e-01 -3.72394294e-01 3.62171931e-03 7.09512457e-02 1.70071542e-01 8.92352685e-02 -3.69379729e-01 3.74079466e-01 2.75431007e-01 5.86208224e-01 -4.12031263e-01 6.73325241e-01 7.38841891e-01 -1.01713613e-01 -4.55345690e-01 -1.20410788e+00 -8.31024826e-01 -1.01757634e+00 -1.55691609e-01 1.01544654e+00 -1.01410973e+00 -5.91420352e-01 7.03613460e-01 -1.07862568e+00 -7.44255543e-01 -2.59786189e-01 6.64157629e-01 -9.90712345e-01 8.33240628e-01 -6.72738254e-01 -5.60965896e-01 1.74533725e-01 -1.19850671e+00 1.28447843e+00 -9.65839624e-03 -2.29754865e-01 -9.49878931e-01 5.93907274e-02 5.63781321e-01 -2.47547433e-01 1.01722367e-01 5.36166012e-01 -5.98901272e-01 -9.01864111e-01 1.43250808e-01 -1.26038581e-01 6.26650333e-01 5.44918835e-01 -1.56044796e-01 -4.48094547e-01 -4.02926505e-01 -9.61704180e-02 -5.72927117e-01 9.53055441e-01 4.50973928e-01 1.40568173e+00 -4.82119232e-01 -3.04243147e-01 6.07145548e-01 1.07275414e+00 2.68944353e-01 5.30921340e-01 2.81696320e-01 1.09308541e+00 5.92451453e-01 1.10242796e+00 3.39349300e-01 1.82923853e-01 1.00551891e+00 5.15119493e-01 -1.49294874e-02 -6.17873669e-02 -3.19949150e-01 6.95793033e-01 3.48808348e-01 -8.74970406e-02 -1.87311560e-01 -4.73774076e-01 5.53477347e-01 -2.37807369e+00 -1.23672819e+00 3.80844116e-01 2.23490691e+00 9.31565166e-01 4.67112847e-03 4.10192639e-01 -4.01861258e-02 6.73631966e-01 4.09337848e-01 -8.31147373e-01 2.07436025e-01 3.59140903e-01 4.63052988e-02 6.07552707e-01 5.38149595e-01 -1.60662401e+00 1.00912118e+00 5.84450626e+00 6.08366132e-01 -7.64866889e-01 6.71463609e-02 5.72209060e-01 -3.83950919e-01 -3.57450452e-03 8.64993408e-02 -6.55125439e-01 7.08587408e-01 5.38322210e-01 1.93544567e-01 5.47546268e-01 8.48151326e-01 3.79088461e-01 -5.27822413e-02 -1.57015634e+00 8.94086301e-01 1.96008548e-01 -1.20735979e+00 1.06885787e-02 2.71708821e-03 8.59809875e-01 -9.15458202e-02 -2.21008435e-01 2.08348528e-01 4.46813375e-01 -8.40666294e-01 7.66946316e-01 2.73036867e-01 4.83584791e-01 -5.34099102e-01 4.77850139e-01 7.63081163e-02 -1.35391295e+00 -3.27929825e-01 -1.61120027e-01 -2.10525319e-01 4.49156582e-01 1.02869339e-01 -4.60570008e-01 4.09965903e-01 6.80037439e-01 1.46955180e+00 -2.27804080e-01 8.88730586e-01 -4.49135393e-01 7.00636446e-01 -4.94072959e-02 3.16750765e-01 4.89457905e-01 -4.10229355e-01 3.12311441e-01 8.39218378e-01 3.48081663e-02 1.07974619e-01 6.96720660e-01 4.48287010e-01 -1.38498753e-01 -1.76434085e-01 -3.97706211e-01 -1.81746706e-01 5.09760439e-01 9.21156645e-01 -6.96756005e-01 -3.77830952e-01 -7.92625606e-01 1.32969296e+00 4.78451908e-01 4.51490909e-01 -1.17163098e+00 -4.88529056e-02 1.00398600e+00 -1.53307393e-02 7.43623853e-01 -2.63731480e-01 2.31342137e-01 -1.01598775e+00 2.95056492e-01 -1.00669324e+00 4.64153200e-01 -4.99720901e-01 -1.07980502e+00 -5.78193031e-02 1.37707517e-01 -1.64405704e+00 -2.67925262e-01 -4.78475451e-01 -4.83713418e-01 2.99915135e-01 -1.41683710e+00 -1.10786581e+00 -1.24684587e-01 6.03998542e-01 9.54723895e-01 -1.55869443e-02 3.06216300e-01 3.83323491e-01 -6.05220020e-01 3.97868305e-01 8.68050102e-03 4.04177070e-01 8.06147516e-01 -1.26830518e+00 1.74243182e-01 9.70913291e-01 4.95520145e-01 2.87957013e-01 2.66747802e-01 -5.73524296e-01 -1.14897668e+00 -1.31940770e+00 6.20196283e-01 -4.52365816e-01 7.64197230e-01 -1.59433767e-01 -8.48661900e-01 1.11098123e+00 8.68843198e-02 2.43112475e-01 6.31549895e-01 -1.24389842e-01 -2.68220276e-01 -7.14890286e-02 -6.59618378e-01 6.84462488e-01 1.62121654e+00 -4.51545954e-01 -4.26752418e-01 7.27648258e-01 8.55445206e-01 -3.86829525e-01 -6.35252059e-01 2.71664351e-01 4.10758346e-01 -9.75212872e-01 9.55218792e-01 -9.79712188e-01 6.93098187e-01 -4.25488621e-01 3.79382866e-03 -1.03149831e+00 -3.57440650e-01 -6.83542430e-01 -2.87911981e-01 1.06924593e+00 2.39306260e-02 -3.11085254e-01 6.76640034e-01 6.09231234e-01 -2.70209491e-01 -8.78916562e-01 -9.02005494e-01 -8.77924442e-01 -2.70721018e-01 -2.48559564e-01 3.28788579e-01 7.58841813e-01 -1.44263685e-01 1.03286326e-01 -7.40138531e-01 1.61021709e-01 5.47011197e-01 2.20011011e-01 9.05270398e-01 -7.56409049e-01 -6.98029637e-01 -3.55770618e-01 -5.15864432e-01 -1.57187402e+00 5.05394101e-01 -5.24142206e-01 3.90263796e-01 -1.34785271e+00 4.95244414e-01 -1.79148674e-01 -4.25265074e-01 8.65354240e-01 -3.92947704e-01 1.88402042e-01 2.44668797e-01 4.59825426e-01 -1.29716825e+00 5.48900843e-01 1.28606617e+00 -1.09152243e-01 -1.56352341e-01 -9.01410133e-02 -4.34636384e-01 9.48405206e-01 5.38140774e-01 -4.94776040e-01 -6.82270527e-01 -5.62386334e-01 -2.24651366e-01 -1.71052992e-01 4.01826471e-01 -9.83464539e-01 5.29868826e-02 -3.99485767e-01 7.49755576e-02 -2.27992281e-01 1.64432779e-01 -8.88387918e-01 -2.05741242e-01 3.86688173e-01 -6.30400658e-01 -2.10599363e-01 -2.30633900e-01 9.54698026e-01 -2.18503848e-01 -1.72769919e-01 6.71005368e-01 -1.14537843e-01 -1.03953457e+00 7.40552962e-01 -5.75646311e-02 3.55342150e-01 1.33463693e+00 -2.61396497e-01 3.00688948e-02 -4.44670111e-01 -4.87933218e-01 5.54824054e-01 6.22319221e-01 4.69063073e-01 4.02446210e-01 -1.38915634e+00 -4.14282590e-01 -8.59670490e-02 1.15536407e-01 2.19426796e-01 2.24315971e-01 1.07887232e+00 -2.75154710e-01 2.43023142e-01 -2.41521865e-01 -7.20469058e-01 -1.38855720e+00 5.26465237e-01 3.51281375e-01 -3.91696423e-01 -7.33955741e-01 8.59261274e-01 4.95009303e-01 -1.27654091e-01 5.29345214e-01 -2.91728407e-01 -1.93574548e-01 -1.82294026e-01 7.36683726e-01 3.16863328e-01 -4.10072148e-01 -6.67807102e-01 -3.64912361e-01 6.25118732e-01 -2.86153167e-01 -1.69142764e-02 1.23092258e+00 -1.01158470e-01 8.85611027e-02 4.28446412e-01 1.28175581e+00 -2.59085059e-01 -2.28591537e+00 -3.62589836e-01 -7.65736476e-02 -8.81353796e-01 -3.92629147e-01 -3.47682118e-01 -1.31485355e+00 5.63396573e-01 2.24428132e-01 -2.59384930e-01 1.08699238e+00 1.66524276e-01 9.80638385e-01 3.93272221e-01 3.62998068e-01 -1.18731356e+00 7.43393064e-01 5.00781357e-01 6.01463318e-01 -1.40758836e+00 -7.39452541e-02 -3.53998750e-01 -9.39822435e-01 7.55948901e-01 8.02745521e-01 -1.16105162e-01 3.54074925e-01 3.83385122e-02 -1.50201842e-01 -2.74704844e-02 -5.71017444e-01 -3.70942831e-01 2.04349175e-01 6.04381502e-01 2.71095365e-01 -2.41941065e-01 -2.27699578e-01 2.53907144e-01 5.29518545e-01 7.91573524e-02 2.10430533e-01 1.16512394e+00 -4.06819165e-01 -1.20675993e+00 -1.66464653e-02 3.97840381e-01 -3.59377652e-01 -6.81748753e-03 -3.85436565e-01 6.59968853e-01 1.13010876e-01 7.93978631e-01 3.28031361e-01 -2.62136430e-01 1.01938948e-01 2.50902921e-02 5.40678680e-01 -8.15832496e-01 -9.15177837e-02 5.20420223e-02 -7.94507787e-02 -1.13408756e+00 -1.01721847e+00 -9.28572476e-01 -1.50613260e+00 1.14744052e-01 -5.09981252e-02 -1.49925977e-01 -1.38336033e-01 1.30783880e+00 4.42537248e-01 3.78918618e-01 8.17502499e-01 -9.95692194e-01 -6.15700245e-01 -7.30420887e-01 -5.37240982e-01 1.00079143e+00 3.66206199e-01 -9.18903768e-01 -3.15588892e-01 5.28757215e-01]
[8.464563369750977, 0.7136480808258057]