text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
point. box the The an values and objectness each outputs bounding network for offset map
The network outputs an objectness map and the bounding box offset values for each point.
35,200
KITTI the evaluation the dataset quantitative evaluations help We with server. carried of out
We carried out quantitative evaluations with the help of the KITTI dataset evaluation server.
35,201
extract and spatial both for pose network features frames. we From the features image
From the spatial network we extract image features and pose features for both frames.
35,202
serve model that (TFF). Temporal temporal Fields These for as Flow input our predicts features
These features serve as input for our temporal model that predicts Temporal Flow Fields (TFF).
35,203
formulate representation similarity novel to joints. of This allows detected a measure
This novel representation allows to formulate a similarity measure of detected joints.
35,204
as driving, crucial estimation are autonomous pose technologies. self-localization/camera applications scene For and parsing such
For applications such as autonomous driving, self-localization/camera pose estimation and scene parsing are crucial technologies.
35,205
propose problems In paper, tackle two this these to framework a simultaneously. unified we
In this paper, we propose a unified framework to tackle these two problems simultaneously.
35,206
semantically-labeled. Both point are and the images the clouds
Both the point clouds and the images are semantically-labeled.
35,207
sensors. Each has highly from video frame motion ground truth pose accurate
Each video frame has ground truth pose from highly accurate motion sensors.
35,208
studies of proposed various effectiveness ablation performed, the the Finally, are system. demonstrate which
Finally, various ablation studies are performed, which demonstrate the effectiveness of the proposed system.
35,209
learning alter behavioral is perception. evidence and physiological that category can This
This is behavioral and physiological evidence that category learning can alter perception.
35,210
net this a We effect. neural model sketch for
We sketch a neural net model for this effect.
35,211
of training be Deep to large learning effective. data requires amounts
Deep learning requires large amounts of training data to be effective.
35,212
iteratively on training, currently predicted the based we errors of clicks add During the segmentation.
During training, we iteratively add clicks based on the errors of the currently predicted segmentation.
35,213
data these and an domain-specific expensive laborious to is Labeling names task. color learn
Labeling data to learn these domain-specific color names is an expensive and laborious task.
35,214
article learn aim labeled color from weakly data. names Therefore, to in we this
Therefore, in this article we aim to learn color names from weakly labeled data.
35,215
we For branch naming network. an color attention purpose, the this add to
For this purpose, we add an attention branch to the color naming network.
35,216
attention pixel-wise used The predictions modulate the to of branch the color is network. naming
The attention branch is used to modulate the pixel-wise color naming predictions of the network.
35,217
the illustrate experiments, relevant branch we the that attention regions. identifies In correctly
In experiments, we illustrate that the attention branch correctly identifies the relevant regions.
35,218
apply of model to self-consistency splices. detecting We task localizing this image and the
We apply this self-consistency model to the task of detecting and localizing image splices.
35,219
optimization experiences Thereby, techniques. important parameter report detailed initialization we concerning and
Thereby, we report important experiences concerning detailed parameter initialization and optimization techniques.
35,220
very they flexible constraints Inequality exact because prior not assume knowledge. are do
Inequality constraints are very flexible because they do not assume exact prior knowledge.
35,221
evaluation conditions. non-studio This us allows present qualitative to in and extension quantitative
This extension allows us to present quantitative and qualitative evaluation in non-studio conditions.
35,222
idea issues address challenging simple This style transfer can methods. existing the effectively in
This simple idea can effectively address the challenging issues in existing style transfer methods.
35,223
we approach. loss, also progressive a on proposed present optimization the feature-domain Based
Based on the proposed loss, we also present a progressive feature-domain optimization approach.
35,224
deadline increases The as bets very rapidly comes betting near. for the of amount
The amount of bets increases very rapidly as the deadline for betting comes near.
35,225
bets gives the bettor which value Each on largest of the expectation benefit. a horse
Each bettor bets on a horse which gives the largest expectation value of the benefit.
35,226
alternative be Using simulations, biased. we approximate-Bayesian (ABC) computation one to on find method based
Using simulations, we find one alternative method based on approximate-Bayesian computation (ABC) to be biased.
35,227
we for to argument, methods them. acquiring improve need and Yet, of apply models
Yet, to improve and apply models of argument, we need methods for acquiring them.
35,228
out fallen of post-processing more In works has favour. however, recent CRF
In more recent works however, CRF post-processing has fallen out of favour.
35,229
can CRFs All backpropagation. optimized using of parameters convolutional easily the be
All parameters of the convolutional CRFs can easily be optimized using backpropagation.
35,230
implementation further research CRF our publicly make facilitating we available. To
To facilitating further CRF research we make our implementation publicly available.
35,231
learning new \emph{e.g. very with a are Humans little of capable concept fine-grained supervision,
Humans are capable of learning a new fine-grained concept with very little supervision, \emph{e.g.
35,232
is collecting challenging. datasets annotated such However,
However, collecting such annotated datasets is challenging.
35,233
complete existing entries. new We databases database by propose to generating
We propose to complete existing databases by generating new database entries.
35,234
(GAN) the generative training algorithm process. networks adopts training adversarial Our
Our training algorithm adopts the generative adversarial networks (GAN) training process.
35,235
facial Robust localization partially are remains challenging landmark faces occluded. task when a
Robust facial landmark localization remains a challenging task when faces are partially occluded.
35,236
on of We dataset RICPR COFW. the evaluate challenging
We evaluate RICPR on the challenging dataset of COFW.
35,237
weights the then fine We propagation. role tune using player back
We then fine tune the player role weights using back propagation.
35,238
and The the interpretability hierarchical of ensures representation. integrity the the group architecture
The hierarchical architecture ensures the interpretability and the integrity of the group representation.
35,239
RMS error. values show of results an improvement of terms Our in prediction
Our results show an improvement in terms of RMS values of prediction error.
35,240
counting paper, In model this to a introduce (PMC). we novel solve algorithm projected
In this paper, we introduce a novel algorithm to solve projected model counting (PMC).
35,241
primal the small of algorithm treewidth input of instance. exploits Our graph the
Our algorithm exploits small treewidth of the primal graph of the input instance.
35,242
deep recently they been incorporated As architectures. such, into have
As such, they have recently been incorporated into deep architectures.
35,243
the kernel boost method MvCCDA. performance develop We a of to further MvCCDA of
We develop a kernel method of MvCCDA to further boost the performance of MvCCDA.
35,244
kernel optimization analysis for MvCCDA extension, and completeness. presented Beyond are of complexity also
Beyond kernel extension, optimization and complexity analysis of MvCCDA are also presented for completeness.
35,245
years. recent analysis become under has the Medical image topic spotlight a in
Medical image analysis has become a topic under the spotlight in recent years.
35,246
medical in of machine a concerning There the is research image usage learning. significant progress
There is a significant progress in medical image research concerning the usage of machine learning.
35,247
and answers questions awaiting still problems and solutions, numerous there However, are respectively.
However, there are still numerous questions and problems awaiting answers and solutions, respectively.
35,248
physical subject. extremely is an optical of relevant of layer the networks Supervision
Supervision of the physical layer of optical networks is an extremely relevant subject.
35,249
in are complexity, Characteristics simulated as time, processing specificity, such and environments. analysed sensitivity,
Characteristics such as sensitivity, specificity, processing time, and complexity, are analysed in simulated environments.
35,250
for an There explain is their to demand algorithms outcomes. increasing
There is an increasing demand for algorithms to explain their outcomes.
35,251
is So produced no far, ranking explains rankings method there the algorithm. a that by
So far, there is no method that explains the rankings produced by a ranking algorithm.
35,252
in generations. network problem deep a focus neural We on training of the
We focus on the problem of training a deep neural network in generations.
35,253
of approach. also Model the and extraction ensemble feature our transfer verify effectiveness
Model ensemble and transfer feature extraction also verify the effectiveness of our approach.
35,254
Carlo algorithm a on Monte inference powerful manifolds. Bayesian Geodesic non-Euclidean is (gMC) for
Geodesic Monte Carlo (gMC) is a powerful algorithm for Bayesian inference on non-Euclidean manifolds.
35,255
object are applications. one Human class interesting numerous faces with
Human faces are one interesting object class with numerous applications.
35,256
problem images deblurring In by this exploiting we the address of structures. face facial paper,
In this paper, we address the problem of deblurring face images by exploiting facial structures.
35,257
affect can the Inter-pathologist test diagnostic accuracy. results in variability
Inter-pathologist variability in the test results can affect diagnostic accuracy.
35,258
in have Multi-label manifested learning machine problems various applications. themselves learning
Multi-label learning problems have manifested themselves in various machine learning applications.
35,259
challenges. framework space Recently, transformation these targeting proposed a label been (LST) has
Recently, a label space transformation (LST) framework has been proposed targeting these challenges.
35,260
(DLST) paper, transformation this In distribution-based we model. propose space label a
In this paper, we propose a distribution-based label space transformation (DLST) model.
35,261
the Consequently, using performance. classifier better codes trained dense latent multi-label yields
Consequently, multi-label classifier trained using the dense latent codes yields better performance.
35,262
The the information of to enables fill out leverage DLST correlations. additional about distribution label
The leverage of distribution enables DLST to fill out additional information about the label correlations.
35,263
employed to ML-KNN the label Then is transformed vector the code. original latent recover from
Then ML-KNN is employed to recover the original label vector from the transformed latent code.
35,264
human the paper, associations faces we In voices. and study this between
In this paper, we study the associations between human faces and voices.
35,265
outcome which increases poor leads rates. mortality treatment This the to
This leads to poor treatment outcome which increases the mortality rates.
35,266
Tehran tested Center. Service Emergency system as in The been has prototype designed a
The designed system has been tested as a prototype in Tehran Emergency Service Center.
35,267
to in changes This leads contact patterns.
This leads to changes in contact patterns.
35,268
population demonstrate negative the at level. models potential to consequences the use network We epidemic
We use epidemic network models to demonstrate the potential negative consequences at the population level.
35,269
through take models. structure of account the several into network population social the We
We take into account the social structure of the population through several network models.
35,270
infectious evolves, the epidemic their themselves distance susceptible individuals contacts. from As may
As the epidemic evolves, susceptible individuals may distance themselves from their infectious contacts.
35,271
replace social ties. Some lost seeking individuals by connections their new
Some individuals replace their lost social connections by seeking new ties.
35,272
real-world in same can negative effect Moreover, arise networks. the
Moreover, the same negative effect can arise in real-world networks.
35,273
importance findings the measures analysis careful models. preventive in epidemic of of highlight These
These findings highlight the importance of careful analysis of preventive measures in epidemic models.
35,274
tasks. related among generalization sharing knowledge improves learning by Multi-task performance
Multi-task learning improves generalization performance by sharing knowledge among related tasks.
35,275
segmentation detection proposed using architecture evaluated We a combination two and on our of datasets.
We evaluated our proposed architecture on a combination of detection and segmentation using two datasets.
35,276
how demonstrate wild with limited Experiments birds CNN from representations general our datasets. learns
Experiments with wild birds demonstrate how our CNN learns general representations from limited datasets.
35,277
finding constraint to a is problems. solutions QGS satisfaction systematic approach of
QGS is a systematic approach to finding solutions of constraint satisfaction problems.
35,278
provide QGS better shows with Comparison that significantly classification. neural network trained networks classifiers other
Comparison with other neural network classifiers shows that QGS trained networks provide significantly better classification.
35,279
generated. material Gene genetic is duplication which through is a new mechanism major
Gene duplication is a major mechanism through which new genetic material is generated.
35,280
tasks is an important computer different and vision for Texture cue applications.
Texture is an important cue for different computer vision tasks and applications.
35,281
efficient yet Pattern the (LBP) of one considered is descriptors. Local Binary best texture
Local Binary Pattern (LBP) is considered one of the best yet efficient texture descriptors.
35,282
noise. sensitivity notable limitations, However, mostly some to has the LBP
However, LBP has some notable limitations, mostly the sensitivity to noise.
35,283
distributions models. are in building a block Gaussian used generative key commonly as many
Gaussian distributions are commonly used as a key building block in many generative models.
35,284
However, well their networks. explored deep applicability been has not in
However, their applicability has not been well explored in deep networks.
35,285
multivariate can model the functions be data. dependence copula to structure used of Empirical
Empirical copula functions can be used to model the dependence structure of multivariate data.
35,286
algorithm numerically benefits the computational theoretically and The and error approximation of is assessed.
The computational benefits and approximation error of the algorithm is theoretically and numerically assessed.
35,287
we for network (DR-ResNet) based counting. deeply-recursive crowd propose a Consequently, blocks on ResNet
Consequently, we propose a deeply-recursive network (DR-ResNet) based on ResNet blocks for crowd counting.
35,288
Beijing data video-monitoring from Besides, we dataset a bus the new generate station. of
Besides, we generate a new dataset from the video-monitoring data of Beijing bus station.
35,289
MEnet several public desirable that Tests performance. show benchmarks has on achieved
Tests on several public benchmarks show that MEnet has achieved desirable performance.
35,290
data tera-voxel current brain interest. neuroanatomy much of light-microscopic is Whole sets using
Whole brain neuroanatomy using tera-voxel light-microscopic data sets is of much current interest.
35,291
tasks. hence They for facial important analysis various are
They are hence important for various facial analysis tasks.
35,292
utilize the shape and to appearance facial ways in information. the They differ
They differ in the ways to utilize the facial appearance and shape information.
35,293
models explicitly global build holistic represent and appearance to shape The the methods information. facial
The holistic methods explicitly build models to represent the global facial appearance and shape information.
35,294
the models. appearance local model build global shape explicitly CLMs The but the leverage
The CLMs explicitly leverage the global shape model but build the local appearance models.
35,295
facial shape capture appearance implicitly information. regression-based methods and The
The regression-based methods implicitly capture facial shape and appearance information.
35,296
theories as discuss we differences. category, each as algorithms well their their within For underlying
For algorithms within each category, we discuss their underlying theories as well as their differences.
35,297
strengths out and their weaknesses. on evaluations, the point respective we Based
Based on the evaluations, we point out their respective strengths and weaknesses.
35,298
section There separate the deep review to learning-based is also a latest algorithms.
There is also a separate section to review the latest deep learning-based algorithms.
35,299