text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
streak its blur first the rain interference with model motion mechanism. We
We first model the rain streak interference with its motion blur mechanism.
38,800
image This a dehazing. single for proposes technique paper novel
This paper proposes a novel technique for single image dehazing.
38,801
method approaches. The combines two the different proposed
The proposed method combines the two different approaches.
38,802
i.e., sequence of a the Given vectors. representation, pair, matrix a a generates video-description network
Given a video-description pair, the network generates a matrix representation, i.e., a sequence of vectors.
38,803
vector a description. video conditioned represents in Each by frame matrix the the
Each vector in the matrix represents a video frame conditioned by the description.
38,804
robust yields the a for representation multi-level multiple Integration video-language abstraction. granularities of
Integration of the multiple granularities yields a robust representation for multi-level video-language abstraction.
38,805
of validate large-scale effectiveness two We ASST datasets. on our the
We validate the effectiveness of our ASST on two large-scale datasets.
38,806
pose to key a intelligent assistance is Person context-aware and systems. detection develop requirement estimation
Person detection and pose estimation is a key requirement to develop intelligent context-aware assistance systems.
38,807
occlusions also and challenges includes in as such visual environments, It present clutter. the such
It also includes the visual challenges present in such environments, such as occlusions and clutter.
38,808
have be convolutional architectures. been shown model a recent competitive to among DenseNets network
DenseNets have been shown to be a competitive model among recent convolutional network architectures.
38,809
introduce early We new a three inference. constructs allow which that ParaNet, architecture for pipelines
We introduce ParaNet, a new architecture that constructs three pipelines which allow for early inference.
38,810
leading among of death are Cardiovascular causes the globally. diseases
Cardiovascular diseases are among the leading causes of death globally.
38,811
We area measures. introduced the on novel distortion scores based spiculation spiculation metric and
We introduced novel spiculation scores based on the area distortion metric and spiculation measures.
38,812
model malignancy introduced. prediction also pathological A is simple
A simple pathological malignancy prediction model is also introduced.
38,813
quantifying We parameter-free interpretable, spiculations nodules. technique developed on a and reproducible for
We developed a reproducible and interpretable, parameter-free technique for quantifying spiculations on nodules.
38,814
location-discriminating We method having learning representation propose higher a novel power.
We propose a novel representation learning method having higher location-discriminating power.
38,815
is by gradients. analyzing better why insights SARE It provides into
It provides insights into why SARE is better by analyzing gradients.
38,816
CNN. any to SARE easy is and implement Our pluggable to loss
Our SARE loss is easy to implement and pluggable to any CNN.
38,817
aims navigation at Indoor navigation performing within buildings.
Indoor navigation aims at performing navigation within buildings.
38,818
we itself scenarios, In an in unseen environment. intelligent most expected navigating of capable device
In most scenarios, we expected an intelligent device capable of navigating itself in unseen environment.
38,819
sort traces a fingerprint. Such represent of camera
Such traces represent a sort of camera fingerprint.
38,820
tasks. breakthrough techniques perceptive are Deep results image-based obtaining in Recent these (DL) Learning
Recent image-based Deep Learning (DL) techniques are obtaining breakthrough results in these perceptive tasks.
38,821
lot Training data. good of deep a a learning annotated requires often model
Training a good deep learning model often requires a lot of annotated data.
38,822
images. can panoramic on detected be ACPs
ACPs can be detected on panoramic images.
38,823
methods, These used barely dentistry. however in been have
These methods, however have been barely used in dentistry.
38,824
used study R-CNN) we Network In Faster the deep this Region-based Convolutional learning. (Faster for
In this study we used the Faster Region-based Convolutional Network (Faster R-CNN) for deep learning.
38,825
Receiver calculate was performed accuracy detection. Characteristic the Operating to (ROC) analysis of
Receiver Operating Characteristic (ROC) analysis was performed to calculate the accuracy of detection.
38,826
on growing. of capturing dynamic scenes environments rapidly are underwater Demands
Demands on capturing dynamic scenes of underwater environments are rapidly growing.
38,827
complex on The state-of-the-art the colorectal dataset. show results performance experimental adenocarcinoma
The experimental results show the state-of-the-art performance on complex colorectal adenocarcinoma dataset.
38,828
during are imperfections unknown or process. calibration measurement there when cases Such parameters the exist
Such cases exist when there are imperfections or unknown calibration parameters during the measurement process.
38,829
version of The is performance numerically. robust VAMP of the demonstrated
The performance of the robust version of VAMP is demonstrated numerically.
38,830
attack for are introduced this paper. Three in discriminative detection face representations presentation
Three discriminative representations for face presentation attack detection are introduced in this paper.
38,831
two also of design kinds detection, attack accurate presentation we combinations. representation For
For accurate presentation attack detection, we also design two kinds of representation combinations.
38,832
Firstly, SSD. cascade combine decision-level we a with to strategy SPMT propose
Firstly, we propose a decision-level cascade strategy to combine SPMT with SSD.
38,833
weaker interferers. in suppres- the detection strong of signals a of This improves presence sion
This suppres- sion improves detection of a weaker signals in the presence of strong interferers.
38,834
deep Iizuka present based a we on Accordingly, et approach learning al.
Accordingly, we present a deep learning approach based on Iizuka et al.
38,835
network boundaries. adversarially training to for a hallucinate image past
for adversarially training a network to hallucinate past image boundaries.
38,836
both outpainting Our that results feasible to deep show approaches are promising. image and learning
Our results show that deep learning approaches to image outpainting are both feasible and promising.
38,837
in strategy training. This considering more testing in than helps when classes
This strategy helps when considering more classes in testing than in training.
38,838
show environments. also not sacrificed is rich the We in performance scattering that
We also show that the performance is not sacrificed in rich scattering environments.
38,839
phenomenon. available enough to generalize The intelligent not are complex this techniques
The available techniques are not intelligent enough to generalize this complex phenomenon.
38,840
learning mass of the in mammograms. for segmentation deep explore We use breast
We explore the use of deep learning for breast mass segmentation in mammograms.
38,841
tedious techniques nor Moreover, required are our pre-processing post-processing not neither algorithm. in
Moreover, neither tedious pre-processing nor post-processing techniques are not required in our algorithm.
38,842
is food preferences events. of the people One interaction with their of
One of the preferences of people is their interaction with food events.
38,843
The an proposed dataset is in-house private on evaluated "EgoFoodPlaces". called model
The proposed model is evaluated on an in-house private dataset called "EgoFoodPlaces".
38,844
places of promising food recognition photo-streams. shows Experimental results egocentric in results classification
Experimental results shows promising results of food places classification recognition in egocentric photo-streams.
38,845
are to Humans movements. innately others' emotional from prepared subtle arguably comprehend body expressions
Humans are arguably innately prepared to comprehend others' emotional expressions from subtle body movements.
38,846
insights. dataset the revealed analysis interesting statistical many of Comprehensive
Comprehensive statistical analysis of the dataset revealed many interesting insights.
38,847
far, features, based on So analysis of is mostly this kind radiomics i.e.
So far, this kind of analysis is mostly based on radiomics features, i.e.
38,848
features priori. a defined that of mathematically fixed a set is
a fixed set of features that is mathematically defined a priori.
38,849
has improved knowledge domain text. Incorporating understanding machine of
Incorporating domain knowledge has improved machine understanding of text.
38,850
architecture this i.e. estimate a to end, CNN is employed To parametric representations
To this end, a CNN architecture is employed to estimate parametric representations i.e.
38,851
pose, hand shape parameters. bone and scales complex
hand pose, bone scales and complex shape parameters.
38,852
data manually real annotating is Also, suboptimal.
Also, manually annotating real data is suboptimal.
38,853
The last section. formulae given in are updating the
The updating formulae are given in the last section.
38,854
integrating the stereo We factor constraints SfM camera exploiting by results graphs. initial refine
We refine the initial SfM results by integrating stereo camera constraints exploiting factor graphs.
38,855
object compute by camera pose background trajectory We information. combining the object and
We compute the object trajectory by combining object and background camera pose information.
38,856
method monocular As our trajectory opposed reconstruction degenerated to no approaches, shows cases.
As opposed to monocular trajectory reconstruction approaches, our method shows no degenerated cases.
38,857
approach data available vehicles of We scenes. using in publicly video our evaluate urban
We evaluate our approach using publicly available video data of vehicles in urban scenes.
38,858
(MC-DART), In deals multi-channel Multi-Channel with paper effectively we this DART present which data.
In this paper we present Multi-Channel DART (MC-DART), which deals effectively with multi-channel data.
38,859
is communication. wireless resource an frequency The essential for spectrum
The frequency spectrum is an essential resource for wireless communication.
38,860
Cognitive the approach Radio solution new comes with (CR) scarcity. to The spectrum feasible
The new Cognitive Radio (CR) approach comes with the feasible solution to spectrum scarcity.
38,861
task networks estimating aerial tackles images. of of from road paper the the topology This
This paper tackles the task of estimating the topology of road networks from aerial images.
38,862
AI. field light We of direction assess omics and of the the Biobanks/PheWAS, atlases, in
We assess the direction of the field in light of Biobanks/PheWAS, omics atlases, and AI.
38,863
detection self-driving Lane vehicles. is for important very
Lane detection is very important for self-driving vehicles.
38,864
algorithm minimal and was for operation very Furthermore, the run-time. fast designed
Furthermore, the algorithm was designed for very fast operation and minimal run-time.
38,865
of global mechanics, stationarity the the minimum. with In corresponds condition unlike particular, action the
In particular, unlike the action of mechanics, the stationarity condition corresponds with the global minimum.
38,866
in problem is of experimented theory computer extraction the Finally, feature the vision. for
Finally, the theory is experimented for the problem of feature extraction in computer vision.
38,867
for representation a Boltzmann (RBM) learning. restricted investigate discriminative the Machine We of potential
We investigate the potential of a restricted Boltzmann Machine (RBM) for discriminative representation learning.
38,868
Experimental effectiveness show the on set benchmark of results method. the our data
Experimental results on the benchmark data set show the effectiveness of our method.
38,869
in novel context. is proposed here method set active The this
The active set method proposed here is novel in this context.
38,870
artifacts main and described characterized. the are this aim, mathematically recurrent To four
To this aim, the four main recurrent artifacts are described and mathematically characterized.
38,871
we and formula. design artifacts measurement attached a these its each For of test-pattern
For each of these artifacts we design a test-pattern and its attached measurement formula.
38,872
contrast, coverage. In accurate yields the interval bootstrap reversed-tails
In contrast, the reversed-tails bootstrap interval yields accurate coverage.
38,873
the also We recursive-design bootstrap with theoretically the bootstrap. compare analyzed fixed-design
We also compare the theoretically analyzed fixed-design bootstrap with the recursive-design bootstrap.
38,874
the empirical estimation. An application illustrates interval
An empirical application illustrates the interval estimation.
38,875
Camera of a image are fingerprints forensics number for precious tasks. tools
Camera fingerprints are precious tools for a number of image forensics tasks.
38,876
well-known example photo the noise pattern, (PRNU) fingerprint. is A a response device non-uniformity powerful
A well-known example is the photo response non-uniformity (PRNU) noise pattern, a powerful device fingerprint.
38,877
affected only by mildly of is residuals content. the However, it high-level PRNU, unlike scene
However, unlike the PRNU, it is only mildly affected by residuals of high-level scene content.
38,878
proposed the The experiments localization the method improves over reference. PRNU-based noiseprint-based that show forgery
The experiments show that the proposed noiseprint-based forgery localization method improves over the PRNU-based reference.
38,879
a of multilingual handwriting for address the system recognition. design We unified
We address the design of a unified multilingual system for handwriting recognition.
38,880
complexity. size reduces the the Dealing with decreases thus model language multigrams strongly lexicon and
Dealing with multigrams strongly reduces the lexicon size and thus decreases the language model complexity.
38,881
identity been applied economic in widely Kotlarski's used research. has
Kotlarski's identity has been widely used in applied economic research.
38,882
for motorcycle Neural system a classification This using paper Convolutional Network urban presents scenarios (CNN).
This paper presents a motorcycle classification system for urban scenarios using Convolutional Neural Network (CNN).
38,883
Nevertheless, extracted trained. be CNNs already features can from
Nevertheless, features can be extracted from CNNs already trained.
38,884
classification. on images performed satisfactory validation set are a showing Further experiments of a
Further experiments are performed on a validation set of images showing a satisfactory classification.
38,885
network is The weak supervision. entire with trainable end-to-end only
The entire network is end-to-end trainable with only weak supervision.
38,886
dosage tablets. forms benefits can Multi-particulate over oral conventional and significant capsules offer
Multi-particulate oral dosage forms can offer significant benefits over conventional capsules and tablets.
38,887
and vitro reported in swallow Both for smaller easier multi-particulates. vivo in tests
Both in vivo and in vitro tests reported easier swallow for smaller multi-particulates.
38,888
oral solids. for optimal appeared of water complete clearance Besides, not thin the liquids
Besides, water thin liquids appeared not optimal for complete oral clearance of the solids.
38,889
A convolutional network is these trained attributes. (CNN) learn to neural
A convolutional neural network (CNN) is trained to learn these attributes.
38,890
the This Autonomous Over paper Architecture describes project. Domains Restricted
This paper describes the Autonomous Architecture Over Restricted Domains project.
38,891
state-of-the-art results of networks. examine and the models deep existing modified neural We
We examine the results of existing state-of-the-art models and modified deep neural networks.
38,892
this the for solving competition we first task. Finally, explain in published detail
Finally, we explain in detail the first published competition for solving this task.
38,893
its scene a stage. in However, to is still preliminary fully social comprehend the ability
However, the ability to fully comprehend a social scene is still in its preliminary stage.
38,894
in guide gaze contextual human the to weakly-supervised a also regions It utilizes attended setting.
It also utilizes human gaze to guide the attended contextual regions in a weakly-supervised setting.
38,895
strategy In mis-grouping. we to hard sampling address negative of problem propose a addition, the
In addition, we propose a hard negative sampling strategy to address the problem of mis-grouping.
38,896
on extensive namely We datasets, perform experiments V-COCO benchmark HICO-DET. two and
We perform extensive experiments on two benchmark datasets, namely V-COCO and HICO-DET.
38,897
component efficacy each proposed The been has validated. also of
The efficacy of each proposed component has also been validated.
38,898
is receiving link only. State Information available side Channel each the (CSI) of at
Channel State Information (CSI) is available at the receiving side of each link only.
38,899