text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
general deep we the neural focus this of work, on attention In in effect networks.
In this work, we focus on the effect of attention in general deep neural networks.
37,500
Our channel along an separate pathways, map spatial. and infers attention module two
Our module infers an attention map along two separate pathways, channel and spatial.
37,501
be and publicly The models code will available.
The code and models will be publicly available.
37,502
less the angles. In uncontrolled results pattern it variations on addition,
In addition, it results less pattern variations on the uncontrolled angles.
37,503
Training to segmentation network labeled a semantic amounts large of perform deep data. requires
Training a deep network to perform semantic segmentation requires large amounts of labeled data.
37,504
network real on relatively performs trained a synthetic data poorly on Unfortunately, images.
Unfortunately, a network trained on synthetic data performs relatively poorly on real images.
37,505
multi-script for in using architecture document end-to-end this retrieval paper. An proposed handwritten signatures is
An end-to-end architecture for multi-script document retrieval using handwritten signatures is proposed in this paper.
37,506
In features second (i.e. stage, from the foreground the
In the second stage, features from the foreground (i.e.
37,507
the (i.e. strokes) and background spatial information signature
signature strokes) and the background spatial information (i.e.
37,508
the characterize to object with match signature combined to the were signature. query
were combined to characterize the signature object to match with the query signature.
37,509
obtained. and results The promising proposed system documents was on were also tested noisy
The proposed system was also tested on noisy documents and promising results were obtained.
37,510
the A comparative approaches. state-of-the-art study that outperforms shows method proposed the
A comparative study shows that the proposed method outperforms the state-of-the-art approaches.
37,511
an Image where ill-posed are restoration role. the important problems plays regularization problems typical term
Image restoration problems are typical ill-posed problems where the regularization term plays an important role.
37,512
interpolations convolution. transposed also of Further, we place in multiple have added
Further, we have also added multiple interpolations in place of transposed convolution.
37,513
computationally obtained perceptually being very while The results efficient. in quality, are similar
The results obtained are perceptually similar in quality, while being computationally very efficient.
37,514
propose colorization. first for exemplar-based the We learning local deep approach
We propose the first deep learning approach for exemplar-based local colorization.
37,515
automatically the The picking by be reference can fully top simply suggestion. colorization performed
The colorization can be performed fully automatically by simply picking the top reference suggestion.
37,516
user and Our through methods. study favorable comparisons the-state-of-the-art is to a quantitative approach validated
Our approach is validated through a user study and favorable quantitative comparisons to the-state-of-the-art methods.
37,517
be Furthermore, extended naturally colorization. approach to our video can
Furthermore, our approach can be naturally extended to video colorization.
37,518
and for models freely code available use. public Our be will
Our code and models will be freely available for public use.
37,519
previous Compared is works, our based neural edge-aware. with consolidation network
Compared with previous neural network based works, our consolidation is edge-aware.
37,520
relays. the Access and assumed the for sources is Multiple Division Time
Time Division Multiple Access is assumed for the sources and the relays.
37,521
Sources the time in slots phase. transmit successively first for
Sources transmit successively in time slots for the first phase.
37,522
limited The second slots phase for a of number time of consists retransmissions.
The second phase consists of a limited number of time slots for retransmissions.
37,523
asymmetric conclusion scenarios. source for both valid same The and rate remains symmetric
The same conclusion remains valid for both symmetric and asymmetric source rate scenarios.
37,524
processes (MDPs) planning Markov such problems. to models capture decision prominent are
Markov decision processes (MDPs) are prominent models to capture such planning problems.
37,525
tradeoffs the discuss exploration and between We safety. sufficient of environment in ensuring progress
We discuss tradeoffs between sufficient progress in exploration of the environment and ensuring safety.
37,526
scenario. selective to in classification poorly are attribute-based Existing classification zero-shot methods the work shown
Existing attribute-based zero-shot classification methods are shown to work poorly in the selective classification scenario.
37,527
the under-complete human attribute the poor accounts for We defined vocabulary argue performance.
We argue the under-complete human defined attribute vocabulary accounts for the poor performance.
37,528
is residual the learning and firstly the proposed jointly. by The attributes constructed classifier defined
The proposed classifier is constructed by firstly learning the defined and the residual attributes jointly.
37,529
defined Then predictions subspace the are the of within attributes. conducted the
Then the predictions are conducted within the subspace of the defined attributes.
37,530
measured the confidence and defined by prediction both Finally, attributes. the residual the is
Finally, the prediction confidence is measured by both the defined and the residual attributes.
37,531
is to be time-consuming medical a task by performed experts. However,it
However,it is a time-consuming task to be performed by medical experts.
37,532
So, automatic are methods reliable required. segmentation and
So, automatic and reliable segmentation methods are required.
37,533
clinical. performs can proposed fast and be The used in method
The proposed method performs fast and can be used in clinical.
37,534
mechanism for enhance recognition. novel a Networks to Convolutional Neural propose fine-grained attention We
We propose a novel attention mechanism to enhance Convolutional Neural Networks for fine-grained recognition.
37,535
and analysis. and Object satellite fundamental challenging is a image detection problem in aerial
Object detection is a fundamental and challenging problem in aerial and satellite image analysis.
37,536
satisfy dense scenes. boxes in can These candidate sparse and tasks detection
These candidate boxes can satisfy detection tasks in sparse and dense scenes.
37,537
to Experimental superiority results the state-of-the-art. of demonstrate the the framework proposed
Experimental results demonstrate the superiority of the proposed framework to the state-of-the-art.
37,538
independently. re-id classical However, per-camera lists generate approaches
However, classical re-id approaches generate per-camera lists independently.
37,539
sequential multi-camera propose a shortcoming, novel re-id we To this approach. address
To address this shortcoming, we propose a novel sequential multi-camera re-id approach.
37,540
custom-designed query improve to incrementally We formulate an procedure optimization representation.
We formulate an optimization procedure custom-designed to incrementally improve query representation.
37,541
a we conduct subject-based comparative human operator Additionally, study of performance.
Additionally, we conduct a comparative subject-based study of human operator performance.
37,542
implementation as of this sort. As publicly know this is far the we first available
As far as we know this is the first publicly available implementation of this sort.
37,543
max/min novel this In a for a we work, shift-register-based architecture stochastic propose function.
In this work, we propose a novel shift-register-based architecture for a stochastic max/min function.
37,544
logo, embroidery) texture, that clothes. original depict the
texture, logo, embroidery) that depict the original clothes.
37,545
is namely, proposed. address Feature) a skeletonbased these To representation, tasks, (Skeleton Pose-Motion SPMF
To address these tasks, a skeletonbased representation, namely, SPMF (Skeleton Pose-Motion Feature) is proposed.
37,546
represent effectively to complex Therefore, able they actions. are
Therefore, they are able to effectively represent complex actions.
37,547
requiring results that computation. indicated less method state-of-the-art methods proposed surpasses Experimental whilst the
Experimental results indicated that the proposed method surpasses state-of-the-art methods whilst requiring less computation.
37,548
theoretical and suggest models for both directions experimental inquiry. Peptide/IMT
Peptide/IMT models suggest directions for both theoretical and experimental inquiry.
37,549
proposed shortcomings with networks capsule of deal to Neural were Networks Convolutional Recently, (ConvNets).
Recently, capsule networks were proposed to deal with shortcomings of Convolutional Neural Networks (ConvNets).
37,550
Power from gain angle-steering-based then beamforming CSI analyzed. tracked is using
Power gain from angle-steering-based beamforming using tracked CSI is then analyzed.
37,551
in intra-cluster spreads. under provides a beam-width different design study angle-steering The angular guideline theoretical
The theoretical study provides a design guideline beam-width in angle-steering under different intra-cluster angular spreads.
37,552
improvements. many have made, been is advancements While room there for continued
While many advancements have been made, there is room for continued improvements.
37,553
DIR for its Yet usefulness not was assessed.
Yet its usefulness for DIR was not assessed.
37,554
two differentiate changes, that (e.g. populations the We which geometric show
We show that geometric changes, which differentiate the two populations (e.g.
37,555
synthesized readily as well. vs. inhale are exhale),
inhale vs. exhale), are readily synthesized as well.
37,556
In radio recent years, as spectrum. cognitive introduced efficient of for use a tool network
In recent years, cognitive radio network introduced as a tool for efficient use of spectrum.
37,557
with primary must users Secondary that users are unauthorized from interferences transmission. avoid ones
Secondary users are unauthorized ones that must avoid from interferences with primary users transmission.
37,558
The structured for hierarchically has tractable importance of been planning representations long acknowledged.
The importance of hierarchically structured representations for tractable planning has long been acknowledged.
37,559
(MRF) Resonance exception. is Fingerprinting an Magnetic not
Magnetic Resonance Fingerprinting (MRF) is not an exception.
37,560
The are existing addressing neither methods nor accurate. PVE scalable
The existing methods addressing PVE are neither scalable nor accurate.
37,561
methods. that PVE, with algorithm outperforms We show, synthetic simulations on through data state-of-the-art our
We show, through simulations on synthetic data with PVE, that our algorithm outperforms state-of-the-art methods.
37,562
and approach in phantom Our on datasets. EUROSPIN vivo is validated the on
Our approach is validated on the EUROSPIN phantom and on in vivo datasets.
37,563
image and This a system text proposes on cross-modal leverages that retrieval paper encoding.
This paper proposes a cross-modal retrieval system that leverages on image and text encoding.
37,564
modalities. can We encodings between that capture semantic show multiple relationships text
We show that text encodings can capture semantic relationships between multiple modalities.
37,565
a removal inpainting. new for and introduce system content image automatic We
We introduce a new system for automatic image content removal and inpainting.
37,566
and a single performed inpainting jointly in are pass. Region segmentation
Region segmentation and inpainting are performed jointly in a single pass.
37,567
more this segmentation module. are inpainting by In way, errors alleviated naturally the potential
In this way, potential segmentation errors are more naturally alleviated by the inpainting module.
37,568
GAN is trained model using The a whole strategy. conditional
The whole model is trained using a conditional GAN strategy.
37,569
simulation illustrate greedy formulation studies proposed of extensive efficacy and algorithm. the the the Our
Our extensive simulation studies illustrate the efficacy of the proposed formulation and the greedy algorithm.
37,570
two-level is proposed. motion A optimization node
A two-level node motion optimization is proposed.
37,571
frame Time-aware a problem video video a fundamental is encoding of in understanding. in sequences
Time-aware encoding of frame sequences in a video is a fundamental problem in video understanding.
37,572
aim video To this we evaluate explicitly. lacuna, fill time to
To fill this lacuna, we aim to evaluate video time explicitly.
37,573
task associated on Based to the a we quantify formulate each able property.
Based on each we formulate a task able to quantify the associated property.
37,574
These in properties domain. are common medical
These properties are common in medical domain.
37,575
multiple of consider classes. most However, don't relationship them between
However, most of them don't consider relationship between multiple classes.
37,576
network designed fine-grained MSML. concept classification have on architecture based that deep incorporates Moreover, we
Moreover, we have designed deep network architecture based on fine-grained classification concept that incorporates MSML.
37,577
two accuracy The and stereo of are speed. factors important vision
The two important factors of stereo vision are accuracy and speed.
37,578
promote species. can On the hand, other chaos coexistence of
On the other hand, chaos can promote coexistence of species.
37,579
we this higher in biodiversity. Moreover that results a confirm
Moreover we confirm that this results in a higher biodiversity.
37,580
structural out. analysis relatively classification, regions exhaustive For and carried of contours is a
For classification, a relatively exhaustive structural analysis of contours and regions is carried out.
37,581
of algorithms at the expense cost. accuracy learning computation significant Deep classification achieve high
Deep learning algorithms achieve high classification accuracy at the expense of significant computation cost.
37,582
heart define disease. causes in and of well-established the the cardiovascular Alterations of function geometry
Alterations in the geometry and function of the heart define well-established causes of cardiovascular disease.
37,583
phenotypes. and factors limit structural the sensitivity complex in quantifying functional Both
Both factors limit the sensitivity in quantifying complex structural and functional phenotypes.
37,584
its time we paper, first for the for this In SAR tomography. demonstrate potential
In this paper, we demonstrate for the first time its potential for SAR tomography.
37,585
set. richer data used is Image a to get enhancement
Image enhancement is used to get a richer data set.
37,586
and the achieves joint testing is good test, used In model results. for the
In the test, the model is used for joint testing and achieves good results.
37,587
aleatoric our combine proposed compare with We and uncertainty uncertainty. model
We compare and combine our proposed aleatoric uncertainty with model uncertainty.
37,588
their image unsatisfactory. is quality However, still
However, their image quality is still unsatisfactory.
37,589
the same proposed outperforms recent quantitative the algorithms addressing and qualitative For algorithm evaluations, problem.
For qualitative and quantitative evaluations, the proposed algorithm outperforms recent algorithms addressing the same problem.
37,590
However, in be itself cue. important an location can
However, location in itself can be an important cue.
37,591
for We adding location approaches information. test two
We test two approaches for adding location information.
37,592
experiments perform with We location. indirect encode ways direct both to and
We perform experiments with both direct and indirect ways to encode location.
37,593
array from Our digital greatly shows multi-user analysis benefits multiplexing.
Our analysis shows digital array benefits greatly from multi-user multiplexing.
37,594
knowledge data detection. from transferring) sensor and low-resolution target enhancing
transferring) knowledge from sensor data and enhancing low-resolution target detection.
37,595
and approaches that performance models. better demonstrate recent achieves we traditional teacher-student Moreover, than it
Moreover, we demonstrate that it achieves better performance than traditional approaches and recent teacher-student models.
37,596
encoder-decoder built structure. segmentation an mainly architectures upon are Semantic
Semantic segmentation architectures are mainly built upon an encoder-decoder structure.
37,597
downsampling perform the operations models in encoder. These subsequent
These models perform subsequent downsampling operations in the encoder.
37,598
commonly ordered diagnostic cost-effective physicians. tests the by Chest imaging Background: X-rays are performed, most
Background: Chest X-rays are the most commonly performed, cost-effective diagnostic imaging tests ordered by physicians.
37,599