text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
segmentation framework, can other. the sufficiently and each detection boost In proposed
In the proposed framework, detection and segmentation can sufficiently boost each other.
39,600
introduced costs extra stage. test the at are no computational Therefore,
Therefore, no extra computational costs are introduced at the test stage.
39,601
way we this confidence approachin paper, In an matching. unsupervised stereo inference in present
In this paper, we present confidence inference approachin an unsupervised way in stereo matching.
39,602
(DNNs) state-of-the-art recently achieving been Deep Networks Neu-ral performance. have
Deep Neu-ral Networks (DNNs) have recently been achieving state-of-the-art performance.
39,603
dense Weshow the confidence is with identical that newly assumption introduced relaxed. map,the
Weshow that with the newly introduced dense confidence map,the identical assumption is relaxed.
39,604
featureless network In pixels (e.g.,noisy occlusions, regions) andfocusonhigh-confidence pixels. learns toattenuatelow-confidence input, practice,the
In practice,the network learns toattenuatelow-confidence pixels (e.g.,noisy input, occlusions, featureless regions) andfocusonhigh-confidence pixels.
39,605
trees (e.g. handle The of various can types package
The package can handle trees of various types (e.g.
39,606
to world. the The in the Knowledge uses (KG) facts graph the triples real describe
The Knowledge graph (KG) uses the triples to describe the facts in the real world.
39,607
widely in applications. analysis has been used It and intelligent
It has been widely used in intelligent analysis and applications.
39,608
possible However, noises in are the constructing. inevitably conflicts introduced of process and
However, possible noises and conflicts are inevitably introduced in the process of constructing.
39,609
is neural a crisscrossing network model structure. The
The model is a crisscrossing neural network structure.
39,610
pillars the computing. Deformable one of biomedical registration image of been has
Deformable registration has been one of the pillars of biomedical image computing.
39,611
using machines structured is support learned vector (LSSVM). This latent combination
This combination is learned using latent structured support vector machines (LSSVM).
39,612
the purpose, we the univocal problem, exploit that matching of i.e. To nature
To that purpose, we exploit the univocal nature of the matching problem, i.e.
39,613
all. or must at a corresponded observation feature with corresponded not single be every
every observation must be corresponded with a single feature or not corresponded at all.
39,614
circles tangent use filling idea first for surrounded the and Afterwards, we our circle.
Afterwards, we use our first idea for filling the tangent circles and surrounded circle.
39,615
propose a deal unified nonlinear simple measurements. to framework with we Here and
Here we propose a simple and unified framework to deal with nonlinear measurements.
39,616
In introduce in R written alphastable. the language we called this paper, package a
In this paper, we introduce a package written in the R language called alphastable.
39,617
results are Further, extensive simulation the considerations NR behind to design. the presented PUCCH explain
Further, extensive simulation results are presented to explain the considerations behind the NR PUCCH design.
39,618
scalable techniques. localization to use An efficient retrieval to and visual approach is image
An efficient and scalable approach to visual localization is to use image retrieval techniques.
39,619
drastically e.g. different image illumination However, across conditions, retrieval
However, image retrieval across drastically different illumination conditions, e.g.
39,620
many achieved models person in Recently, benchmarks. deep re-identification success learning-based great have
Recently, deep learning-based person re-identification models have achieved great success in many benchmarks.
39,621
To be randomly rectangles use pedestrian to images. block we occlude specific,
To be specific, we use block rectangles to randomly occlude pedestrian images.
39,622
use train model. Finally, we baseline augmented the datasets to
Finally, we use the augmented datasets to train baseline model.
39,623
structured-light-based depth indoor sensors For (e.g. scenes,
For indoor scenes, structured-light-based depth sensors (e.g.
39,624
are short-range, depth albeit provide maps. dense, able Kinect) to
Kinect) are able to provide dense, albeit short-range, depth maps.
39,625
captured various camera aspects investigate. to have images The
The camera captured images have various aspects to investigate.
39,626
depends of the emphasis on interesting the research regions. Generally,
Generally, the emphasis of research depends on the interesting regions.
39,627
focus detection be object or could Sometimes color text scene analysis. on the segmentation,
Sometimes the focus could be on color segmentation, object detection or scene text analysis.
39,628
always from associated machines learning the to provided samples. properties learn The
The learning machines always learn from the properties associated to provided samples.
39,629
in results images. on provided text analysis The reasonable convolutional non-cursive approach natural appeared
The convolutional approach provided reasonable results on non-cursive text analysis appeared in natural images.
39,630
empirically kernels. various selected through process image pyramid Each
Each pyramid image process through various empirically selected kernels.
39,631
between We maps the manipulating and of tasks consider analyzing shapes. representing,
We consider the tasks of representing, analyzing and manipulating maps between shapes.
39,632
in end-to-end We model a use image compression. entropy propose for optimized context-adaptive
We propose a context-adaptive entropy model for use in end-to-end optimized image compression.
39,633
The are independent DL. language and results our any of specific
The language and our results are independent of any specific DL.
39,634
of some open and discuss that we state also to be interest. believe We problems
We also state and discuss some open problems that we believe to be of interest.
39,635
neural diverse in (SP) signal applications networks found processing problems. (DNNs) Deep have
Deep neural networks (DNNs) have found applications in diverse signal processing (SP) problems.
39,636
problem. on The (ATR) radar a automatic hybrid-NN is target proposed tested recognition
The proposed hybrid-NN is tested on a radar automatic target recognition (ATR) problem.
39,637
proposed human deep-learning based Conclusion: method outperformed experts. The three CADx
Conclusion: The proposed deep-learning based CADx method outperformed three human experts.
39,638
boundary Meanwhile, proposed is extraction to another generate branch features. information boundary
Meanwhile, another boundary information extraction branch is proposed to generate boundary features.
39,639
our verify results. that framework achieves large-scale on and benchmarks state-of-the-art Experiments four analysis new
Experiments and analysis on four large-scale benchmarks verify that our framework achieves new state-of-the-art results.
39,640
arrays frequencies. antenna millimeter-wave the (mmWave) at This work demonstrates of operating sensitivity lens
This work demonstrates the sensitivity of lens antenna arrays operating at millimeter-wave (mmWave) frequencies.
39,641
and ill-posed inversion problem. a is typical inverse Waveform non-linear
Waveform inversion is a typical non-linear and ill-posed inverse problem.
39,642
years, solve In waveform recent way promising methods data-driven inversion become problem. the to a
In recent years, data-driven methods become a promising way to solve the waveform inversion problem.
39,643
suffer frameworks deep over-fitting and most from issue. learning generalization However,
However, most deep learning frameworks suffer from generalization and over-fitting issue.
39,644
issue. We on strategy a alleviate generalization learning to transfer the based develop VelocityGAN further
We further develop a transfer learning strategy based on VelocityGAN to alleviate the generalization issue.
39,645
to uncertainty scores way efficient We propose an output better calibrated from neural networks.
We propose an efficient way to output better calibrated uncertainty scores from neural networks.
39,646
increase fused performance. be information can segmentation The to harnessed
The fused information can be harnessed to increase segmentation performance.
39,647
for schemes classification. simultaneous propose fracture analyze and localization thoroughly Therefore, several we and
Therefore, we propose and thoroughly analyze several schemes for simultaneous fracture localization and classification.
39,648
localization auxiliary We general, performance. task, an improves in show the that using classification
We show that using an auxiliary localization task, in general, improves the classification performance.
39,649
make plan and We available. dataset code both to the
We plan to make both the dataset and code available.
39,650
networks. distribution the of angular stone grids represent Electric
Electric grids represent the angular stone of distribution networks.
39,651
real meant recognition systems are to world. Visual in the work
Visual recognition systems are meant to work in the real world.
39,652
this i.e. deals realistic Within generalization, context, very domain scenario with a
Within this context, a very realistic scenario deals with domain generalization, i.e.
39,653
neighboring consider The distribution pattern. the social interaction and their metrics both
The social metrics consider both the neighboring distribution and their interaction pattern.
39,654
independently. Traditionally, two tackled been tasks the have
Traditionally, the two tasks have been tackled independently.
39,655
is randomly varying eventually on lighting conditions based multi-shot strategy A discussed.
A multi-shot strategy based on randomly varying lighting conditions is eventually discussed.
39,656
Our major steps. of approach consists two
Our approach consists of two major steps.
39,657
for results. to supplementary Please the materials all refer
Please refer to the supplementary materials for all results.
39,658
to features. Specifically, deep adaptive we convolutional the new a exploit learned layer propose
Specifically, we propose a new adaptive convolutional layer to exploit the learned deep features.
39,659
independent each other overall are of the and These choices performance. impact substantially not
These choices are not independent of each other and substantially impact the overall performance.
39,660
also follows are of datasets predictors misleading. It inherently that different comparisons using
It also follows that comparisons of predictors using different datasets are inherently misleading.
39,661
only such prevails in imaging. as fields astronomical noise traditionally specific Poisson
Poisson noise traditionally prevails only in specific fields such as astronomical imaging.
39,662
deconvolutional connections. The with and architecture along convolutional hybrid layers symmetric a incorporates of
The architecture incorporates a hybrid of convolutional and deconvolutional layers along with symmetric connections.
39,663
of based SR highly performance depends on representation coeffcients. reconstruction, learned however, The learning
The performance of learning based SR reconstruction, however, highly depends on learned representation coeffcients.
39,664
makes localize position the of to best guess rough Instead, a iris. it the
Instead, it makes a best guess to localize the rough position of the iris.
39,665
of to are verify benchmark researchers their Lots for algorithms. released datasets
Lots of benchmark datasets are released for researchers to verify their algorithms.
39,666
segmentation many Semantic studied been has years. for
Semantic segmentation has been studied for many years.
39,667
has a made Since (DNN), tremendous the of Network Deep Neural segmentation progress. emergence
Since the emergence of Deep Neural Network (DNN), segmentation has made a tremendous progress.
39,668
is area a conclusion drawn. Finally, this in
Finally, a conclusion in this area is drawn.
39,669
process often information their ability sources. observers are in simultaneous However, human to limited
However, human observers are often limited in their ability to process simultaneous information sources.
39,670
tracking-by-detection Convolutional Neural the receives framework integration the The Networks growing attentions through with (CNNs).
The tracking-by-detection framework receives growing attentions through the integration with the Convolutional Neural Networks (CNNs).
39,671
variations. objects track based appearance Existing fail with severe tracking-by-detection to methods, however,
Existing tracking-by-detection based methods, however, fail to track objects with severe appearance variations.
39,672
assigned as which are the non-orthogonal In sequences signatures. their with CDMA, serve users non-orthogonal
In non-orthogonal CDMA, the users are assigned with non-orthogonal sequences which serve as their signatures.
39,673
propose point paper, edge and clouds. this novel unorganized for algorithms detection In corner we
In this paper, we propose novel edge and corner detection algorithms for unorganized point clouds.
39,674
various Ras present of superfamily. tokens revealed on GTPases protein the Application in token
Application of protein token on Ras GTPases revealed various tokens present in the superfamily.
39,675
to cloud systems image patches morphological and separable This magnitudes boundaries. gradient integrates method
This method integrates morphological image gradient magnitudes to separable cloud systems and patches boundaries.
39,676
flexible is extendable to imagery. single- method from The proposed and multi-spectral
The proposed method is flexible and extendable from single- to multi-spectral imagery.
39,677
vehicles unmanned system. of considered is and as distributed A manned a mixed aerial group
A mixed group of manned and unmanned aerial vehicles is considered as a distributed system.
39,678
of A be the fulfilled lattice which it. by to may matches system tasks
A lattice of tasks which may be fulfilled by the system matches to it.
39,679
external operations. defined multiplication which defines the linear is operation An correspondingly logic at lattice,
An external multiplication operation is defined at the lattice, which defines correspondingly linear logic operations.
39,680
mainly the purpose Thus, system choice the consequence. is
Thus, the choice is mainly the system purpose consequence.
39,681
resources. as Microgrids, islandable viable of energy power as considered small-scale are promotors renewable systems,
Microgrids, as small-scale islandable power systems, are considered as viable promotors of renewable energy resources.
39,682
how limit by a on microgrid. be there hosted is PV However, much a can
However, there is a limit on how much PV can be hosted by a microgrid.
39,683
introduced impacts, potential capacity avoid hosting of the negative concept used. is such and To
To avoid such potential negative impacts, the concept of hosting capacity is introduced and used.
39,684
based primarily analyses. steadystate are Hosting studies done capacity on
Hosting capacity studies are primarily done based on steadystate analyses.
39,685
result, An exhibits good commonly a EM set over-complete i.e. unmixing
An over-complete EM set commonly exhibits a good unmixing result, i.e.
39,686
are the reliable. unmixed abundances values less
the unmixed abundances values are less reliable.
39,687
greedy condition set scheme i.e. the Our number, EM reduction improves reducing by the
Our greedy reduction scheme improves the EM set by reducing the condition number, i.e.
39,688
possible. the while set's stability, as keeping error low reconstruction the enhancing as
enhancing the set's stability, while keeping the reconstruction error as low as possible.
39,689
size. The optimal set hint sequence the its set EM resulting gives and to
The resulting set sequence gives hint to the optimal EM set and its size.
39,690
allows periods. time of infectious for short high probability and unrealistically exposed This a
This allows for a high probability of unrealistically short exposed and infectious time periods.
39,691
effectively can its improve energy utilization. The RBC adaptive (ARBC)
The adaptive RBC (ARBC) can effectively improve its energy utilization.
39,692
DeepGlobe This to Challenge. contribution presents Detection Building our paper the
This paper presents our contribution to the DeepGlobe Building Detection Challenge.
39,693
enhancement next made the should to buildings separation of be adjacent The solution. the
The separation of adjacent buildings should be the next enhancement made to the solution.
39,694
these beyond fully a We by unsupervised go proposing approaches strategy.
We go beyond these approaches by proposing a fully unsupervised strategy.
39,695
principal the two subtasks. tackle problem We challenging splitting into by scenario this
We tackle this challenging scenario by splitting the problem into two principal subtasks.
39,696
algorithm The Results: proposed OCT successfully denoised single-frame unseen B-scans.
Results: The proposed algorithm successfully denoised unseen single-frame OCT B-scans.
39,697
data. histopathology amounts require labeled of large CNNs However,
However, CNNs require large amounts of labeled histopathology data.
39,698
order regression This compared models. consistency adversarial framework conventional enforces higher to when CNN
This adversarial regression framework enforces higher order consistency when compared to conventional CNN models.
39,699