text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
crowds a dense even Counting is for demanding humans. task in people
Counting people in dense crowds is a demanding task even for humans.
37,700
appearance to variability primarily large the in people. of due This is
This is primarily due to the large variability in appearance of people.
37,701
of seen a only Often blobs. people as are bunch
Often people are only seen as a bunch of blobs.
37,702
clutter further pose difficulty. variations Occlusions, the background and compound
Occlusions, pose variations and background clutter further compound the difficulty.
37,703
In semantics scenario, the requires a identifying of and spatial context larger scene. this person
In this scenario, identifying a person requires larger spatial context and semantics of the scene.
37,704
to prediction the we of Hence, correct the initial CNN. feedback propose top-down
Hence, we propose top-down feedback to correct the initial prediction of the CNN.
37,705
are various bottom-up the the top-down from fed of CNN layers network. to Features
Features from various layers of the bottom-up CNN are fed to the top-down network.
37,706
of crucial. step is process, dermoscopic the first in Segmentation is accuracy images thus this
Segmentation of dermoscopic images is the first step in this process, thus accuracy is crucial.
37,707
tackle paper problem. to an re-identification approach This presents the
This paper presents an approach to tackle the re-identification problem.
37,708
for person are machine to More datasets models train and more learning available re-identification.
More and more datasets are available to train machine learning models for person re-identification.
37,709
These positions, season, i.e. in conditions: numbers, location, cameras camera in datasets size, vary
These datasets vary in conditions: cameras numbers, camera positions, location, season, in size, i.e.
37,710
different number of identities. of images, number
number of images, number of different identities.
37,711
labeling: annotated Finally there in while are are with others datasets not. attributes
Finally in labeling: there are datasets annotated with attributes while others are not.
37,712
on Convolutional multitask learning. trained is (CNN) Our Neural Network a based using and model
Our model is based on a Convolutional Neural Network (CNN) and trained using multitask learning.
37,713
to different are available Several in datasets. losses extract the the information used different
Several losses are used to extract the different information available in the different datasets.
37,714
task classification learned Our with main is loss. a
Our main task is learned with a classification loss.
37,715
the reduce To we with intra-class variation the experiment loss. center
To reduce the intra-class variation we experiment with the center loss.
37,716
We works datasets. our that also outperforms two recent show re-identification system on
We also show that our system outperforms recent re-identification works on two datasets.
37,717
consider streams. the video of highlight-detection problem automatic We in game
We consider the problem of automatic highlight-detection in video game streams.
37,718
present unsupervised methodology for multi-view deep novelty-based learning this we In light, a detection. highlight
In this light, we present a multi-view unsupervised deep learning methodology for novelty-based highlight detection.
37,719
frequency windowed confinement. (OFDM) signals with multiplexing low pulses exhibit spectral Orthogonal division rectangularly
Orthogonal frequency division multiplexing (OFDM) signals with rectangularly windowed pulses exhibit low spectral confinement.
37,720
two This strategies. method and shaping proposes that a unifies generalizes paper spectral these
This paper proposes a spectral shaping method that generalizes and unifies these two strategies.
37,721
hierarchical. structures inherently key are indoor that observation Our scene is
Our key observation is that indoor scene structures are inherently hierarchical.
37,722
is or not recursive network a neural convolutional; it RvNN. our is network Hence,
Hence, our network is not convolutional; it is a recursive neural network or RvNN.
37,723
By a autoencoder a (VAE), resulting training roughly variational the distribution. follow Gaussian codes fixed-length
By training a variational autoencoder (VAE), the resulting fixed-length codes roughly follow a Gaussian distribution.
37,724
for method GRAINS, for We Generative our Autoencoders INdoor coin Scenes. Recursive
We coin our method GRAINS, for Generative Recursive Autoencoders for INdoor Scenes.
37,725
localization on methods based generate produced usually maps supervised attention networks. results by classification Weakly
Weakly supervised methods usually generate localization results based on attention maps produced by classification networks.
37,726
progressively are regions learn SPG confident utilized to maps masks. the high within The attention
The high confident regions within attention maps are utilized to progressively learn the SPG masks.
37,727
object Extensive high-quality demonstrate is producing that ILSVRC maps. effective on localizations SPG in experiments
Extensive experiments on ILSVRC demonstrate that SPG is effective in producing high-quality object localizations maps.
37,728
However, the by of is heterogeneous often challenged datasets. labeling this
However, this is often challenged by heterogeneous labeling of the datasets.
37,729
human analyzing role Facial cognitive a state. significant in expression has
Facial expression has a significant role in analyzing human cognitive state.
37,730
scale. performance influenced user-specified inlier much is the by its First,
First, its performance is much influenced by the user-specified inlier scale.
37,731
large is it data. Second, computationally inefficient for
Second, it is computationally inefficient for large data.
37,732
algorithm reduction problem MaxFS the for data the of This computationally makes realistic.
This reduction of data for the MaxFS problem makes the algorithm computationally realistic.
37,733
the As object a reconstruct LSN cluttered result, and effectively can backgrounds skeletons. suppress
As a result, LSN can effectively suppress the cluttered backgrounds and reconstruct object skeletons.
37,734
results proposed of state-of-the-art the validate performance LSN. the Experimental
Experimental results validate the state-of-the-art performance of the proposed LSN.
37,735
images. to human learning the in consider pose still task We estimate of
We consider the task of learning to estimate human pose in still images.
37,736
denoising application essential imaging. medical which Our is fluoroscopy is image in
Our application is medical image denoising which is essential in fluoroscopy imaging.
37,737
needs a physician, flavor and tailored each different to quality individual. image towards be has
physician, has a different flavor and image quality needs to be tailored towards each individual.
37,738
Also user specific best users test this perform on trained data. models a for
Also models trained for a specific user perform best on this users test data.
37,739
and spatial efficiency high owing is resolution. Brovey method its popular to pan-sharpening a The
The Brovey is a popular pan-sharpening method owing to its efficiency and high spatial resolution.
37,740
be method mathematical optical This sensing can model sensors. explained remote of by
This method can be explained by mathematical model of optical remote sensing sensors.
37,741
model of MNIST handwritten A is presented recognition digit here. simple
A simple model of MNIST handwritten digit recognition is presented here.
37,742
adaptation model is face recognition. of of an previous The theory a
The model is an adaptation of a previous theory of face recognition.
37,743
We present, of achieved example, way by the algorithms. by results state-of-the-art
We present, by way of example, the results achieved by state-of-the-art algorithms.
37,744
tasks this the team all in challenge. in participates Our
Our team participates in all the tasks in this challenge.
37,745
in a is attribute detection way. task, lesion In multi-label also PSPNet the adopted modified
In lesion attribute detection task, the modified PSPNet is also adopted in a multi-label way.
37,746
higher higher accuracy achieves SAW-SS than the cost sensing of computational complexity. at SAE-TF
SAE-TF achieves higher sensing accuracy than SAW-SS at the cost of higher computational complexity.
37,747
i.e. multiscale in applied method This version, is a
This method is applied in a multiscale version, i.e.
37,748
the variable index instead time. is scale of
the variable index is scale instead of time.
37,749
model consistent natural show with We is PAN MS our images. that and more
We show that our model is more consistent with natural MS and PAN images.
37,750
for sampling single diffusion schemes novel and MRI. This presents paper multi-shell
This paper presents novel single and multi-shell sampling schemes for diffusion MRI.
37,751
is . data used also qualitatively Human brain evaluate to reconstruction
Human brain data is also used to qualitatively evaluate reconstruction .
37,752
and algorithms systems family wireless in radar. communication The constant-modulus used widely is of in
The family of constant-modulus algorithms is widely used in wireless communication systems and in radar.
37,753
Instead, entire spectrum. it the equalizes spatial
Instead, it equalizes the entire spatial spectrum.
37,754
We a of example high-SNR provide under conditions. approach the numerical the demonstrate validity to
We provide a numerical example to demonstrate the validity of the approach under high-SNR conditions.
37,755
response. model counting strongest simply the by order, the the roots with
the model order, by simply counting the roots with the strongest response.
37,756
class a to justify agents Existing generating explanation learn prediction. fluently visual
Existing visual explanation generating agents learn to fluently justify a class prediction.
37,757
trust such human agents This concerning users. particularly building fail with is ultimately as in
This is particularly concerning as ultimately such agents fail in building trust with human users.
37,758
prediction, Our capable AI an for is providing arguments of alternative counter explainable agent i.e.
Our explainable AI agent is capable of providing counter arguments for an alternative prediction, i.e.
37,759
counterfactuals, along explanations that classification correct decisions. with the justify
counterfactuals, along with explanations that justify the correct classification decisions.
37,760
seamlessly method be into any proposed CNN The architecture. integrated general can and is
The proposed method is general and can be seamlessly integrated into any CNN architecture.
37,761
proposed experiments Extensive effectiveness the datasets four on approach. of benchmark show the
Extensive experiments on four benchmark datasets show the effectiveness of the proposed approach.
37,762
a and modified based on multi-task the a DeepUNet way. It perform is in segmentation
It is based on a modified DeepUNet and perform the segmentation in a multi-task way.
37,763
are different task groups The and into processed pipelines. clustered channels on
The channels are clustered into groups and processed on different task pipelines.
37,764
show network the fusion efficient. Experiment is that feature results
Experiment results show that the feature fusion network is efficient.
37,765
between controls strength the association unconditional parameters parameters. the top-level of of auxiliary One these
One of these top-level parameters controls the unconditional strength of association between the auxiliary parameters.
37,766
contrast algorithms, assumptions. distributed this approach In Carlo requires similar few to Monte distributional
In contrast to similar distributed Monte Carlo algorithms, this approach requires few distributional assumptions.
37,767
illustrated performance of of examples. algorithms with simulated is the number The a
The performance of the algorithms is illustrated with a number of simulated examples.
37,768
feature paper a of vision: In we attention. visual this focus specific on
In this paper we focus on a specific feature of vision: visual attention.
37,769
field-of-view review for of we extensive cameras. this provide paper, models large an existing In
In this paper, we provide an extensive review of existing models for large field-of-view cameras.
37,770
results We the also qualitative discuss performance models. provide of all and
We also provide qualitative results and discuss the performance of all models.
37,771
ability implicitly the vehicles to automated lack Today's others. cooperate with
Today's automated vehicles lack the ability to cooperate implicitly with others.
37,772
policies predefined policies within for algorithm simultaneously learns over the Without and macro-actions, macro-actions.
Without predefined policies for macro-actions, the algorithm simultaneously learns policies over and within macro-actions.
37,773
present and novel algorithm We for robust generating a model hypotheses consistent multiple-structure for fitting.
We present a novel algorithm for generating robust and consistent hypotheses for multiple-structure model fitting.
37,774
influence parameter on experiments. our settings clearly in of the the expose results We algorithm
We clearly expose the influence of algorithm parameter settings on the results in our experiments.
37,775
strategy multi-model an comparison. we testing propose effective Besides, also by applying
Besides, we also propose an effective testing strategy by applying multi-model comparison.
37,776
CCTV are surveillance views full where not often available. scenario, person practical This reflects
This reflects practical CCTV surveillance scenario, where full person views are often not available.
37,777
samples and gels, filled emulsion two-layered sausages. filled gels, emulsion were The
The samples were emulsion filled gels, two-layered emulsion filled gels, and sausages.
37,778
simplifying principle mastication a indicate and subsequent curve The may during master swallowing.
The master curve may indicate a simplifying principle during mastication and subsequent swallowing.
37,779
also particular complexity We have a calculated measure.
We have also calculated a particular complexity measure.
37,780
swallowing. an increase before displays This just measure
This measure displays an increase just before swallowing.
37,781
may of of different. In a transmitter addition, and time receiver rates a be
In addition, rates of time of a transmitter and a receiver may be different.
37,782
genetic might redundant. female and deplete variance However, choice preferences make
However, female preferences might deplete genetic variance and make choice redundant.
37,783
evolutionary empirical the biology. of We and our results to genetics relevance conservation discuss
We discuss the relevance of our results to conservation genetics and empirical evolutionary biology.
37,784
great vision computer neural have achieved successes in Convolutional (CNNs) many problems. networks
Convolutional neural networks (CNNs) have achieved great successes in many computer vision problems.
37,785
generalization on as Cityscapes) (e.g. as domain another well its capacity
Cityscapes) as well as its generalization capacity on another domain (e.g.
37,786
synthesized as images. in show results data unseen realistic We promising our well in as
We show promising results in our synthesized data as well as in unseen realistic images.
37,787
maximization. for paper, efficient deterministic this an algorithm optimization we In propose consensus
In this paper, we propose an efficient deterministic optimization algorithm for consensus maximization.
37,788
fact, worse the return a challenging previous solution. In techniques instances, even may off on
In fact, on challenging instances, the previous techniques may even return a worse off solution.
37,789
between since show challenging different very data detection Change can data the characteristics. be multimodal
Change detection between multimodal data can be very challenging since the data show different characteristics.
37,790
detect CNN between to the employed two candidate changes epochs. A Siamese then is
A Siamese CNN is then employed to detect candidate changes between the two epochs.
37,791
individual grouped the object patch-based changes. as Finally, changes candidate and verified are
Finally, the candidate patch-based changes are grouped and verified as individual object changes.
37,792
this so explored much However, not been far in direction. has
However, not much has so far been explored in this direction.
37,793
quality Meanwhile, the control. suggestions are on given photogrammetric
Meanwhile, suggestions are given on the photogrammetric quality control.
37,794
indefinite the to FAQ Due global find to maximum. guaranteed the relaxation, is not
Due to the indefinite relaxation, FAQ is not guaranteed to find the global maximum.
37,795
generalization superior demonstrating capabilities. proposed and the U-Net, learning methods The outperform
The proposed methods outperform the U-Net, demonstrating superior learning and generalization capabilities.
37,796
(coCSIT) focus channel covariance exploiting Here CSIT on only. we
Here we focus on exploiting channel covariance CSIT (coCSIT) only.
37,797
Spatial upsampling precision is learned achieved by edge-aware a function. employing
Spatial precision is achieved by employing a learned edge-aware upsampling function.
37,798
network Our image. Siamese and left from uses to right extract a model the features
Our model uses a Siamese network to extract features from the left and right image.
37,799