text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
clustering In iteratively optimized. this manner, image is
In this manner, image clustering is iteratively optimized.
38,500
Regenerative the by train braking is energy produced deceleration. a energy during
Regenerative braking energy is the energy produced by a train during deceleration.
38,501
and train a as produce motors decelerates, generators electricity. act the When
When a train decelerates, the motors act as generators and produce electricity.
38,502
paper of to regenerative aim braking energy. and is this simulate The model
The aim of this paper is to model and simulate regenerative braking energy.
38,503
examples parameters the influence network These to when trained can them. correct strongly is
These examples can strongly influence parameters when the network is trained to correct them.
38,504
often in training and they obtain. sparse Unfortunately, data, expensive to the are are
Unfortunately, they are often sparse in the training data, and are expensive to obtain.
38,505
improves experiments on obtained automatically show these significantly performance. examples that Our retraining detectors often
Our experiments show that retraining detectors on these automatically obtained examples often significantly improves performance.
38,506
such influences movement. We greatly future path of prediction that information show common
We show that such common path information greatly influences prediction of future movement.
38,507
that further filter information. movement scene non-linear a design holds We important data
We further design a scene data filter that holds important non-linear movement information.
38,508
We datasets. model on evaluate two challenging our
We evaluate our model on two challenging datasets.
38,509
memory this However, RAM-like and is objects. unstructured does not matrix naturally thus structured encode
However, this RAM-like memory matrix is unstructured and thus does not naturally encode structured objects.
38,510
queries objects programming. about to learns without a of graph-structured explicit set answer RMDN Thus
Thus RMDN learns to answer queries about a set of graph-structured objects without explicit programming.
38,511
demonstrate model. of proposed Results the efficacy the
Results demonstrate the efficacy of the proposed model.
38,512
step transition Detection of shot video crucial video is analysis. pre-processing in a
Detection of video shot transition is a crucial pre-processing step in video analysis.
38,513
smart trade-offs, framework. Considering performance a we speed design
Considering speed performance trade-offs, we design a smart framework.
38,514
Our autoencoder the branches, for consists structure two for and one geometry. of the one
Our autoencoder consists of two branches, one for the structure and one for the geometry.
38,515
evaluate study. an and performance conduct method the ablation of We our
We evaluate the performance of our method and conduct an ablation study.
38,516
We geometry. that similarities of explicitly for structure shapes show encoding both in and accounts
We explicitly show that encoding of shapes accounts for both similarities in structure and geometry.
38,517
generated variety of results presented. A SAGNet quality are by
A variety of quality results generated by SAGNet are presented.
38,518
multibox SSD---single improvements shot propose two We detector. to the
We propose two improvements to the SSD---single shot multibox detector.
38,519
approach in SSD. First, default propose an selection we adaptive box for
First, we propose an adaptive approach for default box selection in SSD.
38,520
environments in ways cannot. fluidly Humans their machines adapt interest in can complex that
Humans can fluidly adapt their interest in complex environments in ways that machines cannot.
38,521
twofold. of The work are this contributions
The contributions of this work are twofold.
38,522
issues, a To signaling we antenna regularized counteract selection implementation practical scheme. based propose
To counteract practical implementation issues, we propose a regularized antenna selection based signaling scheme.
38,523
rates analyzed rate error the is The symbol/bit and derived. are data possible
The possible data rate is analyzed and the symbol/bit error rates are derived.
38,524
pairwise discriminative PRN The among to is capture identities. and relations unique trained different
The PRN is trained to capture unique and discriminative pairwise relations among different identities.
38,525
recent rapidly Service growing in has (VA been a Analytics Video as years. Software SaaS)
Video Analytics Software as a Service (VA SaaS) has been rapidly growing in recent years.
38,526
accessed a lightweight is by users VA using typically SaaS client.
VA SaaS is typically accessed by users using a lightweight client.
38,527
proposed TSD advantages. three The appealing has
The proposed TSD has three appealing advantages.
38,528
Firstly, TSD the architecture and deployed eg. lightweight can be a client, in has
Firstly, TSD has a lightweight architecture and can be deployed in the client, eg.
38,529
save transmission frames mobile bandwidth. devices, to to compressed representative produce
mobile devices, to produce compressed representative frames to save transmission bandwidth.
38,530
Extensive show experiments characteristics and the effectiveness TSD. of
Extensive experiments show the effectiveness and characteristics of TSD.
38,531
scheme no such the effects. partition statistical nonlinearity to Furthermore, devised biodiversity can under be
Furthermore, under such nonlinearity no statistical scheme can be devised to partition the biodiversity effects.
38,532
alternative more We estimate present a metric also biodiversity providing of effect. reasonable an
We also present an alternative metric providing a more reasonable estimate of biodiversity effect.
38,533
features its effectiveness. has for recently shown Aggregating different image retrieval image
Aggregating different image features for image retrieval has recently shown its effectiveness.
38,534
we a the end, distribution. entropy To exploit membership-score of cluster this
To this end, we exploit the entropy of a cluster membership-score distribution.
38,535
NN addition, voting a bypasses scheme. heuristic In final the set
In addition, the final NN set bypasses a heuristic voting scheme.
38,536
on Experiments several benchmark result. show datasets our retrieval method state-of-the-art the that can improve
Experiments on several retrieval benchmark datasets show that our method can improve the state-of-the-art result.
38,537
of in effect transfer learning the transfer of both settings. beneficial Results indicate experiments
Results of experiments indicate the beneficial effect of transfer learning in both transfer settings.
38,538
using in increase sensitivity. that gives an additional Results exam level indicate annotations
Results indicate that using exam level annotations gives an additional increase in sensitivity.
38,539
the in are decision-making belief reviewed. to uncertainty function under Approaches framework
Approaches to decision-making under uncertainty in the belief function framework are reviewed.
38,540
However, bit-widths drastically degraded with quantization yields generally accuracy. decreasing
However, decreasing bit-widths with quantization generally yields drastically degraded accuracy.
38,541
an essential and inherent of part Response is human actions. delay
Response delay is an inherent and essential part of human actions.
38,542
To simple the data this, support task. balancing a virtual present experimental stick from we
To support this, we present the experimental data from a simple virtual stick balancing task.
38,543
architecture. problem in utilizing address mechanism routing-by We developed this our recently the agreement by
We address this problem by utilizing the recently developed routing-by agreement mechanism in our architecture.
38,544
is problem important robotics. an relocalization Image-based in camera vision computer and
Image-based camera relocalization is an important problem in computer vision and robotics.
38,545
final coordinates. then optimization scheme RANSAC-based a the predicted solved is via The using pose
The final pose is then solved via a RANSAC-based optimization scheme using the predicted coordinates.
38,546
to The multi-view loss which enables available also new us utilize further constraints, improve performance.
The new loss also enables us to utilize available multi-view constraints, which further improve performance.
38,547
predictions also refined. three blocks the of can Meanwhile, nearest the be
Meanwhile, the predictions of the three nearest blocks can also be refined.
38,548
based optimization are cost on The weighted landmark in errors. function points these the
The landmark points are weighted in the optimization cost function based on these errors.
38,549
different Obviously, locate on semantics landmarks facial these because components. have different they
Obviously, these landmarks have different semantics because they locate on different facial components.
38,550
necessary each consider individually. is to it landmark Thus,
Thus, it is necessary to consider each landmark individually.
38,551
landmarks. The weight the adaptive residuals is with estimation of
The weight is adaptive with the estimation residuals of landmarks.
38,552
applications of robotic important Rotation manipulation. as dexterous is for such rigid known estimation objects
Rotation estimation of known rigid objects is important for robotic applications such as dexterous manipulation.
38,553
a rank ordering Elo across method broad The dataset. plausible the produced
The Elo method produced a plausible rank ordering across the broad dataset.
38,554
of to information, those relating correlation with degree highest spatial e.g. The the were metrics
The metrics with the highest degree of correlation were those relating to spatial information, e.g.
38,555
spatial this on the variation was measured. agreement scale However, was dependent from the which
However, this agreement was dependent on the scale from which the spatial variation was measured.
38,556
lacunarity, will and many for presented including image measures entropy. also other Results be compression,
Results will also be presented for many other measures including lacunarity, image compression, and entropy.
38,557
on We and method PASCAL COCO MS dataset. evaluate VOC our
We evaluate our method on VOC PASCAL and MS COCO dataset.
38,558
on and parameters. structures We by conducting demonstrate several experiments SAN
We demonstrate SAN by conducting several experiments on structures and parameters.
38,559
objects with smooth a interaction in for virtual environments experience. user AR/VR makes Natural
Natural interaction with virtual objects in AR/VR environments makes for a smooth user experience.
38,560
achieve world to space these a to are extension interactions. augmented Gestures natural from real
Gestures are a natural extension from real world to augmented space to achieve these interactions.
38,561
The of the and is amenable timescales. averaging to theory synapse assumption slow method multiple
The slow synapse assumption is amenable to averaging theory and the method of multiple timescales.
38,562
demonstrate We examples. theory our two then using
We then demonstrate our theory using two examples.
38,563
we In framework a modeling propose deformable unified work for this material.
In this work we propose a unified framework for modeling deformable material.
38,564
explicitly network a encapsulates complex to The is neural hard function that model.
The neural network encapsulates a complex function that is hard to explicitly model.
38,565
synthetic a of with set is examples. demonstrated method Our
Our method is demonstrated with a set of synthetic examples.
38,566
Generative Adversarial Network Gesture hand this a (GestureGAN). we tackle propose novel problem, To
To tackle this problem, we propose a novel hand Gesture Generative Adversarial Network (GestureGAN).
38,567
loss back-propagating while of issue proposed gradients. the color the "channel handles pollution" The
The proposed color loss handles the issue of "channel pollution" while back-propagating the gradients.
38,568
experiments the FIMA. verify real on superiority of applications Extensive
Extensive experiments on real applications verify the superiority of FIMA.
38,569
load with models. simulated power approach systems tested different in increment is various Our
Our approach is tested in various simulated power systems with different load increment models.
38,570
collapse for approach speed cases our the of and point. show Other estimating accuracy voltage
Other cases show the accuracy and speed of our approach for estimating voltage collapse point.
38,571
and approach. successfully effectiveness These testing cases the accuracy of demonstrate our
These testing cases successfully demonstrate the accuracy and effectiveness of our approach.
38,572
images processing pipelines. image medical of essential for is of automation Quality assessment complete
Quality assessment of medical images is essential for complete automation of image processing pipelines.
38,573
need image an assessment there techniques. for quality urgent is automatic Therefore,
Therefore, there is an urgent need for automatic image quality assessment techniques.
38,574
to methods. compare state-of-the-art range of quality We assessment our a approach
We compare our approach to a range of state-of-the-art quality assessment methods.
38,575
Comparisons presented. with from ImageNet and scratch classifiers classifiers are pre-trained trained
Comparisons with ImageNet pre-trained classifiers and classifiers trained from scratch are presented.
38,576
are on factors that Covariates debilitating face have influence a verification performance.
Covariates are factors that have a debilitating influence on face verification performance.
38,577
both and of faces. characteristics intrinsic covariates cover factors extrinsic These subject-specific
These covariates cover both intrinsic subject-specific characteristics and extrinsic factors of faces.
38,578
more patches representative. At every smaller extracted step, the the and levels are than previous
At every step, the patches extracted are smaller than the previous levels and more representative.
38,579
is to detect therefore crucial images out-of-distribution The applications. capability real-world for many
The capability to detect out-of-distribution images is therefore crucial for many real-world applications.
38,580
concept, new the RPIfield. carefully collected multi-shot dataset To person called re-id illustrate a we
To carefully illustrate the concept, we collected a new multi-shot person re-id dataset called RPIfield.
38,581
is systems simulation validated in using DFRC of examples. utilization The sparse arrays
The utilization of sparse arrays in DFRC systems is validated using simulation examples.
38,582
and sensing applications. task a remote challenging aerial Vehicle is detection significant in
Vehicle detection is a significant and challenging task in aerial remote sensing applications.
38,583
be our R-RPN In can model, R-DN trained and jointly.
In our model, R-RPN and R-DN can be trained jointly.
38,584
with is a newly-recorded original the lip task. Manual tedious the movements of alignment speech
Manual alignment of the newly-recorded speech with the original lip movements is a tedious task.
38,585
successful well a human quantitatively, alignment both as using We perception-inspired as demonstrate metric, qualitatively.
We demonstrate successful alignment both quantitatively, using a human perception-inspired metric, as well as qualitatively.
38,586
state-of-the-art methods will cases these In fail.
In these cases state-of-the-art methods will fail.
38,587
the the could of system. off rate pulse be improved further trading These performances by
These performances could be further improved by trading off the pulse rate of the system.
38,588
work focused cameras. has forward-facing exclusively Recent on vision automotive processing almost
Recent automotive vision work has focused almost exclusively on processing forward-facing cameras.
38,589
which powerful. architecture image simple fully-convolutional yet a for propose denoising We neural-network is
We propose a fully-convolutional neural-network architecture for image denoising which is simple yet powerful.
38,590
collected code the the publish in analysis. dataset, We InstaBarcelona used and the
We publish the collected dataset, InstaBarcelona and the code used in the analysis.
38,591
In a (Neural novel this paper, propose Fitting (NBF)). Body approach we
In this paper, we propose a novel approach (Neural Body Fitting (NBF)).
38,592
core-shell diffusion a modeled. Fickian geometry is into
Fickian diffusion into a core-shell geometry is modeled.
38,593
shell core Langerhan interior The mimics acts as exterior islets pancreatic and inert protection. the
The interior core mimics pancreatic Langerhan islets and the exterior shell acts as inert protection.
38,594
the oxygen The using consumption is diffusing approximated Michaelis-Menten cells into kinetics. of
The consumption of oxygen diffusing into the cells is approximated using Michaelis-Menten kinetics.
38,595
problem to units and numerically. is transformed The solved dimensionless
The problem is transformed to dimensionless units and solved numerically.
38,596
diffusion are other limited. is regimes one Two the consumption limited that and identified,
Two regimes are identified, one that is diffusion limited and the other consumption limited.
38,597
the and hand, separating good identification scalable has On property. other loss
On the other hand, identification loss has good separating and scalable property.
38,598
of material for representation the ubiquitous BRDF tools are models appearance.
BRDF models are ubiquitous tools for the representation of material appearance.
38,599