text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
relevance Video most of the one prediction streaming service. tasks online important for is
Video relevance prediction is one of the most important tasks for online streaming service.
36,000
video approach i.e. to analyzing promising content One solve itself, it is
One promising approach to solve it is analyzing video content itself, i.e.
36,001
predicting by relevance and subtitle video metadata. audio, frame, video
predicting video relevance by video frame, audio, subtitle and metadata.
36,002
experiments achieve accuracy with methods. traditional compared Our classification outstanding
Our experiments achieve outstanding classification accuracy compared with traditional methods.
36,003
from measurements. is a undersampled method Compressive recover image to the sensing original
Compressive sensing is a method to recover the original image from undersampled measurements.
36,004
methods to data. also They truth yielded ground with the performances comparable
They also yielded comparable performances to the methods with ground truth data.
36,005
in show are modifications neglected some of required We often which examples implementations.
We show some examples of required modifications which are often neglected in implementations.
36,006
(CP) technique propose In Cube follows. contrast, simple as effective Padding a and we
In contrast, we propose a simple and effective Cube Padding (CP) technique as follows.
36,007
methods in experiments, outperforms In speed our method and baseline both quality.
In experiments, our method outperforms baseline methods in both speed and quality.
36,008
be the should cost cost saved lower CFA. using Moreover, by the associated than
Moreover, the associated cost should be lower than the cost saved by using CFA.
36,009
a Inspired sensing weakness, low scheme. strength we edge very and cost design HA's by
Inspired by HA's strength and weakness, we design a very low cost edge sensing scheme.
36,010
between directional of it Briefly, variations. functional a difference logistic guides by demosaicking the
Briefly, it guides demosaicking by a logistic functional of the difference between directional variations.
36,011
theoretical experiments Carlo Monte the in setting dynamic to conducted finding. data support are panel
Monte Carlo experiments are conducted in dynamic panel data setting to support the theoretical finding.
36,012
result that higher. suggests the be The may to returns schooling
The result suggests that the returns to schooling may be higher.
36,013
surgical to outcomes costly patients Adverse and hospitals. are
Adverse surgical outcomes are costly to patients and hospitals.
36,014
efficiency performance of and suggest a framework. results generation our high morphology The
The results suggest a high efficiency and performance of our morphology generation framework.
36,015
Each standard backpropagation. learns through with a along specialization of its kernel location weights
Each kernel learns a location of specialization along with its weights through standard backpropagation.
36,016
ROIs. differentiable optimize A supervision attention to is no the used requiring additional mechanism
A differentiable attention mechanism requiring no additional supervision is used to optimize the ROIs.
36,017
synthetic domains. architecture both We natural tasks proposed multiple on and evaluate across our
We evaluate our proposed architecture on both synthetic and natural tasks across multiple domains.
36,018
et. approach by extend (Xie (BNs) the learning We for networks Bayesian proposed decomposition
We extend the decomposition approach for learning Bayesian networks (BNs) proposed by (Xie et.
36,019
CGs), graphs special as which regression to multivariate (MVR chain BNs case. a learning include
to learning multivariate regression chain graphs (MVR CGs), which include BNs as a special case.
36,020
running except PC-like the algorithm the decomposition-based in outperforms fact, time. algorithm usually In
In fact, the decomposition-based algorithm usually outperforms the PC-like algorithm except in running time.
36,021
when sparse. better The is is of underlying the much algorithms graph both performance
The performance of both algorithms is much better when the underlying graph is sparse.
36,022
believe field. the be in We researchers instructive this for will directions potential these
We believe these potential directions will be instructive for the researchers in this field.
36,023
aims transfer arbitrary to content Style styles to visual images. transfer
Style transfer aims to transfer arbitrary visual styles to content images.
36,024
strategies appealing compare measure these We visually images. up to and compare they produce how
We compare these strategies and compare how they measure up to produce visually appealing images.
36,025
adopt recurrence or based mainly methods Existing networks. convolution
Existing methods mainly adopt recurrence or convolution based networks.
36,026
a supervised neuroanatomical We propose for network neural shapes. deep learning on
We propose a deep neural network for supervised learning on neuroanatomical shapes.
36,027
unary and DCCRF consists The terms. pairwise of
The DCCRF consists of unary and pairwise terms.
36,028
objects' unary based across displacements on tracked estimate appearance time visual The information. terms
The unary terms estimate tracked objects' displacements across time based on visual appearance information.
36,029
visual use learning makes spectral from to range. a transfer It wider to possible
It makes possible to use transfer learning from visual to a wider spectral range.
36,030
This real-time drone a paper areas. introduces individuals in violent to system public surveillance identify
This paper introduces a real-time drone surveillance system to identify violent individuals in public areas.
36,031
Network images. from uses detect to first The aerial system Pyramid Feature humans the
The system first uses the Feature Pyramid Network to detect humans from aerial images.
36,032
identification and pose is violent the with compared estimation The individuals state-of-the-art techniques. performance
The pose estimation and violent individuals identification performance is compared with the state-of-the-art techniques.
36,033
lesions cystic classification for on certain achieved have Previous CT. in segmented accuracy computer-aided developments
Previous computer-aided developments have achieved certain accuracy for classification on segmented cystic lesions in CT.
36,034
to data. We novel iterative to a image propose graph adapt d-dimensional a a method
We propose a novel iterative method to adapt a a graph to d-dimensional image data.
36,035
the image The features. graph drives of the method nodes towards
The method drives the nodes of the graph towards image features.
36,036
method fully is parallelisable. efficient proposed and The computationally
The proposed method is computationally efficient and fully parallelisable.
36,037
behavior finite In sample the Monte-Carlo simulations analyze we of estimator. our
In Monte-Carlo simulations we analyze the finite sample behavior of our estimator.
36,038
health Connected security and is communication technologies. devices, medical technology devices that a associates
Connected health is a technology that associates medical devices, security devices and communication technologies.
36,039
and be enables patients to home. It monitored remotely treated their from
It enables patients to be monitored and treated remotely from their home.
36,040
detection. investigation division) A (cell step the fundamental mitotic is cell of
A fundamental step of the investigation is mitotic cell (cell division) detection.
36,041
recognition. action complementary information key for temporal features are two human Spatial and and
Spatial and temporal features are two key and complementary information for human action recognition.
36,042
feature recalibration. operations the and implement to are Squeeze Excitation The used
The Squeeze and Excitation operations are used to implement the feature recalibration.
36,043
is The of proposed using the minimization energy optimization. performed stochastic
The minimization of the proposed energy is performed using stochastic optimization.
36,044
of dataset order created the degraded to method. images realistically consisting is in evaluate A
A dataset consisting of realistically degraded images is created in order to evaluate the method.
36,045
source information flow in of the is estimation. optical context Temporal coherence valuable of a
Temporal coherence is a valuable source of information in the context of optical flow estimation.
36,046
However, suitable a a motion leverage non-trivial finding task. model information is this to
However, finding a suitable motion model to leverage this information is a non-trivial task.
36,047
usefulness demonstrate Experiments our approach. new the of
Experiments demonstrate the usefulness of our new approach.
36,048
the search Similarity of of strategy speed final searching and accelerates accuracy the results. boosts
Similarity search strategy accelerates the speed of searching and boosts the accuracy of final results.
36,049
years, (DL) biomedical powerful In learning segmentation. recent have image become deep methods tools for
In recent years, deep learning (DL) methods have become powerful tools for biomedical image segmentation.
36,050
inserted devices. supporting Catheters commonly are life
Catheters are commonly inserted life supporting devices.
36,051
results. deep learning, resurgence With begining are of the supervised training showing approaches to promising
With the resurgence of deep learning, supervised training approaches are begining to showing promising results.
36,052
algorithms that suggests the Experimental RED evidence state-of-the-art. are
Experimental evidence suggests that the RED algorithms are state-of-the-art.
36,053
that the explicit does regularization We not algorithms. claim, explain RED however,
We claim, however, that explicit regularization does not explain the RED algorithms.
36,054
al. Furthermore, et interpret the algorithms Romano from we RED
Furthermore, we interpret the RED algorithms from Romano et al.
36,055
algorithms new and acceleration guarantees. and convergence propose with
and propose new algorithms with acceleration and convergence guarantees.
36,056
in analysis and variance models many regressors. tests with of restrictions of linear Applications include
Applications include analysis of variance and tests of linear restrictions in models with many regressors.
36,057
RGB vs sketch near-infrared. vs or RGB
RGB vs sketch or RGB vs near-infrared.
36,058
these of optimal. none We is scenarios two that observed
We observed that none of these two scenarios is optimal.
36,059
monocular Automatically challenging determining from image three-dimensional RGB human is pose a data problem.
Automatically determining three-dimensional human pose from monocular RGB image data is a challenging problem.
36,060
maintaining coherent heatmaps whilst differentiability. visually produces model, MargiPose, Our resulting
Our resulting model, MargiPose, produces visually coherent heatmaps whilst maintaining differentiability.
36,061
segmentation topic processing computer vision. a and in Image is image research fundamental
Image segmentation is a fundamental research topic in image processing and computer vision.
36,062
the of algorithms large decades, applications. for number developed various In a segmentation last researchers
In the last decades, researchers developed a large number of segmentation algorithms for various applications.
36,063
undesirable results. to lead will segmentation This some flaw
This flaw will lead to some undesirable segmentation results.
36,064
method can Our measure similarity the adaptively update estimating by parameters. function some
Our method can adaptively update the similarity measure function by estimating some parameters.
36,065
computational cost greater of at have the networks power increased representational generally Larger complexity.
Larger networks generally have greater representational power at the cost of increased computational complexity.
36,066
as all-in-focus can generated potentially images The reference be images. used
The generated images can potentially be used as reference all-in-focus images.
36,067
the achieves dataset. competitive and state-of-the-art on on the results performance FVTA MovieQA MemexQA dataset
FVTA achieves state-of-the-art performance on the MemexQA dataset and competitive results on the MovieQA dataset.
36,068
for temperatures calculation method A presented. engine is
A calculation method for engine temperatures is presented.
36,069
behaviors thermal measuring of can be various simulated positions successfully. Different
Different thermal behaviors of various measuring positions can be simulated successfully.
36,070
Previous jointly successful interactions. and model social physical been work to has not
Previous work has not been successful to jointly model physical and social interactions.
36,071
integration foundation the and Concepts learning, human understanding, of and knowledge deep transfer. are
Concepts are the foundation of human deep learning, understanding, and knowledge integration and transfer.
36,072
stable behavior and is general is maximum learning How reached? once
How stable and general is behavior once maximum learning is reached?
36,073
of Harris al. experiment the Using Pavlovian data et by and this another
Using the data of this and another Pavlovian experiment by Harris et al.
36,074
the research personalized of empowering for is areas education. one key tracing Knowledge
Knowledge tracing is one of the key research areas for empowering personalized education.
36,075
problems our have in major extensive two model. experimentation, through However, the we DKT noticed
However, through our extensive experimentation, we have noticed two major problems in the DKT model.
36,076
the The that the fails model reconstruct first is problem to input. observed
The first problem is that the model fails to reconstruct the observed input.
36,077
time-steps predicted consistent. KCs not the for across is performance Second,
Second, the predicted performance for KCs across time-steps is not consistent.
36,078
time. and expected transit gradually to over performance is because is undesirable This student's unreasonable
This is undesirable and unreasonable because student's performance is expected to transit gradually over time.
36,079
Two variables. sets and are provided: data invalid Combining examples instrumental
Two examples are provided: Combining data sets and invalid instrumental variables.
36,080
adaptive indirect An variations to is fuzzy parameter resistant proposed. controller, then
An indirect adaptive fuzzy controller, resistant to parameter variations is then proposed.
36,081
structure of The first in specified the the controller is general stage. adaptive
The general structure of the adaptive controller is specified in the first stage.
36,082
of two second In fuzzy with its are stage, regulated aid the systems. parameters the
In the second stage, its parameters are regulated with the aid of two fuzzy systems.
36,083
approaches Finally, conventional compared. and results the of proposed are the the
Finally, the results of the proposed and the conventional approaches are compared.
36,084
configurations, For observed no or we contrast inducers' color the other induction.
For the other inducers' configurations, we observed color contrast or no induction.
36,085
representation solution to tracking. a is visual Sparse viable
Sparse representation is a viable solution to visual tracking.
36,086
With system a limited power, task. becomes develop how a computing challenging robust to
With limited computing power, how to develop a robust system becomes a challenging task.
36,087
are small inflammation. and related bioactive species mediators critical, human and physiology Eicosanoids of
Eicosanoids and related species are critical, small bioactive mediators of human physiology and inflammation.
36,088
of diagnostic a making. essential for decision measure Furthermore, uncertainty is
Furthermore, a measure of uncertainty is essential for diagnostic decision making.
36,089
energy makes Mitochondrial currency phosphorylation oxidative (mOxPhos) ATP, of life. the
Mitochondrial oxidative phosphorylation (mOxPhos) makes ATP, the energy currency of life.
36,090
is so, channel new combination create crucial. merged In layers, to FMs are convolutional FMs,
In convolutional layers, FMs are merged to create new FMs, so, channel combination is crucial.
36,091
class. all not relevance FMs given However, the have a same for
However, not all FMs have the same relevance for a given class.
36,092
we brain evaluate segmentation, data. available using Finally, the in tumor proposed publicly methods
Finally, we evaluate the proposed methods in brain tumor segmentation, using publicly available data.
36,093
Pavement that capacity. one road important is the most affects of the reasons cracks
Pavement cracks is one of the most important reasons that affects the road capacity.
36,094
important new Therefore, significance. has a detection designing method
Therefore, designing a new detection method has important significance.
36,095
Hausdorff evaluate is used distance At to detection precision. last,
At last, Hausdorff distance is used to evaluate detection precision.
36,096
However, of urban phenotypic remains on traits selection directional in empirical scarce. evidence areas
However, empirical evidence of directional selection on phenotypic traits in urban areas remains scarce.
36,097
between survival intensities. difference different of colouration was local in adults No found
No difference in local survival was found between adults of different colouration intensities.
36,098
critical rate in been deep issue network Learning training. scheduler has the neural a
Learning rate scheduler has been a critical issue in the deep neural network training.
36,099