text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
a for human in novel We recognizing videos. method such present activities
We present a novel method for recognizing human activities in such videos.
37,200
maintains activity. information for the recognizing the proposed However, the system motion required
However, the proposed system maintains the required motion information for recognizing the activity.
37,201
of plays classification in key nephropathy. diagnosis glomerulus a Pathological role the
Pathological glomerulus classification plays a key role in the diagnosis of nephropathy.
37,202
alternative for provides paper glomerulus an for multi-stained This cues visual classification. solution integrating
This paper provides an alternative solution for integrating multi-stained visual cues for glomerulus classification.
37,203
in We these joint a stages manner. optimize two
We optimize these two stages in a joint manner.
37,204
undergo the are that Objects common real deformation in non-rigid world.
Objects that undergo non-rigid deformation are common in the real world.
37,205
A and is example human faces. challenging typical the
A typical and challenging example is the human faces.
37,206
and both classification are tasks. These registration useful for quantitative of evaluation
These are useful for quantitative evaluation of both registration and classification tasks.
37,207
Rule Set for Multi-vAlue in-hospital model patient (MRS) predicting a We propose mortality.
We propose a Multi-vAlue Rule Set (MRS) model for in-hospital predicting patient mortality.
37,208
of blood for management. is glucose diabetes essential Control
Control of blood glucose is essential for diabetes management.
37,209
Performance convolutional of benchmarked is network recurrent neural against algorithms. the four
Performance of the recurrent convolutional neural network is benchmarked against four algorithms.
37,210
share pooling. We predictors max all between with scales box and replace across scales, convolution
We share box predictors across all scales, and replace convolution between scales with max pooling.
37,211
model these SSD. changes hurt to empirically We show not that quality vanilla compared do
We empirically show that these changes do not hurt model quality compared to vanilla SSD.
37,212
background or light was applied. artificial no
no artificial background or light was applied.
37,213
and with approach It high-throughput. novel non-invasive, inexpensive objective for is phenotyping a
It is a novel approach for non-invasive, inexpensive and objective phenotyping with high-throughput.
37,214
image First, inflorescences regions depicting and were identified localized.
First, image regions depicting inflorescences were identified and localized.
37,215
using "inflorescence"-areas from the flowers were Hough Finally, Circular individual extracted Transform.
Finally, individual flowers were extracted from the "inflorescence"-areas using Circular Hough Transform.
37,216
is the reported Australian Age-progression in on data. cancer risk then
Age-progression in the cancer risk is then reported on Australian data.
37,217
of tracking and estimation useful is Head pose medical applications. in variety
Head pose estimation and tracking is useful in variety of medical applications.
37,218
increased without to important the patient is head For track markers. comfort, it
For increased patient comfort, it is important to track the head without markers.
37,219
pose using different head this regard, In the estimation address problem approaches. we two
In this regard, we address the head pose estimation problem using two different approaches.
37,220
are obtained analog (ADC). to by samples a The digital (SR) modulo self-reset converter
The modulo samples are obtained by a self-reset (SR) analog to digital converter (ADC).
37,221
a the propose also we by functions, second-order-statistics objective FT-PCA. Motivated approach, geometrical
Motivated by the geometrical objective functions, we also propose a second-order-statistics approach, FT-PCA.
37,222
of upon supervised Under category built the detectors are usually weakly supervision classifiers. labels,
Under the supervision of category labels, weakly supervised detectors are usually built upon classifiers.
37,223
been few objects environments. in uncontrolled has However, research there generating real-world
However, there has been few research generating objects in uncontrolled real-world environments.
37,224
this vehicle approach in a scenes. we paper, image for In propose novel generation real-world
In this paper, we propose a novel approach for vehicle image generation in real-world scenes.
37,225
object networks neural largely detection. Region-based (R-CNN)~\cite{fast_rcnn,faster_rcnn,mask_rcnn} have dominated convolutional
Region-based convolutional neural networks (R-CNN)~\cite{fast_rcnn,faster_rcnn,mask_rcnn} have largely dominated object detection.
37,226
propose this network paper, WSOD. novel deep In we for a
In this paper, we propose a novel deep network for WSOD.
37,227
The adjacent spatially the proposals cluster associated are same and the with in object. same
The proposals in the same cluster are spatially adjacent and associated with the same object.
37,228
ImageNet are for MS-COCO Experiments the benchmarks conducted on VOC, WSOD. and detection, PASCAL
Experiments are conducted on the PASCAL VOC, ImageNet detection, and MS-COCO benchmarks for WSOD.
37,229
of method outperforms our previous Results art the that state significantly. the show
Results show that our method outperforms the previous state of the art significantly.
37,230
outliers module false to reduce used filter alarms. post-processing A and is out
A post-processing module is used to filter out outliers and reduce false alarms.
37,231
Typical guidance or texture local includes, coordinates field. a e.g., values, normal displacement
Typical local guidance includes, e.g., normal values, texture coordinates or a displacement field.
37,232
our transfer style techniques, approach expensive Contrary not does previous involve computationally any to optimization.
Contrary to previous style transfer techniques, our approach does not involve any computationally expensive optimization.
37,233
maintain of matrix scattering mode also can information encoding completely. This polarimetric
This encoding mode can also maintain polarimetric information of scattering matrix completely.
37,234
observers. by OCT intra-coronary performed manually using characterization is images expert plaque Commonly,
Commonly, plaque characterization using intra-coronary OCT images is performed manually by expert observers.
37,235
in result Creative journeys follow memes, or unique works, form. whether their that paintings final
Creative works, whether paintings or memes, follow unique journeys that result in their final form.
37,236
irregular for inpainting approach This a holes. to semi-parametric image introduces paper
This paper introduces a semi-parametric approach to image inpainting for irregular holes.
37,237
nonparametric of database. part The an external consists image
The nonparametric part consists of an external image database.
37,238
Markov on transition a based first of We multi-view probability chain. matrices construct the tensor
We first construct a tensor based on multi-view transition probability matrices of the Markov chain.
37,239
method be direction efficiently of alternating can optimized by method multipliers~(ADMM). The the proposed
The proposed method can be efficiently optimized by the alternating direction method of multipliers~(ADMM).
37,240
or supervised either unsupervised video Most based are learning. on summarisation existing methods
Most existing video summarisation methods are based on either supervised or unsupervised learning.
37,241
to rewards training for A DQSN. trained provide network is also the companion classification
A companion classification network is also trained to provide rewards for training the DQSN.
37,242
classification network, a result. reward on the based global the we recognisability With classification develop
With the classification network, we develop a global recognisability reward based on the classification result.
37,243
on proposed achieves show state-of-the-art the approach performance. two that experiments datasets benchmark Extensive
Extensive experiments on two benchmark datasets show that the proposed approach achieves state-of-the-art performance.
37,244
powerful proposed code-rates with and varying The constructs block-lengths. scheme LDPC codes
The proposed scheme constructs powerful LDPC codes with varying code-rates and block-lengths.
37,245
systems. the gains obtained over compared show The large results
The results obtained show large gains over the compared systems.
37,246
for (EL) the photovoltaic of useful inspection a modules. Electroluminescence is imaging modality (PV)
Electroluminescence (EL) imaging is a useful modality for the inspection of photovoltaic (PV) modules.
37,247
which application are differ their dictated by their hardware in requirements, approaches respective scenarios. The
The approaches differ in their hardware requirements, which are dictated by their respective application scenarios.
37,248
compare processing variants. To and a performance, various strong investigate obtain we
To obtain a strong performance, we investigate and compare various processing variants.
37,249
make PV highly approaches cells of Both automated continuous, monitoring accurate feasible.
Both automated approaches make continuous, highly accurate monitoring of PV cells feasible.
37,250
bilateral major with significant have alterations white-matter fascicles. Patients IGE in
Patients with IGE have significant bilateral alterations in major white-matter fascicles.
37,251
scale-invariant, orientation- affine and from An (ACs) for approach is proposed correspondences recovering e.g.
An approach is proposed for recovering affine correspondences (ACs) from orientation- and scale-invariant, e.g.
37,252
locally RANSAC, magnitude order optimized is faster. of an by
by locally optimized RANSAC, is an order of magnitude faster.
37,253
particular factors depend?" the of heuristics ", the performance does (relative) which "On
", "On which factors does the (relative) performance of the particular heuristics depend?"
37,254
circumstances use or heuristic?" "Under I which should which
or "Under which circumstances should I use which heuristic?"
37,255
UWCNN train underwater for models each image we multiple separately type. formation Then,
Then, we separately train multiple UWCNN models for each underwater image formation type.
37,256
the cross-validation and V-Net of testing. outperformed used both implementation Results: the U-Net The in
Results: The U-Net outperformed the used implementation of the V-Net in both cross-validation and testing.
37,257
task We and novel subject prediction: on Variability also identity perspective BOLD (BV). introduce a
We also introduce a novel perspective on task and subject identity prediction: BOLD Variability (BV).
37,258
each based variance Conceptually, a region-specific brain BV is within region. the on measure
Conceptually, BV is a region-specific measure based on the variance within each brain region.
37,259
to compute, simple interpret, is and BV visualize.
BV is simple to compute, interpret, and visualize.
37,260
extremely unfavourable Underwater from suffer conditions. images
Underwater images suffer from extremely unfavourable conditions.
37,261
veiling scattering causes called change creates so hue, in Attenuation light.
Attenuation creates change in hue, scattering causes so called veiling light.
37,262
its understanding effects. lending provides a approach deeper the and of This bank channel
This approach provides a deeper understanding of the bank lending channel and its effects.
37,263
A disparity between can the observed. be countries small north-south
A small north-south disparity between the countries can be observed.
37,264
create to are and architectures a vision necessary AGI? What system for frameworks
What frameworks and architectures are necessary to create a vision system for AGI?
37,265
existing and our demonstrate some discriminative their We discuss for generative purposes. insufficiency models and
We discuss some existing generative and discriminative models and demonstrate their insufficiency for our purposes.
37,266
some dilemmas open and Finally, we discuss architectural questions.
Finally, we discuss some architectural dilemmas and open questions.
37,267
at the the blockage. likelihood BS Placing height reduces greater a of
Placing the BS at a greater height reduces the likelihood of blockage.
37,268
Kingdomino. surprisingly is shows MCE for than game like The MCTS that a evaluation stronger
The evaluation shows that surprisingly MCE is stronger than MCTS for a game like Kingdomino.
37,269
novel decoupled we paper, propose In a filtering this algorithm. Kalman
In this paper, we propose a novel decoupled Kalman filtering algorithm.
37,270
equivalence also The is the example. confirmed numerical via
The equivalence is also confirmed via the numerical example.
37,271
fully tutorials. automatic method video This game for paper generating introduces a
This paper introduces a fully automatic method for generating video game tutorials.
37,272
of wound doctors, significant Postoperative patients. for cause a expense are hospitals, complications and
Postoperative wound complications are a significant cause of expense for hospitals, doctors, and patients.
37,273
the method an wound desired. of onset to diagnose is Hence, effective strongly complications
Hence, an effective method to diagnose the onset of wound complications is strongly desired.
37,274
accessible due care to deliver means to ubiquity. wound provide a their Smartphones increasing
Smartphones provide a means to deliver accessible wound care due to their increasing ubiquity.
37,275
extremely of Networks Modern tasks. Neural powerful Convolutional computer on (CNN) vision range are a
Modern Convolutional Neural Networks (CNN) are extremely powerful on a range of computer vision tasks.
37,276
algorithms by proposed measurement numerical and data. by of the evaluated simulations performance is The
The performance of the proposed algorithms is evaluated by numerical simulations and by measurement data.
37,277
We main result two this develop in steps.
We develop this result in two main steps.
37,278
associated The that uncovers first the be fact classes to specific image-space with directions. tend
The first uncovers the fact that classes tend to be associated with specific image-space directions.
37,279
This perspective on provides of existence the universal a adversarial perturbations. novel
This provides a novel perspective on the existence of universal adversarial perturbations.
37,280
contradiction apparent and our between resolves analysis accuracy vulnerability. Thus, the
Thus, our analysis resolves the apparent contradiction between accuracy and vulnerability.
37,281
an is captions captioning. in for emerging image topic image an stylized Generating
Generating stylized captions for an image is an emerging topic in image captioning.
37,282
topic. image weak a medical unexplored remains with relatively network annotations a segmentation Training
Training a medical image segmentation network with weak annotations remains a relatively unexplored topic.
37,283
We the techniques public segmentation evaluate (NCI-ISBI) and (ACDC) prostate datasets. cardiac on
We evaluate the techniques on public cardiac (ACDC) and prostate (NCI-ISBI) segmentation datasets.
37,284
and tested publicly is proposed The datasets. on two used available method widely
The proposed method is tested on two publicly available and widely used datasets.
37,285
jointly and pose MultiPoseNet problems. segmentation can person detection, keypoint handle estimation person detection,
MultiPoseNet can jointly handle person detection, keypoint detection, person segmentation and pose estimation problems.
37,286
factors that highly behavior. impacted are One interference important the of is
One of the important factors that are highly impacted is interference behavior.
37,287
simulations. Monte-Carlo Analytical results validated are via
Analytical results are validated via Monte-Carlo simulations.
37,288
in the of the we setting study problem Calculus. the Here, relational Situation
Here, we study the problem in the relational setting of the Situation Calculus.
37,289
techniques We the actually provide synthesizing also controller. for
We also provide techniques for actually synthesizing the controller.
37,290
version, capsule to generative This variational paper (VCs). a extends the named capsules
This paper extends the capsule to a generative version, named variational capsules (VCs).
37,291
model as a a probabilistic Variational model. in entities composition of capsules an image
Variational capsules model an image as a composition of entities in a probabilistic model.
37,292
multiple a encode in variational addition, capsules way. semantically-disentangling In entities
In addition, variational capsules encode multiple entities in a semantically-disentangling way.
37,293
semantic not enforce This is because, invertibility correctness. does necessarily
This is because, invertibility does not necessarily enforce semantic correctness.
37,294
(RNNs) neural networks of are of sequential capable temporal information. dynamics complex modeling the Recurrent
Recurrent neural networks (RNNs) are capable of modeling the temporal dynamics of complex sequential information.
37,295
block an to with We EleAttG an RNN block. an equipped EleAtt-RNN as refer
We refer to an RNN block equipped with an EleAttG as an EleAtt-RNN block.
37,296
The uniqueness algorithm the existence and proposed and are of the solution convergence analysed. of
The existence and uniqueness of the solution and convergence of the proposed algorithm are analysed.
37,297
work includes main The following aspects. contribution of this three the
The main contribution of this work includes the following three aspects.
37,298
the DSurVD. pedestrian by evaluated detection seven robustness the is built of algorithms existing Third,
Third, the robustness of seven existing pedestrian detection algorithms is evaluated by the built DSurVD.
37,299