text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
an increasing new Thus, developed recent number have in years. of been algorithms
Thus, an increasing number of new algorithms have been developed in recent years.
39,500
of requirements describes platform. structure, this paper the building and This steps,
This paper describes the steps, structure, and requirements of building this platform.
39,501
this understand to help is study Case platform. presented
Case study is presented to help understand this platform.
39,502
real data often losses. not while Synthetic unsupervised heuristic to videos, generalize require methods does
Synthetic data often does not generalize to real videos, while unsupervised methods require heuristic losses.
39,503
as for interpolation optical a task flow. investigate frame we Here, proxy
Here, we investigate frame interpolation as a proxy task for optical flow.
39,504
a we train Using interpolation. real for unsupervised movies, CNN temporal
Using real movies, we train a CNN unsupervised for temporal interpolation.
39,505
but Such cannot a implicitly handle untextured network estimates regions. motion,
Such a network implicitly estimates motion, but cannot handle untextured regions.
39,506
starting complex a to point model interactions. models are suitable social Multi-agent
Multi-agent models are a suitable starting point to model complex social interactions.
39,507
practices agents in practical We argue that standard facilitate interactions. social of social the reasoning
We argue that social practices facilitate the practical reasoning of agents in standard social interactions.
39,508
available many only form. In cases, pixel is however, in data
In many cases, however, data is only available in pixel form.
39,509
a domain high-level a method generating We (e.g. directly in images for present
We present a method for generating images directly in a high-level domain (e.g.
39,510
for pairwise data. the real strokes), need brush without
brush strokes), without the need for real pairwise data.
39,511
the of embeddings knowledge increases. gradually the RNKN exhibits number as training of epochs interpretation
RNKN gradually exhibits the interpretation of knowledge embeddings as the number of training epochs increases.
39,512
network multilevel (PMA). corresponding pose-guided contents, a attention To propose visual multi-granularity we the exploit
To exploit the multilevel corresponding visual contents, we propose a pose-guided multi-granularity attention network (PMA).
39,513
fitting is One of studied of the the main statistics models. in problems
One of the main problems studied in statistics is the fitting of models.
39,514
with explain would Ideally, like we as parameters a few dataset large as possible. to
Ideally, we would like to explain a large dataset with as few parameters as possible.
39,515
this automatizing been have numerous process. at There attempts
There have been numerous attempts at automatizing this process.
39,516
angle algorithm, the holonomic HELARS. least We call this extended or regression
We call this the holonomic extended least angle regression algorithm, or HELARS.
39,517
are for SINR both compared resulting The experiments and interference discussed.
The resulting SINR for both interference experiments are compared and discussed.
39,518
and HCI the these to MAHNOB DEAP applied schemes thus databases. We
We thus applied these schemes to the HCI MAHNOB and DEAP databases.
39,519
we Head model. paper, an Detector, this trainable propose Convolutional head detection end-to-end In FCHD-Fully
In this paper, we propose FCHD-Fully Convolutional Head Detector, an end-to-end trainable head detection model.
39,520
This makes memory our low lightweight with and inference model requirements. time
This makes our model lightweight with low inference time and memory requirements.
39,521
are bounding when labeling that still the ambiguities observe introduced However, we boxes.
However, we observe that ambiguities are still introduced when labeling the bounding boxes.
39,522
architectures the loss with improves nearly additional no various of greatly computation. Our accuracies localization
Our loss greatly improves the localization accuracies of various architectures with nearly no additional computation.
39,523
github.com/yihui-he/KL-Loss code at: Our and are available models .
Our code and models are available at: github.com/yihui-he/KL-Loss .
39,524
model the several once fine-tuning. without tasks to This allows explicit solve at
This allows the model to solve several tasks at once without explicit fine-tuning.
39,525
designed approaches, our e.g. When model comparing to are for specifically state-of-the-art which
When comparing our model to state-of-the-art approaches, which are specifically designed for e.g.
39,526
augment analyses data. services by can qualitative with gerontologists Such quantitative done
Such services can augment qualitative analyses done by gerontologists with quantitative data.
39,527
The system as a Arianna+. mechanism home implemented referred reasoning has to in been smart
The reasoning mechanism has been implemented in a smart home system referred to as Arianna+.
39,528
Results processes. indicate and scalability human recognition for activity robustness
Results indicate scalability and robustness for human activity recognition processes.
39,529
and important number simple allows analytical a restriction a yet obtaining Such results. of
Such a restriction allows obtaining a number of simple and yet important analytical results.
39,530
findings variance illustrate by in-depth evolution. our We of analysis the
We illustrate our findings by in-depth analysis of the variance evolution.
39,531
This understand aims problem how two to (sisters, are brothers, and if people etc.) family-related
This problem aims to understand if two people are family-related and how (sisters, brothers, etc.)
39,532
Solving applications. give to can varied problem this and rise tasks
Solving this problem can give rise to varied tasks and applications.
39,533
problem. the article, we this Kinship-Verification solving Deep propose approach a using Learning for In
In this article, we propose using a Deep Learning approach for solving the Kinship-Verification problem.
39,534
mask harmonized multi-site with imaging Even can advantages of studies. these the sequences, site-dependent variability
Even with harmonized imaging sequences, site-dependent variability can mask the advantages of these multi-site studies.
39,535
edges increase notable for observed. in A was and ICC functional networks values resting-state also
A notable increase in ICC values for functional edges and resting-state networks was also observed.
39,536
in to regression. found identifiability were affected Improvements not by be signal global
Improvements in identifiability were not found to be affected by global signal regression.
39,537
include only the metabolic information models current gene However, level. at
However, current metabolic models only include information at the gene level.
39,538
in context license our method the We of plates. automotive demonstrate
We demonstrate our method in the context of automotive license plates.
39,539
Estimation of a facial shapes central face plays for animation. transfer and role
Estimation of facial shapes plays a central role for face transfer and animation.
39,540
very tissue this therefore, is, segmentation Brain at challenging. age
Brain tissue segmentation at this age is, therefore, very challenging.
39,541
dataset code be source Our open-sourced. will synthetic and
Our source code and synthetic dataset will be open-sourced.
39,542
significant SR. in The accurate of a role correspondence plays generation video
The generation of accurate correspondence plays a significant role in video SR.
39,543
deep in generation. are methods flows optical correspondence for LR However, existing based used learning
However, LR optical flows are used in existing deep learning based methods for correspondence generation.
39,544
flows. the Then, optical motion is to performed compensation HR according
Then, motion compensation is performed according to the HR optical flows.
39,545
real perform images. numerical we Finally, and synthetic experiments the on both
Finally, we perform the numerical experiments on both synthetic and real images.
39,546
The vessel retinal on out quantitive is carried images. validation
The quantitive validation is carried out on retinal vessel images.
39,547
that a the proposed promising The results performance. indeed indicate achieves model
The results indicate that the proposed model indeed achieves a promising performance.
39,548
the task. in which vital We plays role this pooling a investigate method global
We investigate the global pooling method which plays a vital role in this task.
39,549
the ILSVRC and the on We dataset. evaluate localization ability pixel-level classification benchmark
We evaluate the classification and pixel-level localization ability on the ILSVRC benchmark dataset.
39,550
We that multi-task with GWAP a our module combines propose class-specific R-FCN. framework
We propose a multi-task framework that combines our class-specific GWAP module with R-FCN.
39,551
with framework truth image-level and bounding The labels. ground few trained boxes is large-scale
The framework is trained with few ground truth bounding boxes and large-scale image-level labels.
39,552
We dataset. on this PASCAL evaluate VOC framework
We evaluate this framework on PASCAL VOC dataset.
39,553
accuracy model is of The hierarchical through the ensures which detection. obtained the training,
The model is obtained through hierarchical training, which ensures the accuracy of the detection.
39,554
advances recent on An learning. capitalize in alternative approach deep to would be
An alternative approach would be to capitalize on recent advances in deep learning.
39,555
vision. estimation an computer in challenging yet pose task essential is Human
Human pose estimation is an essential yet challenging task in computer vision.
39,556
architecture model. convolutional network with work, this In attention the a novel we combined proposed
In this work, we proposed a convolutional network architecture combined with the novel attention model.
39,557
attention (ACNN). We network neural convolutional named it
We named it attention convolutional neural network (ACNN).
39,558
ACNN to focus regions features. different on of input learns specific
ACNN learns to focus on specific regions of different input features.
39,559
as background the "nothing-regions", body filtrate Early and parts. such stages redundant
Early stages filtrate the "nothing-regions", such as background and redundant body parts.
39,560
State-of-the-art obtained the is MPII on performance benchmarks.
State-of-the-art performance is obtained on the MPII benchmarks.
39,561
to model's Adversarial aim images by a a perturbing decision minimally target image. change target
Adversarial images aim to change a target model's decision by minimally perturbing a target image.
39,562
task. person we scheme attention improve paper, present this mechanism In an re-identification to
In this paper, we present an attention mechanism scheme to improve person re-identification task.
39,563
measurement, are oscilloscope digital improving and discussed. impairments system Device resolution
Device measurement, digital oscilloscope impairments and improving system resolution are discussed.
39,564
recognition text for paired supervisory a using without any This work data. visual method presents
This work presents a method for visual text recognition without using any paired supervisory data.
39,565
two-stage novel intra-mode. in approach a decision propose HEVC partition Therefore, CU we
Therefore, we propose a novel two-stage CU partition decision approach in HEVC intra-mode.
39,566
Most effects apply image. existing the to the whole methods artistic of
Most of the existing methods apply artistic effects to the whole image.
39,567
is filters. the to allow enough new method flexible of our Also, addition
Also, our method is flexible enough to allow the addition of new filters.
39,568
We segmentation pre-trained segmentation dataset R-CNN instance on the network. COCO as use the Mask
We use pre-trained Mask R-CNN instance segmentation on the COCO dataset as the segmentation network.
39,569
filters performed. use the results different on of are Experimental
Experimental results on the use of different filters are performed.
39,570
of amount target perform propose a to limited we data. Then, with adaptation
Then, we propose to perform adaptation with a limited amount of target data.
39,571
three by observations. The is framework motivated proposed
The proposed framework is motivated by three observations.
39,572
between a Finally, evaluate association posteriori mode connectivity longitudinal outcomes. the we weights with neurocognitive
Finally, we evaluate a posteriori the association between connectivity mode weights with longitudinal neurocognitive outcomes.
39,573
engineer inspections authorities municipal volunteers. structural These gfare by typically through conducted
These inspections gfare typically conducted by municipal authorities through structural engineer volunteers.
39,574
limited spatial However, phosphenic very of images terms resolution (e.g. are in
However, phosphenic images are very limited in terms of spatial resolution (e.g.
39,575
this of recognition under the indoor environments we study In vision. prosthetic work, simulated
In this work, we study the recognition of indoor environments under simulated prosthetic vision.
39,576
can become easily. more sensors depth prevalent Meanwhile, be so maps depth acquired nowadays,
Meanwhile, depth sensors become prevalent nowadays, so depth maps can be acquired more easily.
39,577
semantic studies are the segmentation task. few focus on there RGB-D However, that
However, there are few studies that focus on the RGB-D semantic segmentation task.
39,578
to challenge. effectiveness depth information the a improve Exploiting is performance
Exploiting the depth information effectiveness to improve performance is a challenge.
39,579
concept multiple modalities. sensor this We to data from extend incorporate Stixel
We extend this Stixel concept to incorporate data from multiple sensor modalities.
39,580
variograms on signals extensions of Graph variograms are defined continuous for Euclidean space. used
Graph variograms are extensions of variograms used for signals defined on continuous Euclidean space.
39,581
datasets However, to large are acquire. costly such labeled
However, such large labeled datasets are costly to acquire.
39,582
scratch. SNNs from addresses the neuromorphic learning for temporal tutorial This
This tutorial addresses neuromorphic SNNs for temporal learning from the scratch.
39,583
large remains in significant challenge a Detecting over objects areas satellite imagery analytics. small
Detecting small objects over large areas remains a significant challenge in satellite imagery analytics.
39,584
scales For (e.g. very different objects of
For objects of very different scales (e.g.
39,585
this In a proposed. enhancement algorithm is low-light paper, image gradient-based
In this paper, a gradient-based low-light image enhancement algorithm is proposed.
39,586
addition, constraints In intensity-range the the for image we integration. involve
In addition, we involve the intensity-range constraints for the image integration.
39,587
Experiments effectively proposed the demonstrate image enhancement images. low-light enhance that gradient-based can low-light the
Experiments demonstrate that the proposed gradient-based low-light image enhancement can effectively enhance the low-light images.
39,588
In we of detectors scratch. object learning principles for DSOD, set contribute a design from
In DSOD, we contribute a set of design principles for learning object detectors from scratch.
39,589
than models. better consistently more much compact with DSOD methods achieves state-of-the-art results the
DSOD achieves consistently better results than the state-of-the-art methods with much more compact models.
39,590
the analytical also verify an based robustness our We of individual model develop to prediction.
We also develop an individual based model to verify the robustness of our analytical prediction.
39,591
coevolutionary population cause unexpected However, prey patterns. small selection can sizes, for
However, for small population sizes, prey selection can cause unexpected coevolutionary patterns.
39,592
Creativity of value, often in interestingness. and novelty, terms formulated is surprise
Creativity is often formulated in terms of novelty, value, surprise and interestingness.
39,593
Experiments benefits of the network. proposed benefit show the the gLGCN
Experiments show the benefit of the benefits the proposed gLGCN network.
39,594
Play-style learning. video introduce gameplay via problem Let's commentary the We machine of of generating
We introduce the problem of generating Let's Play-style commentary of gameplay video via machine learning.
39,595
future commentary Play demonstrate promising generation. We towards Let's results
We demonstrate promising results towards future Let's Play commentary generation.
39,596
pair of It removesthe ofdual-lens offline calibration every need individual for cameras.
It removesthe need of individual offline calibration for every pair ofdual-lens cameras.
39,597
markers the introduced fiducial are to improve precision. overall Visual
Visual fiducial markers are introduced to improve the overall precision.
39,598
segmentation semantic supervision prior detection. objectness an object and Class-agnostic knowledge both provides for segmentation
Class-agnostic segmentation supervision provides an objectness prior knowledge for both semantic segmentation and object detection.
39,599