text
stringlengths
27
153
label
stringlengths
27
153
id
int64
0
40k
the some stationary be distribution cases, effect on substantial. the In can
In some cases, the effect on the stationary distribution can be substantial.
13,600
several even scalable This instances. to large with method our million makes training datasets
This makes our method scalable even to large datasets with several million training instances.
13,601
Neural image classification (CNN) state-of-the-art Networks models Convolutional many tasks. for are
Convolutional Neural Networks (CNN) are state-of-the-art models for many image classification tasks.
13,602
agreement our accuracy similar between the is pathologists. method inter-observer classification to of The
The classification accuracy of our method is similar to the inter-observer agreement between pathologists.
13,603
data task. challenging mining is a mining pattern under constraints Sequential
Sequential pattern mining under constraints is a challenging data mining task.
13,604
to what This interpretable our is us and object representation easily visualize allows detector "sees".
This representation is easily interpretable and allows us to visualize what our object detector "sees".
13,605
correspondences estimation between The of resp. two images
The estimation of correspondences between two images resp.
13,606
core point problem in vision. sets a computer is
point sets is a core problem in computer vision.
13,607
so to second have this been methods Several proposed solve called problem. order
Several so called second order methods have been proposed to solve this problem.
13,608
no guarantees order theoretical are known. third algorithms these most of For
For most of these third order algorithms no theoretical guarantees are known.
13,609
cameras. behaviors of using front are work, individuals' In web in this computers studied
In this work, individuals' behaviors in front of computers are studied using web cameras.
13,610
for data in non-Gaussian. of the problem analysis that are general We factor discriminative consider
We consider the problem of discriminative factor analysis for data that are in general non-Gaussian.
13,611
the proposed. model based data the Bayesian of is on A ranks
A Bayesian model based on the ranks of the data is proposed.
13,612
rank-likelihood. We first {\em introduce the of new max-margin} a version
We first introduce a new {\em max-margin} version of the rank-likelihood.
13,613
developed robust technique. we work, In and have this departure a detection lane warning
In this work, we have developed a robust lane detection and departure warning technique.
13,614
on single sensor. system camera based Our is
Our system is based on single camera sensor.
13,615
using fitted to curves cubic and RANSAC. robust assumed are be Lanes
Lanes are assumed to be cubic curves and fitted using robust RANSAC.
13,616
the and method This can detect lanes road boundary. its reliably of
This method can reliably detect lanes of the road and its boundary.
13,617
flow lane angle optical an For departure based used. method were
For lane departure angle an optical flow based method were used.
13,618
for probabilistic this we spoken systems. a framework In paper, goal-driven present dialog
In this paper, we present a probabilistic framework for goal-driven spoken dialog systems.
13,619
that also We DM efficient are DM. based more demonstrate than entropy-based non-entropy strategies
We also demonstrate that entropy-based DM strategies are more efficient than non-entropy based DM.
13,620
powerful limited from Deep learning and flexibility. suffer expensive tools but methods computation often are
Deep learning methods are powerful tools but often suffer from expensive computation and limited flexibility.
13,621
deep alternative is An models representations. combine with light-weight to
An alternative is to combine light-weight models with deep representations.
13,622
As absent. in problems, visual several exist unified cases is framework a successful
As successful cases exist in several visual problems, a unified framework is absent.
13,623
boosting pre-trained transfers feed CCF CNN features to from forest low-level models model. the
CCF transfers low-level features from pre-trained CNN models to feed the boosting forest model.
13,624
increasingly is popular, very is egocentric browsing video While it becoming difficult.
While egocentric video is becoming increasingly popular, browsing it is very difficult.
13,625
segment Recognizing (index) and to egocentric activities us unstructured temporally videos. long-term long enables
Recognizing long-term activities enables us to temporally segment (index) long and unstructured egocentric videos.
13,626
input, flow a sparse camera optical classifies activity. Given as wearer's the volume CNN our
Given a sparse optical flow volume as input, our CNN classifies the camera wearer's activity.
13,627
widely cancer in Bladder the world. spread is
Bladder cancer is widely spread in the world.
13,628
exploration remains clinical surface. internal for of procedure bladder Video-endoscopy visual the standard the
Video-endoscopy remains the standard clinical procedure for visual exploration of the bladder internal surface.
13,629
And, lesions over several typically, images. spread are,
And, lesions are, typically, spread over several images.
13,630
problems. addresses classes different, optimization two of paper yet This interrelated
This paper addresses two classes of different, yet interrelated optimization problems.
13,631
We consider (aka existential for specifying formalism ontologies. Datalog+) as rules a
We consider existential rules (aka Datalog+) as a formalism for specifying ontologies.
13,632
inconsistent imply Taken tree together, under inference mild results these conditions.
Taken together, these results imply inconsistent tree inference under mild conditions.
13,633
convergence world significant and Our demonstrates sets. on data algorithm synthetic improvements real
Our algorithm demonstrates significant convergence improvements on synthetic and real world data sets.
13,634
demonstrate neural ensembles that of the modeled. can activity be quantitatively we Here
Here we demonstrate that the activity of neural ensembles can be quantitatively modeled.
13,635
this bias in computer been ignored mostly However, has vision. volumetric
However, this volumetric bias has been mostly ignored in computer vision.
13,636
significant We segmentation to this methods due artifacts demonstrate reconstruction and in standard bias.
We demonstrate significant artifacts in standard segmentation and reconstruction methods due to this bias.
13,637
of centers dynamics Epidemic is stochastic a interacting network considered. epidemic in
Epidemic dynamics in a stochastic network of interacting epidemic centers is considered.
13,638
The migration epidemic modelled by processes and are chains. Markov's
The epidemic and migration processes are modelled by Markov's chains.
13,639
probability the derived. Explicit formulas migration process for are of distribution
Explicit formulas for probability distribution of the migration process are derived.
13,640
and numerically. Dependence parameters population, of outbreak parameters coupling analytically initial is parameters, on examined
Dependence of outbreak parameters on initial parameters, population, coupling parameters is examined analytically and numerically.
13,641
The for process general approximation a derived. is mean migration field
The mean field approximation for a general migration process is derived.
13,642
in from are We cues. interested geometric automatic scene understanding
We are interested in automatic scene understanding from geometric cues.
13,643
real-time semantic we the reconstruction. of aim To loop segmentation to this end, bring in
To this end, we aim to bring semantic segmentation in the loop of real-time reconstruction.
13,644
world segment any real without scenes Importantly, to noise our network is modelling. able
Importantly, our network is able to segment real world scenes without any noise modelling.
13,645
corner presents a descriptors. paper and novel shape This scale multi gradient based point a
This paper presents a novel multi scale gradient and a corner point based shape descriptors.
13,646
large for method runtime The it the images. quasilinear suitable makes of segmenting
The quasilinear runtime of the method makes it suitable for segmenting large images.
13,647
are medicine. of frequently in new special of today's Discoveries importance biomarkers for occurring diseases
Discoveries of new biomarkers for frequently occurring diseases are of special importance in today's medicine.
13,648
interaction. of social strategy a Ethnocentrism behavioral scale seen on every is
Ethnocentrism is a behavioral strategy seen on every scale of social interaction.
13,649
selects models boosts demonstrate it which increases that Game-theory ethnocentrism reproductive evolution because cooperation, fitness.
Game-theory models demonstrate that evolution selects ethnocentrism because it boosts cooperation, which increases reproductive fitness.
13,650
simulations we test this hypothesis. Here, use agent-based to computer
Here, we use agent-based computer simulations to test this hypothesis.
13,651
end-to-end This optimization. as lends opportunities as advanced introspection to for well
This lends to advanced introspection as well as opportunities for end-to-end optimization.
13,652
HOG state-of-the-art applications improve Both on approaches. respective the
Both applications improve on the respective state-of-the-art HOG approaches.
13,653
good necessary learning supervision representation? strong for visual a Is
Is strong supervision necessary for learning a good visual representation?
13,654
Neural images millions a Convolutional semantically-labeled Do to train need of (CNN)? really Network we
Do we really need millions of semantically-labeled images to train a Convolutional Neural Network (CNN)?
13,655
visual Our provides the key idea supervision. that tracking is
Our key idea is that visual tracking provides the supervision.
13,656
a function Siamese-triplet network to representation. this We a design CNN loss train with ranking
We design a Siamese-triplet network with a ranking loss function to train this CNN representation.
13,657
toward the surveys detection. This current review made anomaly article video-based progresses
This review article surveys the current progresses made toward video-based anomaly detection.
13,658
aspect fundamental for that address anomaly most is, feature video video representation. the We detection,
We address the most fundamental aspect for video anomaly detection, that is, video feature representation.
13,659
obtained Large data of Internet. quantities can the easily video from unlabeled be
Large quantities of unlabeled video data can be easily obtained from the Internet.
13,660
trainable solution document a images. this binarization fully present In for paper degraded we
In this paper we present a fully trainable binarization solution for degraded document images.
13,661
out-of-domain generalization data. good capabilities shows the trained binarization Additionally, classifier on document
Additionally, the trained document binarization classifier shows good generalization capabilities on out-of-domain data.
13,662
it direction will a Given move, scene, move? in is to and what what going
Given a scene, what is going to move, and in what direction will it move?
13,663
could non-semantic action form a be considered question Such a prediction. of
Such a question could be considered a non-semantic form of action prediction.
13,664
this convolutional network approach present prediction. we (CNN) for motion based a In neural work,
In this work, we present a convolutional neural network (CNN) based approach for motion prediction.
13,665
large previous approaches all margins. outperform We by
We outperform all previous approaches by large margins.
13,666
is a Particle Swarm (PSO) problems. solving optimization continuous Optimization nature-inspired meta-heuristic for
Particle Swarm Optimization (PSO) is a nature-inspired meta-heuristic for solving continuous optimization problems.
13,667
purpose, the is dimension mathematically. in each analyzed For particles' potential this
For this purpose, the particles' potential in each dimension is analyzed mathematically.
13,668
the reasonable particles' made. some assumptions if Additionally, are behavior on potential the
Additionally, some reasonable assumptions on the behavior if the particles' potential are made.
13,669
KITTI on We model also for estimation dataset. the car train distance a
We also train a model for car distance estimation on the KITTI dataset.
13,670
real can approach to direct Results well our perception that show generalize driving images.
Results show that our direct perception approach can generalize well to real driving images.
13,671
on website. project available data and Source code are our
Source code and data are available on our project website.
13,672
regularity. presents and examples intriguing processes proceed high multiple precision at of which Nature
Nature presents multiple intriguing examples of processes which proceed at high precision and regularity.
13,673
activity among (ST). are precancerous of studied The tissues S-transform through and evolution development
The evolution and development of precancerous activity among tissues are studied through S-transform (ST).
13,674
are Additionally, particularly environmental non-bilaterian high not levels. bodyplans expected oxygen require to
Additionally, non-bilaterian bodyplans are not expected to require particularly high environmental oxygen levels.
13,675
the panicea. with is demonstrating consistent of low-oxygen Halichondria sponge experimental evidence tolerance the This
This is consistent with experimental evidence demonstrating the low-oxygen tolerance of the sponge Halichondria panicea.
13,676
analyze We positions. of choosing strategies propose these several and
We propose and analyze several strategies of choosing these positions.
13,677
This plant for video paper proposes novel framework fluorescence processing. a
This paper proposes a novel framework for fluorescence plant video processing.
13,678
within plant interested research The photosynthetic in analysis community plant. the is a leaf-level
The plant research community is interested in the leaf-level photosynthetic analysis within a plant.
13,679
as We tracking identify alignment, segmentation, this problem. and multi-leaf a joint
We identify this as a joint multi-leaf segmentation, alignment, and tracking problem.
13,680
of also The limitation algorithm studied. is our
The limitation of our algorithm is also studied.
13,681
and show of results the robustness effectiveness, Experimental method. proposed the efficiency,
Experimental results show the effectiveness, efficiency, and robustness of the proposed method.
13,682
great from still Event image images for understanding. is importance recognition of
Event recognition from still images is of great importance for image understanding.
13,683
called a new we Neural Object-Scene Specifically, Convolutional architecture, (OS-CNN). design Network
Specifically, we design a new architecture, called Object-Scene Convolutional Neural Network (OS-CNN).
13,684
Furthermore, are complementary to networks and very-deep other. find deep we that the each
Furthermore, we find that the deep and very-deep networks are complementary to each other.
13,685
to Potential Application: of development methodology include of reduce study carsickness. a applications this
Application: Potential applications of this study include development of a methodology to reduce carsickness.
13,686
significant It problem. potentially is yet under-researched a
It is a potentially significant yet under-researched problem.
13,687
dimension of spatial signal, Independently image etc. image, (one-dimensional images, volume of the
Independently of the image spatial dimension (one-dimensional signal, image, volume of images, etc.
13,688
can for problematic These learning established many requirements for to act. methods be
These requirements can be problematic for many established methods for learning to act.
13,689
time free However, settings. planning is most not in real-world
However, planning time is not free in most real-world settings.
13,690
Markov and for We the analyze (MDPs). problem Decision formalize Processes metareasoning
We formalize and analyze the metareasoning problem for Markov Decision Processes (MDPs).
13,691
approximations. discuss, motivating optimal we be general metareasoning impractical, to reasons turns out For
For reasons we discuss, optimal general metareasoning turns out to be impractical, motivating approximations.
13,692
two and cons approaches pros are elucidated. the The of
The pros and cons of the two approaches are elucidated.
13,693
can work view The and non-overlapping have layout. in the of independently cameras larger field
The cameras work independently and can have larger field of view in the non-overlapping layout.
13,694
with. dealt be a should scale ambiguity However, factor
However, a scale factor ambiguity should be dealt with.
13,695
the Kalman filter extended is For approaches, real-time recursive as a estimator. both used
For both approaches, the extended Kalman filter is used as a real-time recursive estimator.
13,696
with The are and studied alike. verified real experiments approaches synthetic
The approaches studied are verified with synthetic and real experiments alike.
13,697
enumerate provide notes. We a features few and its implementation novel
We enumerate its novel features and provide a few implementation notes.
13,698
weighted the also We a algorithm filtering. fast provide for
We also provide a fast algorithm for the weighted filtering.
13,699