Filters








8,460 Hits in 5.7 sec

Learning viewpoint invariant object representations using a temporal coherence principle

Wolfgang Einhäuser, Jörg Hipp, Julian Eggert, Edgar Körner, Peter König
2005 Biological cybernetics  
In the present study we train cells on a particular variant of the general principle of temporal coherence, the "stability" objective.  ...  In summary, here we show that unsupervised learning using a general coding principle facilitates the classification of realworld objects, that are not segmented from the background and undergo complex,  ...  Körding for making his MATLAB code for optimizing the stability objective available.  ... 
doi:10.1007/s00422-005-0585-8 pmid:16021516 fatcat:rffhjrxvwjg2zp3ajwvipmch6q

JPMAX: Learning to Recognize Moving Objects as a Model-fitting Problem

Suzanna Becker
1994 Neural Information Processing Systems  
When trained with this algorithm on image sequences of moving geometric objects a two-layer network can learn to perform accurate position-invariant object classification.  ...  So far, however, they have had limited success in learning higher-order representations, e.g., of objects in visual images.  ...  viewpoint-specific whole objects and in the top layer viewpoint-invariance, in principle, could be achieved.  ... 
dblp:conf/nips/Becker94 fatcat:v7ljp2n265afbpgb7trufdrnie

Learning to Categorize Objects Using Temporal Coherence

Suzanna Becker
1992 Neural Information Processing Systems  
The invariance of an objects' identity as it transformed over time provides a powerful cue for perceptual learning.  ...  We present an unsupervised learning procedure which maximizes the mutual information between the representations adopted by a feed-forward network at consecutive time steps.  ...  TEMPORAL-COHERENCE BASED LEARNING One way to capture the constraint of temporal coherence in a learning procedure is to build it into the objective function.  ... 
dblp:conf/nips/Becker92 fatcat:v6ewjdgnsrg3db6f5b5hs334f4

Time-Contrastive Networks: Self-Supervised Learning from Video [article]

Pierre Sermanet, Corey Lynch, Yevgen Chebotar, Jasmine Hsu, Eric Jang, Stefan Schaal, Sergey Levine
2018 arXiv   pre-print
We train our representations using a metric learning loss, where multiple simultaneous viewpoints of the same observation are attracted in the embedding space, while being repelled from temporal neighbors  ...  Imitation of human behavior requires a viewpoint-invariant representation that captures the relationships between end-effectors (hands or robot grippers) and the environment, object attributes, and body  ...  A number of prior works use temporal coherence [28, 29, 30, 31] . Others also train for viewpoint invariance using metric learning [22, 32, 33] .  ... 
arXiv:1704.06888v3 fatcat:mqt2bdjvobc7lidrtvrc3rtnoi

Object-Centric Representation Learning from Unlabeled Videos [article]

Ruohan Gao, Dinesh Jayaraman, Kristen Grauman
2016 arXiv   pre-print
We introduce a novel object-centric approach to temporal coherence that encourages similar representations to be learned for object-like regions segmented from nearby frames.  ...  from which to learn invariances.  ...  Object-Centric Representation Learning from Unlabeled Videos  ... 
arXiv:1612.00500v1 fatcat:p4bnb4p2ubgvjkiizpicex4sti

Slow and Steady Feature Analysis: Higher Order Temporal Coherence in Video

Dinesh Jayaraman, Kristen Grauman
2016 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)  
How can unlabeled video augment visual learning? Existing methods perform "slow" feature analysis, encouraging the representations of temporally close frames to exhibit only small differences.  ...  Using five diverse datasets, including unlabeled YouTube and KITTI videos, we demonstrate our method's impact on object, scene, and action recognition tasks.  ...  This work was supported in part by ONR YIP N00014-15-1-2291. used throughout our results, which let us leverage standard CNN architectures known to work well with tiny images [1] .  ... 
doi:10.1109/cvpr.2016.418 dblp:conf/cvpr/JayaramanG16 fatcat:l2wmapkndba4jngntgsghgu6bm

What happens next and when "next" happens: Mechanisms of spatial and temporal prediction [article]

Dean Wyatte
2014 arXiv   pre-print
Synaptic weight scaling from prolonged learning built viewpoint invariance, but led to confusion between ambiguous views of objects, producing slightly lower performance on average.  ...  This counterintuitive pattern of results was accounted for by a neural network model that learned three-dimensional viewpoint invariance with LeabraTI's spatiotemporal prediction rule.  ...  viewpoint invariance learned from spatially and temporally predictable training), although this would need to be quantified via the representation similarity methods used in Chapter 5 (Kriegeskorte et  ... 
arXiv:1407.5328v1 fatcat:q6x7qbx225fh7jblaju2kyrzeq

Slow and steady feature analysis: higher order temporal coherence in video [article]

Dinesh Jayaraman, Kristen Grauman
2016 arXiv   pre-print
How can unlabeled video augment visual learning? Existing methods perform "slow" feature analysis, encouraging the representations of temporally close frames to exhibit only small differences.  ...  Using five diverse datasets, including unlabeled YouTube and KITTI videos, we demonstrate our method's impact on object, scene, and action recognition tasks.  ...  This work was supported in part by ONR YIP N00014-15-1-2291. used throughout our results, which let us leverage standard CNN architectures known to work well with tiny images [1] .  ... 
arXiv:1506.04714v2 fatcat:2mskrntojfcf3kyxdevggsykxm

SIMONe: View-Invariant, Temporally-Abstracted Object Representations via Unsupervised Video Decomposition [article]

Rishabh Kabra, Daniel Zoran, Goker Erdogan, Loic Matthey, Antonia Creswell, Matthew Botvinick, Alexander Lerchner, Christopher P. Burgess
2021 arXiv   pre-print
Leveraging the shared structure that exists across different scenes, our model learns to infer two sets of latent representations from RGB video input alone: a set of "object" latents, corresponding to  ...  the time-invariant, object-level contents of the scene, as well as a set of "frame" latents, corresponding to global time-varying elements such as viewpoint.  ...  It also uses a β-weighted KL loss to disentangle object representations.  ... 
arXiv:2106.03849v2 fatcat:vtgeigfcrbgj5jnwka2e2djgxe

View and Style-Independent Action Manifolds for Human Activity Recognition [chapter]

Michał Lewandowski, Dimitrios Makris, Jean-Christophe Nebel
2010 Lecture Notes in Computer Science  
Each action descriptor is produced, first, by applying Temporal Laplacian Eigenmaps to view-dependent videos in order to produce a stylistic invariant embedded manifold for each view separately.  ...  Then, all view-dependent manifolds are automatically combined to discover a unified representation which model in a single three dimensional space an action independently from style and viewpoint.  ...  Then, these models are combined to learn a single compact and view invariant generative model of the action using generative decomposable model [12] . Fig. 1 provides an overview of our method.  ... 
doi:10.1007/978-3-642-15567-3_40 fatcat:xrhzykgjdbhblgj7vfbyqmjvtu

Implicit Learning in 3D Object Recognition: The Importance of Temporal Context

Suzanna Becker
1999 Neural Computation  
The model learns a representation that categorizes people's faces according to identity, independent of viewpoint, by taking advantage of the temporal continuity in image sequences.  ...  In this case, the model learns a two-tiered representation, starting with a coarse view-based clustering and proceeding to a finer clustering of more specific stimulus features.  ...  All simulations were carried out using the Xerion neural network simulator developed in Hinton's lab and additional  ... 
doi:10.1162/089976699300016683 pmid:9950735 fatcat:xuurqzk7zfhmbnkqc672ovyrra

Visual working memory retains movement information within an allocentric reference frame

Justin N. Wood
2010 Visual Cognition  
Similarly, human adults readily learn locations in virtual reality environments defined entirely by a continuous colour gradient without individual objects that may be used as landmarks, in qualitative  ...  These four signatures characterize an object tracking 1 By innate I simply mean not learned.  ... 
doi:10.1080/13506285.2010.502430 fatcat:3fvoplmp2jb4hia755xrbymq4i

Learning Image Representations Tied to Ego-Motion

Dinesh Jayaraman, Kristen Grauman
2015 2015 IEEE International Conference on Computer Vision (ICCV)  
Understanding how images of objects and scenes behave in response to specific ego-motions is a crucial aspect of proper visual development, yet existing visual learning methods are conspicuously disconnected  ...  We propose to exploit proprioceptive motor signals to provide unsupervised regularization in convolutional neural networks to learn visual representations from egocentric video.  ...  Acknowledgements: This research is supported in part by ONR PECASE Award N00014-15-1-2291 and a gift from Intel.  ... 
doi:10.1109/iccv.2015.166 dblp:conf/iccv/JayaramanG15 fatcat:lt3mnzxdbzhftlnr4dzroglyae

Book reports

1993 Computers and Mathematics with Applications  
Viewpoint-invariant reconstruction of visible surface. 14. Invariant surface reconstructions using weak continuity cczustralnts. 15.  ...  Estimation of stereo and motion parameters using a variational principle. 7. On the reconstruction of a scene from two ~tered images. 8. A pipelined architecture for the Canny edge detector. 9.  ... 
doi:10.1016/0898-1221(93)90315-m fatcat:tnmkz3axfnekvgr4srsgsivuvy

Multiple levels of visual object constancy revealed by event-related fMRI of repetition priming

P. Vuilleumier, R. N. Henson, J. Driver, R. J. Dolan
2002 Nature Neuroscience  
We used event-related fMRI during a repetition-priming task to identify neural subsystems that process visual objects at distinct stages of representation, with particular interest in brain regions showing  ...  These data show that dissociable subsystems in ventral visual cortex maintain distinct view-dependent and view-invariant object representations.  ...  in a size-invariant manner.  ... 
doi:10.1038/nn839 pmid:11967545 fatcat:q5vbhaltdfb6hf4g6ukdruhy24
« Previous Showing results 1 — 15 out of 8,460 results