CaltechAUTHORS
  A Caltech Library Service

Learning viewpoint invariant object representations using a temporal coherence principle

Einhäuser, Wolfgang and Hipp, Joerg and Eggert, Julian and Körner, Edgar and König, Peter (2005) Learning viewpoint invariant object representations using a temporal coherence principle. Biological Cybernetics, 93 (1). pp. 79-90. ISSN 0340-1200. http://resolver.caltech.edu/CaltechAUTHORS:20130816-103140484

[img]
Preview
PDF - Published Version
See Usage Policy.

549Kb

Use this Persistent URL to link to this item: http://resolver.caltech.edu/CaltechAUTHORS:20130816-103140484

Abstract

Invariant object recognition is arguably one of the major challenges for contemporary machine vision systems. In contrast, the mammalian visual system performs this task virtually effortlessly. How can we exploit our knowledge on the biological system to improve artificial systems? Our understanding of the mammalian early visual system has been augmented by the discovery that general coding principles could explain many aspects of neuronal response properties. How can such schemes be transferred to system level performance? In the present study we train cells on a particular variant of the general principle of temporal coherence, the “stability” objective. These cells are trained on unlabeled real-world images without a teaching signal. We show that after training, the cells form a representation that is largely independent of the viewpoint from which the stimulus is looked at. This finding includes generalization to previously unseen viewpoints. The achieved representation is better suited for view-point invariant object classification than the cells’ input patterns. This property to facilitate view-point invariant classification is maintained even if training and classification take place in the presence of an – also unlabeled – distractor object. In summary, here we show that unsupervised learning using a general coding principle facilitates the classification of real-world objects, that are not segmented from the background and undergo complex, non-isomorphic, transformations.


Item Type:Article
Related URLs:
URLURL TypeDescription
http:/dx.doi.org/ 10.1007/s00422-005-0585-8DOIArticle
http://link.springer.com/article/10.1007%2Fs00422-005-0585-8PublisherArticle
Additional Information:Received: 19 February 2004 / Accepted: 23 May 2005 / Published online: 13 July 2005 © Springer-Verlag 2005. This work was supported by the EU-AMOUSE (JH) project and the Swiss National Science Foundation (PK, grant-no. 31-61415.01). We are grateful to K.P. Kording for making his MATLAB code for optimizing the stability objective available.
Group:Koch Laboratory, KLAB
Funders:
Funding AgencyGrant Number
EU-AMOUSEUNSPECIFIED
Swiss National Science Foundation31-61415.01
Record Number:CaltechAUTHORS:20130816-103140484
Persistent URL:http://resolver.caltech.edu/CaltechAUTHORS:20130816-103140484
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:40365
Collection:CaltechAUTHORS
Deposited By: KLAB Import
Deposited On:12 Jan 2008 00:54
Last Modified:02 Dec 2013 23:52

Repository Staff Only: item control page