A modular network scheme for unsupervised 3D object recognition


This paper presents an unsupervised learning scheme for recognizing 3D objects from their 2D projected images. The scheme consists of a mixture of nonlinear autoencoders which can compress various views of 3D objects into representations that indicate the view direction. We evaluate the performance of the proposed modular network scheme through simulations using 3D wire-frame objects and discuss its related issues on object representations in the primate visual cortex. ( 2000 Elsevier Science B.V. All rights reserved.

DOI: 10.1016/S0925-2312(99)00148-4

Extracted Key Phrases

6 Figures and Tables

Cite this paper

@article{Suzuki2000AMN, title={A modular network scheme for unsupervised 3D object recognition}, author={Satoshi Suzuki and Hiroshi Ando}, journal={Neurocomputing}, year={2000}, volume={31}, pages={15-28} }