Presentation 1997/3/17
Learning and Recognition of Translated, Rotated and Scaled Patterns by Multilayer Nets
Hiromu GOTANDA, Kousaku KAWAI, Tatsuya YAMAOKA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This report describes a scheme for learning and recognizing translated, rotated and scaled patterns by multilayer nets to which inputs are given by meshed patterns (feature vectors) extracted using well-known geometrical characteristics (center of gravity, angle of principal axis, variance). It also describes detection of the rotated angle and the scaling factor of target patterns to their template based on the characteristics. Experiments show that small nets can learn the patterns faster and classify them at high recognition rate, and that the rotated angle and the scaling factor can be detected at high accuracy.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) multilayer nets / BP learning / pattern recognition / translation / rotation / scaling
Paper # NC96-148
Date of Issue

Conference Information
Committee NC
Conference Date 1997/3/17(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Learning and Recognition of Translated, Rotated and Scaled Patterns by Multilayer Nets
Sub Title (in English)
Keyword(1) multilayer nets
Keyword(2) BP learning
Keyword(3) pattern recognition
Keyword(4) translation
Keyword(5) rotation
Keyword(6) scaling
1st Author's Name Hiromu GOTANDA
1st Author's Affiliation Kinki University in Kyushu, Faculty of Engineering()
2nd Author's Name Kousaku KAWAI
2nd Author's Affiliation Kinki University, Graduate School of Advanced Technology
3rd Author's Name Tatsuya YAMAOKA
3rd Author's Affiliation Kinki University, Graduate School of Advanced Technology
Date 1997/3/17
Paper # NC96-148
Volume (vol) vol.96
Number (no) 583
Page pp.pp.-
#Pages 8
Date of Issue