Learning Transformation Invariance from Global to Local

Jens Hocke, Thomas Martinetz

Abstract

Learning representations invariant to image transformations is fundamental to improving object recognition. We explore the connections between i-theory, Toroidal Subspace Analysis and slow subspace learning. All these methods can only achieve invariance to one transformation. Motivated by this limitation of these global methods we adapt the slow subspace approach to a local convolutional setting. Experimentally we show invariance to multiple transformations, and test object recognition performance.
OriginalspracheEnglisch
TitelWorkshop New Challenges in Neural Computation 2015
Redakteure/-innenBarbara Hammer, Thomas Villmann
Seitenumfang9
Band03
Erscheinungsdatum2015
Seiten16-24
PublikationsstatusVeröffentlicht - 2015
Veranstaltung37th German Conference on Pattern Recognition - Workshop Program: New Challenges in Neural Computation (NC2) - RWTH Aachen, Aachen, Deutschland
Dauer: 10.10.201510.10.2015
http://vmv2015.rwth-aachen.de/workshops.html

Fingerprint

Untersuchen Sie die Forschungsthemen von „Learning Transformation Invariance from Global to Local“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren