DocumentCode :
3688640
Title :
Scalable multi-neighborhood learning for convolutional networks
Author :
Elnaz Barshan;Paul Fieguth;Alexander Wong
Author_Institution :
Department of System Design Engineering, University of Waterloo, Waterloo, Canada
fYear :
2015
Firstpage :
1
Lastpage :
6
Abstract :
In this paper we explore the role of scale for improved feature learning in convolutional networks. We propose multi-neighborhood convolutional networks, designed to learn image features at different levels of detail. Utilizing nonlinear scale-space models, the proposed multi-neighborhood model can effectively capture fine-scale image characteristics (i.e., appearance) using a small-size neighborhood, while coarse-scale image structures (i.e., shape) are detected through a larger neighborhood. In addition, we introduce a scalable learning method for the proposed multi-neighborhood architecture and show how one can use an already-trained single-scale network to extract image features at multiple levels of detail. The experimental results demonstrate the superior performance of the proposed multi-scale multi-neighborhood models over their single-scale counterparts without an increase in training cost.
Keywords :
"Feature extraction","Computer architecture","Computational modeling","Image representation","Training","Convolution","Shape"
Publisher :
ieee
Conference_Titel :
Machine Learning for Signal Processing (MLSP), 2015 IEEE 25th International Workshop on
Type :
conf
DOI :
10.1109/MLSP.2015.7324361
Filename :
7324361
Link To Document :
بازگشت