ABM and CNN application in ventral stream of visual system
Bardia Yousefi, Paria Yousefi · 2015
This paper addresses an investigation regarding the suitability of two different techniques, Active Basis Model (ABM) and Gabor based Convolutional Neural Network (CNN or G-ConvNets) in the mechanism for recognition of biological movement (mammalian visual system model). This method inspired by ventral streams which provide the form information. Both of these approaches contain information of the shape of human object obtained by the Gabor features. The comparison of these methods concludes advantages and drawbacks of both methods that shown CNN have advantage of recognition of the action (for CNN only walking, running and waving) however ABM basically used for object recognition task (not particularly for action recognition).