Skip to main navigation menu Skip to main content Skip to site footer

Comparison of Foveated Downsampling Techniques in Image Recognition

Abstract

Foveation is an important part of human vision, and a number of deep networks have also used foveation. However, there have been few systematic comparisons between foveating and non-foveating deep networks, and between different variable-resolution downsampling methods. Here we define several such methods, and compare their performance on ImageNet recognition with a custom Densenet network. The best variable-resolution method slightly outperforms uniform downsampling. Thus in our experiments, foveation does not substantially help or hinder object recognition in deep networks.
PDF