Visual attention and novelty detection: experiments with automatic scale selection

Abstract

We present experiments with an autonomous inspection robot, whose task was to highlight novel features in its environment using camera images. Experiments were conducted with two different attention mechanisms — saliency map and multiscale Harris detector — and two different novelty detection mechanisms — the Grow-When- Required neural network and incremental PCA. For both mechanisms we compared fixed-scale image encoding with automatically scaled image patches. Results show that using automatic scale selection provides a more efficient representation of the visual input space, but that performance is generally better using a fixed-scale image encoding.