Forming a sparse representation for visual place recognition using a neurorobotic approach

TL;DR

This paper presents an unsupervised neural network model inspired by the visual cortex for encoding visual information, significantly improving speed and accuracy in large-scale visual place recognition tasks.

Contribution

The novel HSD model introduces a bio-inspired sparse coding and pooling architecture for visual localization, enhancing existing VPR systems.

Findings

HSD doubles the runtime speed of LPMP.

HSD improves localization accuracy by 10%.

HSD outperforms CoHog in accuracy.

Abstract

This paper introduces a novel unsupervised neural network model for visual information encoding which aims to address the problem of large-scale visual localization. Inspired by the structure of the visual cortex, the model (namely HSD) alternates layers of topologic sparse coding and pooling to build a more compact code of visual information. Intended for visual place recognition (VPR) systems that use local descriptors, the impact of its integration in a bio-inpired model for self-localization (LPMP) is evaluated. Our experimental results on the KITTI dataset show that HSD improves the runtime speed of LPMP by a factor of at least 2 and its localization accuracy by 10%. A comparison with CoHog, a state-of-the-art VPR approach, showed that our method achieves slightly better results.