Abstract

A novel cognition-inspired, agnostic framework is proposed for building mapsin mobile robotics that are efficient in terms of image matching/retrieval forsolving Visual Place Recognition (VPR) problem. A dataset, 'ESSEX3IN1', is alsopresented to demonstrate the significantly enhanced performance ofstate-of-the-art VPR techniques when combined with the proposed framework.

 

Quick Read (beta)