Add to Chrome
Add to Firefox
Add to EdgeVPR-Bench: An Open-Source Visual Place Recognition Evaluation Framework with Quantifiable Viewpoint and Appearance Change
May 17, 2020
Visual Place Recognition (VPR) is the process of recognising a previously visited place using visual information, often under varying appearance conditions and viewpoint changes and with computational constraints. VPR is a critical component of many autonomous navigation systems ranging from autonomous vehicles to drones. While the concept of place recognition has been around for many years, VPR research has grown rapidly as a field over the past decade due to both improving camera hardware technologies and its suitability for application of deep learning-based techniques. With this growth however has come field fragmentation, lack of standardisation and a disconnect between current performance metrics and the actual utility of a VPR technique at application-deployment. In this paper we address these key challenges through a new comprehensive open-source evaluation framework, dubbed 'VPR-Bench'. VPR-Bench introduces two much-needed capabilities for researchers: firstly, quantification of viewpoint and illumination variation, replacing what has largely been assessed qualitatively in the past, and secondly, new metrics 'Extended precision' (EP), 'Performance-Per-Compute-Unit' (PCU) and 'Number of Prospective Place Matching Candidates' (NPPMC). These new metrics complement the limitations of traditional Precision-Recall curves, by providing measures that are more informative to the wide range of potential VPR applications. Mechanistically, we develop new unified templates that facilitate the implementation, deployment and evaluation of a wide range of VPR techniques and datasets. We incorporate the most comprehensive combination of state-of-the-art VPR techniques and datasets to date into VPR-Bench and demonstrate how it provides a rich range of previously inaccessible insights, such as the nuanced relationship between viewpoint invariance, different types of VPR techniques and datasets.
