Improving Worst Case Visual Localization Coverage via Place-Specific Sub-Selection in Multi-Camera Systems

Simple item page
dc.contributor.authorHausler, S.
dc.contributor.authorXu, M.
dc.contributor.authorGarg, S.
dc.contributor.authorChakravarty, P.
dc.contributor.authorShrivastava, S.
dc.contributor.authorVora, A.
dc.contributor.authorMilford, M.
dc.date.issued2022
dc.description.abstract6-DoF visual localization systems utilize principled approaches rooted in 3D geometry to perform accurate camera pose estimation of images to a map. Current techniques use hierarchical pipelines and learned 2D feature extractors to improve scalability and increase performance. However, despite gains in typical recall@0.25mtype metrics, these systems still have limited utility for real-world applications like autonomous vehicles because of their worst areas of performance - the locations where they provide insufficient recall at a certain required error tolerance. Here we investigate the utility of using place specific configurations, where a map is segmented into a number of places, each with its own configuration for modulating the pose estimation step, in this case selecting a camera within a multi-camera system. On the Ford AV benchmark dataset, we demonstrate substantially improved worst-case localization performance compared to using off-the-shelf pipelines - minimizing the percentage of the dataset which has low recall at a certain error tolerance, as well as improved overall localization performance. Our proposed approach is particularly applicable to the crowdsharingmodel of autonomous vehicle deployment, where a fleet of AVs are regularly traversing a known route.
dc.description.statementofresponsibilityStephen Hausler, Ming Xu, Sourav Garg, Punarjay Chakravarty, Shubham Shrivastava, Ankit Vora, and Michael Milford
dc.identifier.citationIEEE Robotics and Automation Letters, 2022; 7(4):10112-10119
dc.identifier.doi10.1109/LRA.2022.3191174
dc.identifier.issn2377-3766
dc.identifier.issn2377-3766
dc.identifier.orcidGarg, S. [0000-0001-6068-3307]
dc.identifier.urihttps://hdl.handle.net/2440/138436
dc.language.isoen
dc.publisherInstitute of Electrical and Electronics Engineers
dc.relation.granthttp://purl.org/au-research/grants/arc/FL210100156
dc.rights© 2022 IEEE.
dc.source.urihttps://doi.org/10.1109/lra.2022.3191174
dc.subjectAutonomous vehicle navigation; deep learning methods; localization; multi camera system
dc.titleImproving Worst Case Visual Localization Coverage via Place-Specific Sub-Selection in Multi-Camera Systems
dc.typeJournal article
pubs.publication-statusPublished

Files

Collections

Australian Institute for Machine Learning publications