Research papers · Posted 2023.12.19 · Preprint
M-BEV: Masked BEV Perception for Robust Autonomous Driving
M-BEV: Masked BEV Perception for Robust Autonomous Driving
Overview
M-BEV trains with masked camera features and reconstructs missing views from cross-view spatiotemporal context. A modular reconstruction component is evaluated for 3D perception resilience when one or more camera streams fail.
Author-supplied source information
Author-supplied information on arXiv.
Github repository: https://github.com/Sranc3/M-BEV
Version: v1