BEVUDA: Multi-geometric Space Alignments for Domain Adaptive BEV 3D Object Detection

arXiv:2211.17126v2 Announce Type: replace
Abstract: Vision-centric bird-eye-view (BEV) perception has shown promising potential in autonomous driving. Recent works mainly focus on improving efficiency or accuracy but neglect the challenges when facing environment changing, resulting in severe degradation of transfer performance. For BEV perception, we figure out the significant domain gaps existing in typical real-world cross-domain scenarios and comprehensively solve the Domain Adaption (DA) problem for multi-view 3D object detection. Since BEV perception approaches are complicated and contain several components, the …

3d object 3d object detection abstract accuracy arxiv autonomous autonomous driving bird challenges cs.cv detection domain driving efficiency environment figure focus improving object perception performance space transfer type view vision