Multi-Modal Data-Efficient 3D Scene Understanding for Autonomous Driving

arXiv:2405.05258v1 Announce Type: cross
Abstract: Efficient data utilization is crucial for advancing 3D scene understanding in autonomous driving, where reliance on heavily human-annotated LiDAR point clouds challenges fully supervised methods. Addressing this, our study extends into semi-supervised learning for LiDAR semantic segmentation, leveraging the intrinsic spatial priors of driving scenes and multi-sensor complements to augment the efficacy of unlabeled datasets. We introduce LaserMix++, an evolved framework that integrates laser beam manipulations from disparate LiDAR scans and incorporates LiDAR-camera correspondences to …

abstract arxiv autonomous autonomous driving challenges cs.cv cs.lg cs.ro data driving human intrinsic lidar modal multi-modal reliance segmentation semantic semi semi-supervised semi-supervised learning sensor spatial study supervised learning type understanding