UniPAD: A Universal Pre-training Paradigm for Autonomous Driving

arXiv:2310.08370v2 Announce Type: replace
Abstract: In the context of autonomous driving, the significance of effective feature learning is widely acknowledged. While conventional 3D self-supervised pre-training methods have shown widespread success, most methods follow the ideas originally designed for 2D images. In this paper, we present UniPAD, a novel self-supervised learning paradigm applying 3D volumetric differentiable rendering. UniPAD implicitly encodes 3D space, facilitating the reconstruction of continuous 3D shape structures and the intricate appearance characteristics of their 2D projections. The flexibility …

arxiv autonomous autonomous driving cs.cv driving paradigm pre-training training type universal