VisionPAD: A Vision-Centric Pre-training Paradigm for Autonomous Driving
This addresses the need for efficient pre-training in autonomous driving by offering a novel paradigm that reduces reliance on explicit depth supervision, though it is incremental in advancing existing self-supervised techniques.
The paper tackles the problem of vision-centric pre-training for autonomous driving by introducing VisionPAD, a self-supervised method that uses 3D Gaussian Splatting and multi-view reconstruction with only image supervision, resulting in significant performance improvements in 3D object detection, occupancy prediction, and map segmentation over state-of-the-art strategies.
This paper introduces VisionPAD, a novel self-supervised pre-training paradigm designed for vision-centric algorithms in autonomous driving. In contrast to previous approaches that employ neural rendering with explicit depth supervision, VisionPAD utilizes more efficient 3D Gaussian Splatting to reconstruct multi-view representations using only images as supervision. Specifically, we introduce a self-supervised method for voxel velocity estimation. By warping voxels to adjacent frames and supervising the rendered outputs, the model effectively learns motion cues in the sequential data. Furthermore, we adopt a multi-frame photometric consistency approach to enhance geometric perception. It projects adjacent frames to the current frame based on rendered depths and relative poses, boosting the 3D geometric representation through pure image supervision. Extensive experiments on autonomous driving datasets demonstrate that VisionPAD significantly improves performance in 3D object detection, occupancy prediction and map segmentation, surpassing state-of-the-art pre-training strategies by a considerable margin.