Physics-Aware Diffusion for LiDAR Point Cloud Densification
For autonomous driving perception, this method provides fast, physically plausible LiDAR densification that improves downstream detection without retraining.
LiDAR point clouds suffer from distance-dependent sparsity. The proposed Scanline-Consistent Range-Aware Diffusion framework densifies point clouds via probabilistic refinement, achieving state-of-the-art results on KITTI-360 and nuScenes in 156ms, and directly boosting off-the-shelf 3D detectors without retraining.
LiDAR perception is severely limited by the distance-dependent sparsity of distant objects. While diffusion models can recover dense geometry, they suffer from prohibitive latency and physical hallucinations manifesting as ghost points. We propose Scanline-Consistent Range-Aware Diffusion, a framework that treats densification as probabilistic refinement rather than generation. By leveraging Partial Diffusion (SDEdit) on a coarse prior, we achieve high-fidelity results in just 156ms. Our novel Ray-Consistency loss and Negative Ray Augmentation enforce sensor physics to suppress artifacts. Our method achieves state-of-the-art results on KITTI-360 and nuScenes, directly boosting off-the-shelf 3D detectors without retraining. Code will be made available.