Unsupervised Snowy-Weather Point Cloud Denoising via Two-Stage Filter-Network Collaboration (opens in new tab)

Reliable light detection and ranging (LiDAR) point cloud data is fundamental to real-time outdoor perception applications such as autonomous driving and mobile robotics; however, snow-induced noise severely disrupts point cloud structure and significantly degrades the performance of downstream algorithms. Although denoising can mitigate such effects, existing approaches face notable limitations: filter-based methods require no labels but suffer from slow inference, making them difficult to de...