3D-UnOutDet: A Fast and Efficient Unsupervised Snow Removal Algorithm for 3D LiDAR Point Clouds

Abu Mohammed Raisuddin, Idriss Gouigah, Eren Erdal Aksoy

Abstract: In this work, we propose a novel, fast, and memory-efficient unsupervised statistical method, combined with an unsupervised deep learning (DL) model, for de-snowing LiDAR point clouds in a fully unsupervised fashion. The results obtained on the real-scanned WADS dataset show that our model achieves a 6.3% improvement in mIOU over the current state-of-the-art unsupervised DL methods and performs comparable to supervised counterparts, substantially narrowing the performance gap between supervised and unsupervised approaches. Along with that, our model also outperforms its nearest competitor by 12.8% mIOU when tested on our CADC annotations. Additionally, our de-snowing algorithm enhances downstream semantic segmentation and object detection tasks without even requiring any modifications to the base segmentation and detection models. The source code, trained models, and online supplementary information are available at: https://sporsho.github.io/3DUnOutDet.

External IDs:doi:10.36227/techrxiv.172865969.94242670/v1