PSA-SSL is a Pose and Size-aware self-supervised learning framework for LiDAR networks.
PSA-SSL: Pose and Size-aware Self-Supervised Learning on LiDAR Point Clouds
Barza Nisar, Steven L. Waslander
[Paper] | [Video]
Please refer to INSTALL.md for the installation of PSA-SSL.
To cluster and fit target bounding boxes:
cd PSA-SSL
python tools/preprocess_waymo.py
python tools/preprocess_semkitti.py
python tools/preprocess_nuscenes.py
To reproduce the results in the paper, pretrain and finetune with commands given in pretrain_finetune.sh.
If you use PSA-SSL in your research, please consider citing:
@inproceedings{Nisar_2025_CVPR,
author = {Nisar, Barza and Waslander, Steven L.},
title = {PSA-SSL: Pose and Size-aware Self-Supervised Learning on LiDAR Point Clouds},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2025},
pages = {6670-6679}
}
This work is based on and extends ideas and code from the following open-source repositories:
We thank the authors and contributors of these projects for making their work publicly available.