FreePoint: Unsupervised Point Cloud Instance Segmentation

Zhikai Zhang  Jian Ding*  Li Jiang  Dengxin Dai  Gui-Song Xia* 
Wuhan University KAUST CUHK-Shenzhen Huawei Zurich Research Center

News 🔥

• [2024-02] FreePoint is accepted by CVPR 2024. Thanks for the recognition!

TODOs

• Release class-agnostic instance segmentation training codebase
• Release pretrained checkpoints

Prepare

Please refer to Mask3D for detailed dataset and environment preparation.

Code Structure

We adapt the codebase of Mask3D and Mix3D, which provide a highly modularized framework for 3D Segmentation based on the MinkowskiEngine.

FreePoint
│   ├── main_instance_segmentation_freepoint.py <- the main file
│   ├── conf                          <- hydra configuration files
│   ├── datasets
│   │   ├── preprocessing             <- folder with preprocessing scripts
│   │   ├── semseg_freepoint.py                 <- indoor dataset
│   │   └── utils_freepoint.py        
│   ├── models                        <- Mask3D modules
│   ├── trainer
│   │   ├── __init__.py
│   │   └── trainer.py                <- train loop
│   └── utils
├── data
│   ├── processed                     <- folder for preprocessed datasets
│   └── raw                           <- folder for raw datasets
├── scripts                           <- train scripts
├── docs
├── README.md
└── saved                             <- folder that stores models and logs

Dependencies:

The main dependencies of the project are the following:

python: 3.10.9
cuda: 11.3

BibTex

If you find this repository helpful, please cite our work:

@inproceedings{zhang2024freepoint,
  title={Freepoint: Unsupervised point cloud instance segmentation},
  author={Zhang, Zhikai and Ding, Jian and Jiang, Li and Dai, Dengxin and Xia, Guisong},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={28254--28263},
  year={2024}
}