PointDif

Point Cloud Pre-training with Diffusion Models CVPR2024

We propose the first framework for point cloud pre-training based on diffusion models, called PointDif.

1. Installation

PyTorch >= 1.10.0; python >= 3.7.12; CUDA >= 11.1;

pip install -r requirements.txt

# PointNet++
pip install "git+https://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib"

# GPU kNN
pip install --upgrade https://github.com/unlimblue/KNN_CUDA/releases/download/0.2/KNN_CUDA-0.2-py3-none-any.whl

# SoftPool
git clone https://github.com/alexandrosstergiou/SoftPool.git
cd SoftPool-master/pytorch
make install

2. Data Preparation

We use ShapeNet, ScanObjectNN, ScanNet, S3DIS, SemanticKITTI in this work. See DATASET.md for details.

3. Pre-trained Models

The pretrained model is available at: [ Baidu Netdisk ] (code : tw43)

4. PointDif pre-training

Run the following command to pre-train the PointDif model on ShapeNet:

CUDA_VISIBLE_DEVICES=<GPUs> python pointdif_main.py --config cfgs/pretrain.yaml --exp_name <output_file_name>

5. PointDif fine-tuning

Run the following command to fine-tunine the PointDif model on ScanObjectNN:

CUDA_VISIBLE_DEVICES=<GPUs> python pointdif_main.py --config cfgs/finetune_scan_hardest.yaml \
--finetune_model --exp_name <output_file_name> --ckpts <path/to/pre-trained/model>

6. Acknowledgements

Our codes are built upon Point-MAE.

7. Citation

If you find PointDif useful to your research, please cite our work.

@inproceedings{zheng2024point,
  title={Point Cloud Pre-training with Diffusion Models},
  author={Zheng, Xiao and Huang, Xiaoshui and Mei, Guofeng and Hou, Yuenan and Lyu, Zhaoyang and Dai, Bo and Ouyang, Wanli and Gong, Yongshun},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={22935--22945},
  year={2024}
}