🚀 This is the official PyTorch implementation of the AAAI 2025 paper:
Spiking Point Transformer for Point Cloud Classification

Ensure your directory structure follows:

│Spiking_Point_Transformer/
├── config/
├── data/
│   ├── modelnet40_normal_resampled/
│   ├── ScanObjectNN/
├── .......

ModelNet40 Dataset:
Download the aligned ModelNet dataset from here and extract it to data/modelnet40_normal_resampled/.

ScanObjectNN Dataset:
Download the official data from here and extract it into data/ScanObjectNN/.

Tested on Ubuntu 20.04, CUDA 11.7, PyTorch 1.13.1, Python 3.9.

conda create -y -n spt python=3.9
conda activate spt
conda install pytorch==1.13.1 torchvision==0.14.1 pytorch-cuda=11.7 -c pytorch -c nvidia

pip install -r requirements.txt
cd ops/fps/ && python setup.py install

Modify the method in config/cls.yaml and config/model/Hengshuang.yaml, then run:

python train_cls.py

If you find this work useful, please consider citing:

@article{wu2025spiking,
  title={Spiking Point Transformer for Point Cloud Classification},
  author={Wu, Peixi and Chai, Bosong and Li, Hebei and Zheng, Menghua and Peng, Yansong and Wang, Zeyu and Nie, Xuan and Zhang, Yueyi and Sun, Xiaoyan},
  journal={arXiv preprint arXiv:2502.15811},
  year={2025}
}

The project is largely based on Point-Transformer and has incorporated numerous code snippets from Spike-Driven-Transformer, SpikingJelly from SpikingJelly. Many thanks to these three projects for their excellent contributions!

Feel free to contribute and reach out if you have any questions!