/3D-Pretraining

Self-supervised Point Cloud Representation Learning via Separating Mixed Shapes

Primary LanguagePython

License CC BY-NC-SA 4.0 Python 3.6

3D-Pretraining

Self-supervised Point Cloud Representation Learning via Separating Mixed Shapes

pdf

Prerequisites

  • Python 3.6
  • GPU memory >= 8G
  • NumPy
  • PyTorch 1.0+

ShapeNetPart

Download dataset from ShapeNetPart

ln -s <dataset_path> data
python train.py --opt ./mixup.yml

This is the part of the code. The complete code will be released after integration into the new code framework, such as more datasets, more backbones.

ModelNet

TODO

Citation

Please cite this paper if it helps your research:

@article{sun2022self,
  title={Self-supervised Point Cloud Representation Learning via Separating Mixed Shapes},
  author={Sun, Chao and Zheng, Zhedong and Wang, Xiaohan and Xu, Mingliang and Yang, Yi},
  journal={IEEE Transactions on Multimedia},
  year={2022},
  publisher={IEEE}
}