PointVST: Self-Supervised Pre-training for 3D Point Clouds via View-Specific Point-to-Image Translation

This is the official implementation of [PointVST] (TVCG 2023), a self-supervised learning approach for pre-training deep 3D point cloud backbone encoders.

Given a backbone point cloud encoder, we pre-train its model parameters via our proposed pretext task of view-specific point-to-image translation (refer to the demo scripts provided in main/backbone_pretraining).

The pre-trained backbones can be integrated into task-specific learning frameworks for downstream task scenarios (refer to the demo scripts provided in main/example_downstream).

The used datasets can be downloaded from here (password: 7kmq), and the checkpoints of pre-trained backbone networks as well as various task-specific learning frameworks can also be downloaded from here (password: iyxx).

Citation

If you find our work useful in your research, please consider citing:

@article{zhang2023pointvst,
  title={PointVST: Self-Supervised Pre-Training for 3D Point Clouds Via View-Specific Point-to-Image Translation},
  author={Zhang, Qijian and Hou, Junhui},
  journal={IEEE Transactions on Visualization and Computer Graphics},
  year={2023}
}