ToThePoint

Pytorch implementation for ToThePoint results in the paper ToThePoint: Efficient Contrastive Learning of 3D Point Clouds via Recycling by Xinglin Li, Jiajing Chen, Jinhui Ouyang, Hanhui Deng, Senem Velipasalar, Di Wu in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023.

Dependencies

python 3.8

Pytorch 1.7.1

In addition, please add the project folder to PYTHONPATH and pip install the following packages:

• ***
• ***
• ***

Evaluation Results

• 3D object classification

• Few-shot 3D object classification

• 3D object part segmentation:

• ablation study:

Citing ToThePoint

If you find ToThePoint useful in your research, please consider citing: BibTex:

@InProceedings{Li_2023_CVPR,
    author    = {Li, Xinglin and Chen, Jiajing and Ouyang, Jinhui and Deng, Hanhui and Velipasalar, Senem and Wu, Di},
    title     = {ToThePoint: Efficient Contrastive Learning of 3D Point Clouds via Recycling},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2023},
    pages     = {21781-21790}
}

or

Xinglin Li, Jiajing Chen, Jinhui Ouyang, Hanhui Deng, Senem Velipasalar, Di Wu; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 21781-21790.

Reference

• CrossPoint: Self-Supervised Cross-Modal Contrastive Learning for 3D Point Cloud Understanding (CVPR'22) [code]