AdaCoSeg

@inproceedings{zhu2020adacoseg,
  title={Adacoseg: Adaptive shape co-segmentation with group consistency loss},
  author={Zhu, Chenyang and Xu, Kai and Chaudhuri, Siddhartha and Yi, Li and
          Guibas, Leonidas J and Zhang, Hao},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and
             Pattern Recognition},
  pages={8543--8552},
  year={2020}
}

• Deep Learning based
• Unsupervised Learning actually (“weakly” supervised)
• Adaptive 3D Shape Co-Segmentation

Method #

1. Predict noisy part classification using PointNet++
2. Use a pre-trained “part-prior” network to refine per shape parts
   • Supervised training (offline)

• Use a Group consistency loss
  • “Things that are grouped together should be similar in structure”

• Goal: Encourage a low loss by having a “low rank” matrix, meaning that the individual feature vectors should be close together (?)