BAE-NET

@inproceedings{chen2019bae,
  title={BAE-NET: branched autoencoder for shape co-segmentation},
  author={Chen, Zhiqin and Yin, Kangxue and Fisher, Matthew and Chaudhuri, Siddhartha and Zhang, Hao},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={8490--8499},
  year={2019}
}

Goal 3D Shape Co-Segmentation
Uses a Branched autoencoder

When trained Supervised :
- Each branch is responsible for a (predefined) segment
When trained Unsupervised :
- Branches are predicting consistent parts of the shape
- Loss only shape reconstruction loss