PointMixer: MLP-Mixer for Point Cloud Understanding

This is an official implementation for the paper,

PointMixer: MLP-Mixer for Point Cloud Understanding
Jaesung Choe*, Chunghyun Park*, Francois Rameau, Jaesik Park, and In So Kweon
European Conference on Computer Vision (ECCV), Tel Aviv, Israel, 2022
[Paper] [Video] [VideoSlide] [Poster]

(*: equal contribution)

(TL;DR) Pytorch implementation of PointMixer:zap: and Point Transformer:zap:

We are currently updating this repository :fire:

Features

1. Universal point set operator: intra-set, inter-set, and hier-set mixing

• Newly revisit the use of K-Nearest Neighbors
  
• Can process arbitrary number of points
  
  

2. Symmetric encoder-decoder network for point clouds

• Maintain the hierarchical relation among points
  
• Design learning-based transition up/down layers (i.e., hier-set mixing)
  
  

3. Parameter efficient design (6.5M)

References

@article{choe2021pointmixer,
  title={PointMixer: MLP-Mixer for Point Cloud Understanding},
  author={Choe, Jaesung and Park, Chunghyun and Rameau, Francois and Park, Jaesik and Kweon, In So},
  journal={arXiv preprint arXiv:2111.11187},
  year={2021}
}