Awesome

PointMixer: MLP-Mixer for Point Cloud Understanding

This is an official implementation for the paper,

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

<br/>

(*: equal contribution)

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

We are currently updating this repository :fire:

• semseg<br/>
  • methods
    • pointmixer
    • point transformer
  • s3dis weights
  • scannet weights
  • logger option (tensorboard / neptune)
• objcls<br/>
• recon<br/>

Features

1. Universal point set operator: intra-set, inter-set, and hier-set mixing <br/>

• Newly revisit the use of K-Nearest Neighbors <br/>
• Can process arbitrary number of points <br/> <img src="./fig/universal point set operator.PNG" width="560" > <br/>

2. Symmetric encoder-decoder network for point clouds <br/>

• Maintain the hierarchical relation among points <br/>
• Design learning-based transition up/down layers (i.e., hier-set mixing) <br/> <img src="./fig/symmetric.PNG" width="572" > <br/>

3. Parameter efficient design (6.5M) <br/>

<img src="./fig/arch.PNG" width="617" > <br/>

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}
}