Rotation-invariant-deep-pointcloud-analysis

Code for the ICCV 2021 paper: A-closer-look-at-rotation-invariant-deep-pointcloud-analysis.

Authors: Feiran Li, Kent Fujiwara, Fumio Okura, and Yasuyuki Matsushita

1. Environment

The provide codes have been tested with Pytorch-1.6.0 on a Tesla-V100.

1. Run the code

1. Download the PCA-processed datasets (ModelNet40, ShapeNet-PartSeg, and ScanObjectNN) and unzip them to the dataset folder.
2. Note that the ScanObjectNN dataset is originally provided here. Please pay attention to citation.
3. Run respective *_test.ipynb to test the pretrained model and *_train.ipynb to train from scratch.
4. If you want to generate the 24 ambiguities of your own dataset, please see the generate_24_pca_poses.py script.

3. Contact

Please feel free to raise an issue or email to [email protected] if you have any question regarding the paper or any suggestions for further improvements.

4. Citation

If you find this code helpful, thanks for citing our work as

@inproceedings{li2021rotinv,
title = {A Closer Look at Rotation-invariant Deep Point Cloud Analysis},
author = {Feiran Li and Kent Fujiwara and Fumio Okura and Yasuyuki Matsushita},
booktitle = {IEEE/CVF International Conference on Computer Vision (ICCV)},
year = {2021}
}