Skip to content
Projects

qq456cvb/PRIN has 87 GitHub stars and 10 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-18 11:26 UTC. Last push 2026-10-05.

How these counts are measured

Measured star growth

in 24h
0
in 7 days
0
in 15 days
Collecting history
in 1 month
Collecting history

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

Pointwise Rotation-Invariant Network (AAAI 2020)

Machine Learning

README.md

Original

PRIN: Pointwise Rotation-Invariant Network (AAAI 2020)

PyTorch implementation of Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution (AAAI 2020).

PRIN learns point-wise, rotation-invariant features for 3D point clouds: input points are converted to spherical voxels with density-aware adaptive sampling, processed with 3D spherical voxel convolutions (built on s2cnn), and mapped back to per-point features — so part segmentation works on arbitrarily rotated shapes without rotation augmentation.

News: an improved version, SPRIN, achieves much better results and is released at qq456cvb/SPRIN. It is described in PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features (IEEE TPAMI).

Dependencies

  • s2cnn and its dependencies (PyTorch, cupy, lie_learn, pynvrtc).
  • pybind11: compile the sampling extension under src/ with CMake.

Dataset and Pretrained Weights

  • Download the ShapeNet 17-category part segmentation dataset (h5py format) from Google Drive and extract it to ./hdf5_data.
  • Download the pretrained weights (trained on unrotated shapes) from Hugging Face:
hf download qq456cvb/PRIN state.pkl --local-dir .

Usage

Train:

python train.py --log_dir log --model_path ./model.py --num_workers 4

Test with pretrained weights:

python test.py --weight_path ./state.pkl --model_path ./model.py --num_workers 4

License

MIT

Citation

@inproceedings{you2020pointwise,
  title={Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution},
  author={You, Yang and Lou, Yujing and Liu, Qi and Tai, Yu-Wing and Ma, Lizhuang and Lu, Cewu and Wang, Weiming},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={34},
  number={07},
  pages={12717--12724},
  year={2020}
}