search

myavartanoo / PolyNet_PyTorch

Official implementation of the paper "PolyNet: Polynomial Neural Network for 3D Shape Recognition with PolyShape Representation" (3DV 2021)
21 stars 0 forks source link
deep-learning graphcnn graphconv pytorch

readme

PolyNet_Pytorch

This repository contains the official code to reproduce the results from the paper:

PolyNet: Polynomial Neural Network for 3D Shape Recognition with PolyShape Representation (3DV 2021)

[project page] [arXiv] [ResearchGate] [presentation]

Dependencies

Dataset

Download the preprocessed ModelNet dataset with PTQ and √3-subdivision from the follwing link and unzip them in the data directroy. The data type is .npz.

[PTQ] [√3-subdivision]

Train

In config.json you can set dataset type (ModelNet10 or ModelNet40) and the PolyPool type (PTQ, Sqrt3).

To train PolyNet with the desired dataset and PolyPool, simply run, 

CUDA_VISIBLE_DEVICES=0 python train.py --config config.json -t "direction to save the model"

Citation

If you find our paper, code, or provided data useful, please consider citing:

@INPROCEEDINGS{9665897,
  author={Yavartanoo, Mohsen and Hung, Shih-Hsuan and Neshatavar, Reyhaneh and Zhang, Yue and Lee, Kyoung Mu},
  booktitle={2021 International Conference on 3D Vision (3DV)}, 
  title={PolyNet: Polynomial Neural Network for 3D Shape Recognition with PolyShape Representation}, 
  year={2021},
  pages={1014-1023},
  doi={10.1109/3DV53792.2021.00109}}