• Stars
    star
    117
  • Rank 301,828 (Top 6 %)
  • Language
    Python
  • License
    MIT License
  • Created over 5 years ago
  • Updated about 3 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing (ICCV 2019)

DensePoint

This repository contains the code in Pytorch for the paper:

DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing [arXiv] [CVF]
Yongcheng Liu, Bin Fan, Gaofeng Meng, Jiwen Lu, Shiming Xiang and Chunhong Pan
ICCV 2019

Citation

If our paper is helpful for your research, please consider citing:

    @inproceedings{liu2019densepoint,   
        author = {Yongcheng Liu and    
                        Bin Fan and  
                   Gaofeng Meng and
                       Jiwen Lu and
                  Shiming Xiang and   
                       Chunhong Pan},   
        title = {DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing},   
        booktitle = {IEEE International Conference on Computer Vision (ICCV)},    
        pages = {5239--5248},  
        year = {2019}   
    }   

Usage: Preparation

  • Requirement

    • Ubuntu 14.04
    • Python 3 (recommend Anaconda3)
    • Pytorch 0.3.*
    • CMake > 2.8
    • CUDA 8.0 + cuDNN 5.1
  • Building Kernel

    git clone https://github.com/Yochengliu/DensePoint.git 
    cd DensePoint
    mkdir build && cd build
    cmake .. && make
    
  • Dataset

    • Shape Classification: download and unzip ModelNet40 (415M). Replace $data_root$ in cfgs/config_cls.yaml with the dataset parent path.

Usage: Training

  • Shape Classification

    sh train_cls.sh
    

We have trained a 6-layer classification model in cls folder, whose accuracy is 92.38%.

Usage: Evaluation

  • Shape Classification

    Voting script: voting_evaluate_cls.py
    

You can use our model cls/model_cls_L6_iter_36567_acc_0.923825.pth as the checkpoint in config_cls.yaml, and after this voting you will get an accuracy of 92.5% if all things go right.

License

The code is released under MIT License (see LICENSE file for details).

Acknowledgement

The code is heavily borrowed from Pointnet2_PyTorch.

Contact

If you have some ideas or questions about our research to share with us, please contact [email protected]