Awesome

Learning to Discover Reflection Symmetry via Polar Matching Convolution

<p align="center"> Ahyun Seo*, Woohyeon Shim*, Minsu Cho </p> <p align="center"> <a href="https://arxiv.org/abs/2108.12952">[paper]</a> <a href="http://cvlab.postech.ac.kr/research/PMCNet">[project page]</a> </p>

Official PyTorch implementation of Learning to Discover Reflection Symmetry via Polar Matching Convolution (ICCV 2021).

Contributors of this repo: Woohyeon Shim, Ahyun Seo

Environment

    conda create --name pmcnet python=3.7
    conda activate pmcnet
    conda install pytorch==1.7.0 torchvision==0.8.1 cudatoolkit=11.0 -c pytorch
    conda install -c conda-forge matplotlib
    pip install albumentations tqdm parmap scikit-image pycocotools opencv-python
    
    mkdir weights
    # setup coco_path and sym_datasets
    cd bsds
    python setup.py build_ext --inplace

Datasets

symmetry detection datasets (LDRS, SDRW, NYU) (passwd: ldrs2021) sym_datasets
COCO dataset (2014) train2014 val2014 annotations

.
├── coco_path
│   ├── train2014
│   ├── val2014
│   └── annotations
├── sym_datasets
│   ├── NYU
│   ├── SDRW
│   └── LDRS
├── (...) 
└── main.py

Training

The trained weights and arguments will be save to the checkpoint path corresponding to the VERSION_NAME.

    CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py --ver VERSION_NAME

Test

trained weights of PMCNet (passwd: ldrs2021) weights

    CUDA_VISIBLE_DEVICES=0 python main.py --ver ours -t

References

Python port of BSDS 500 link
spb-mil link
NYU database link
SDRW dataset (CVPR 2013 challenge) link
COCO preprocessing link

Citation

If you find our code or paper useful to your research work, please consider citing:

@inproceedings{seoshim2021pmcnet,
    author   = {Seo, Ahyun and Shim, Woohyeon and Cho, Minsu},
    title    = {Learning to Discover Reflection Symmetry via Polar Matching Convolution},
    booktitle= {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    year     = {2021}
}