Pytorch implementation of our method for weakly supervised semantic matching.

Contact: Yun-Chun Chen (ycchen918 at citi dot sinica dot edu dot tw)

Please cite our paper if you find it useful for your research.

@inproceedings{Chen_WeakMatchNet_2018,
  author = {Y.-C. Chen and P.-H. Huang and L.-Y. Yu and J.-B. Huang and M.-H. Yang and Y.-Y. Lin},
  booktitle = {Asian Conference on Computer Vision (ACCV)},
  title = {Deep Semantic Matching with Foreground Detection and Cycle-Consistency},
  year = {2018}
}

• Install PyTorch

• Clone this repo

git clone https://github.com/YunChunChen/WeakMatchNet
cd WeakMatchNet

• Please use the code to download the datasets and put it under the data/ folder.

• Please download the pre-trained model for training here and put it under the trained_models/resnet101/ folder.

• Evaluation command

sh eval.sh

• Training command

sh train.sh

• Rocco et al. Convolutional Neural Network Architecture for Geometric Matching. In CVPR, 2017. [project] [paper] [code]
• Rocco et al. End-to-End Weakly-Supervised Semantic Alignment. In CVPR 2018. [project] [paper] [code]

This code is heavily borrowed from weakalign.

The model and code are available for non-commercial research purposes only.

• Nov 2018: code released!