For more information, checkout the project site [website] and the paper [PDF].

• Python 3.6
• PyTorch >= 1.0.0
• numpy
• pandas

• Pascal VOC 2012 segmentation dataset (excluding images that overlap with the test split in the PF-PASCAL) for training
• PF-Pascal & PF-WILLOW datasets for evaluation
• All datasets are automatically downloaded into the data folder by running download_datasets.py
• FYI: Without the need of downloading the entire data, the csv file of PF-PASCAL test split is available in the following link: [csv file]

git clone https://github.com/cvlab-yonsei/SFNet.git
cd SFNet
python3 download_datasets.py # prepare the datasets for training/evaluation
python3 train.py # for training
python3 eval_pascal.py # evaluation on PF-Pascal dataset
python3 eval_willow.py # evaluation on PF-WILLOW dataset

• Download pre-trained weights into weights folder
  Link: [weights]

@inproceedings{lee2019sfnet,
  title={SFNet: Learning Object-aware Semantic Correspondence},
  author={Lee, Junghyup and Kim, Dohyung and Ponce, Jean and Ham, Bumsub},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={2278--2287},
  year={2019}
}