Pytorch implementation of our method for adapting semantic segmentation from the synthetic dataset (source domain) to the real dataset (target domain).

Contact: Yi-Hsuan Tsai (wasidennis at gmail dot com) and Wei-Chih Hung (whung8 at ucmerced dot edu)

Learning to Adapt Structured Output Space for Semantic Segmentation
Yi-Hsuan Tsai*, Wei-Chih Hung*, Samuel Schulter, Kihyuk Sohn, Ming-Hsuan Yang and Manmohan Chandraker
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018 (* indicates equal contribution).

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

@article{Tsai_adaptseg_2018,
  author = {Y.-H. Tsai and W.-C. Hung and S. Schulter and K. Sohn and M.-H. Yang and M. Chandraker},
  journal = {arXiv preprint arXiv:xxxx.xxxxx},
  title = {Learning to Adapt Structured Output Space for Semantic Segmentation},
  year = {2018}
}

• Install PyTorch from http://pytorch.org

• Clone this repo

git clone https://github.com/wasidennis/AdaptSegNet
cd AdaptSegNet

• Download the GTA5 Dataset as the source domain, and put it in the dataset/gta5 folder

• Download the Cityscapes Dataset as the target domain, and put it in the dataset/cityscapes folder

• Download the pre-trained GTA5-to-Cityscapes model and put it in the model folder

• Test the model

python ...

• Download the initial weight and put it in the model folder

• Train the GTA5-to-Cityscapes model

sh ...

This code is heavily borrowed from Pytorch-Deeplab.

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

• 02/2018: code released