HAIFIT

HAIFIT: Human-to-AI Fashion Image Translation. Arxiv

Our model weights is available checkpoint

1. Data preparation

For datasets that have paired sketch-image data, the path should be formatted as:

./dataset/trainA/  # training reference (sketch)
./dataset/trainB/  # training ground truth (image)
./dataset/testA/  # testing reference (sketch)
./dataset/testB/  # testing ground truth (image)

After that, the configuration should be specified in the config file in:

./config.yml  # config file

Our Sketch-to-Image synthesis dataset is available HAIFashion.

2. Train and Test

2.1 Train

set model=1 in main.py and run:

python main.py

2.2 test

set model=2 in main.py and run:

python main.py

Note: you should change your checkpoint path.

3. Reference

If you find our code or dataset useful for your research, please cite our work. Thank you. 🥰🥰

@article{jiang2024haifit,
  title={HAIFIT: Human-Centered AI for Fashion Image Translation},
  author={Jiang, Jianan and Li, Xinglin and Yu, Weiren and Wu, Di},
  journal={arXiv preprint arXiv:2403.08651},
  year={2024}
}