InST: Industrial Style Transfer with Large-scale Geometric Warping and Content Preservation

This repository contains the source code for our paper:

Industrial Style Transfer with Large-scale Geometric Warping and Content Preservation
CVPR 2022
Jinchao Yang, Fei Guo, Shuo Chen, Jun Li and Jian Yang

Requirement

Install PyTorch 1.6.0+ and corresponding package

pip install -r requirements.txt

Install PointRender backup to segment images during test time.

git clone https://github.com/facebookresearch/detectron2.git
pip install -e detectron2/

Data preparation

You should prepare the source mask dataset and the target mask dataset respectively.

/path/to/source_mask_root
    mask1.jpg
    mask2.jpg

/path/to/target_mask_root
    mask1.jpg
    mask2.jpg

Evaluation

Using PointRender to segment source and target images. This will take a while.(you should make sure the object can be segmented by PointRender)

python test_LGW.py --source_path path/to/source_img --target_path /path/to/target_img --checkpoint /path/to/checkpoint

If you want to save time and generate more accurate results, you can provide source mask and target mask path in the command line.

python test_LGW.py --source_path path/to/source_img --source_mask_path /path/to/source_mask --target_mask_path /path/to/target_mask --checkpoint /path/to/checkpoint

Training

CUDA_VISIBLE_DEVICES=0 python main.py --source_dir /path/to/source_mask_root --target_dir /path/to/target_mask_root 

Acknowledgments

Parts of the code are based on RAFT.

If you find this code useful for your research, please cite

@InProceedings{jcyang2022InST,
    title = {Industrial Style Transfer with Large-scale Geometric Warping and Content Preservation},
    author = {Jinchao Yang and Fei Guo and Shuo Chen and Jun Li and Jian Yang},
    booktitle = {CVPR},
    year = {2022},
}