EADT

Face sketch synthesis, image-to-image translation, generative adversarial network, edge-aware domain transformation, edge-preserving filter.

We provide PyTorch implementations for our TIFS2022 paper EADT:

@article{zhang2022edge,
  title={Edge Aware Domain Transformation for Face Sketch Synthesis},
  author={Zhang, Congyu and Liu, Decheng and Peng, Chunlei and Wang, Nannan and Gao, Xinbo},
  journal={IEEE Transactions on Information Forensics and Security},
  volume={17},
  pages={2761--2770},
  year={2022},
  publisher={IEEE}
}

This project can generate face sketch from photos using the GAN-based model. Paper@IEEE Code@Github

Framework

Prerequisites

Linux
Python 3.7
Pytorch-lightning 0.7.5
NVIDIA GPU + CUDA + CuDNN

Getting Started

Installation

Clone this repo:

git clone https://github.com/shenhaiyoualn/EADT
cd EADT

Install all the dependencies by:
```
pip install -r requirements.txt
```

train/test

Download CUFS and CUFSF dataset and put it in the dataset directory
Download parsing model and move it to /model/parsing/cp/
You need to process the file into the format of the /dataset/list_train_sample.txt directory (mat1 and mat2 represent the results of the segmentation process of photos and sketches, respectively）

Train a model

python train.py --dataset_train_list train_sample.txt --dataset_test_list test_sample.txt   --name eadt

Test the model

python test.py  --input_size 256  --checkpoint_dir /home/sd01/EADT/checkpoint/eadt.ckpt

Preprocessing steps

If you need to use your own data, please align all faces by eyes and the face parsing is segmented by face-parsing

Citation

If you use this code for your research, please cite our paper.

Zhang, C., Liu, D., Peng, C., Wang, N., & Gao, X. (2022). Edge Aware Domain Transformation for Face Sketch Synthesis. IEEE Transactions on Information Forensics and Security, 17, 2761-2770. (Accepted)

bibtex:

@article{zhang2022edge,
  title={Edge Aware Domain Transformation for Face Sketch Synthesis},
  author={Zhang, Congyu and Liu, Decheng and Peng, Chunlei and Wang, Nannan and Gao, Xinbo},
  journal={IEEE Transactions on Information Forensics and Security},
  volume={17},
  pages={2761--2770},
  year={2022},
  publisher={IEEE}
}

Acknowledgments

Our code is inspired by GENRE and SPADE/GauGAN.

clpeng/EADT