kmjz/photosketchman

Code for paper "High-Quality Facial Photo-Sketch Synthesis Using Multi-Adversarial Networks"

PhotoSketchMAN

High-Quality Facial Photo-Sketch Synthesis Using Multi-Adversarial Networks

Lidan Wang, Vishwanath Singagi, Vishal M. Patel

Paper Link

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

@inproceedings{wang2018high,
  title={High-quality facial photo-sketch synthesis using multi-adversarial networks},
  author={Wang, Lidan and Sindagi, Vishwanath and Patel, Vishal},
  booktitle={Automatic Face \& Gesture Recognition (FG 2018), 2018 13th IEEE International Conference on},
  pages={83--90},
  year={2018},
  organization={IEEE}
}

Prerequisites

• Linux
• Python 2 or Python 3
• NVIDIA GPU + CUDA CuDNN

Getting Started

Installation

• Install PyTorch and dependencies from http://pytorch.org/
• Install Torch vision from the source.
• Install python libraries visdom.

pip install visdom

Datasets

• CUHK Face Sketch Database (CUFS)
• CUHK Face Sketch FERET Database (CUFSF)

Training

make sure the data folder has subfolders trainA, trainB, valA, valB, testA, testB

python PS2MAN.py --dataroot 'data/cuhkdata_augmented' --ckpt_path 'ckpt'

Related Projects

CycleGAN: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
pix2pix: Image-to-image translation with conditional adversarial nets
DualGAN: DualGAN: Unsupervised Dual Learning for Image-to-Image Translation

Acknowledgments

This code is modified based on:

CycleGAN: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks

We thank the authors for their great work.