acht7111020 / DSMAP

[ECCV 2020] Domain-Specific Mappings for Generative Adversarial Style Transfer

Domain-Specific Mappings for Generative Adversarial Style Transfer

[Project Page][Paper]

Pytorch implementation for our paper Domain-Specific Mappings for Generative Adversarial Style Transfer.

Example Results

Paper

Domain-Specific Mappings for Generative Adversarial Style Transfer
Hsin-Yu Chang, Zhixiang Wang, and Yung-Yu Chuang
European Conference on Computer Vision (ECCV), 2020
[arxiv]

Citation

If you find the work is useful in your research, please consider citing:

@inproceedings{chang2020dsmap,
    author    = {Chang, Hsin-Yu and Wang, Zhixiang and Chuang, Yung-Yu},
    title     = {Domain-Specific Mappings for Generative Adversarial Style Transfers},
    booktitle = {European Conference on Computer Vision},
    year      = {2020}

Usage

Requirements

1. Python 3.6 or higher
2. Pytorch 1.2.0 or higher, torchvision 0.4.0 or higher
3. Tensorboard, TensorboardX, Pyyaml, pillow

Dataset

Download dataset from the following github repo.

• Cat2Dog (DRIT)
• Portrait (DRIT)
• Monet (CycleGAN)
• Yosemite (CycleGAN)

Train

CUDA_VISIBLE_DEVICES=[gpu] python3 train.py --config [config_path] --save_name [path_to_save]

Test

1. Download model here!

CUDA_VISIBLE_DEVICES=[gpu] python3 test.py --config [config_path] --checkpoint [checkpoint_path] --test_path [test_folder_path] --output_path [path_to_output_images] --a2b [1 or 0]

LICENSE

Copyright (C) 2020 Hsin-Yu Chang. Licensed under the CC BY-NC-SA 4.0 license.

Reference

Code inspired from:

1. https://github.com/NVlabs/MUNIT
2. https://github.com/HelenMao/MSGAN