VITON-GAN

Implementation of the paper "VITON-GAN: Virtual Try-on Image Generator Trained with Adversarial Loss" by Shion Honda.
https://diglib.eg.org/handle/10.2312/egp20191043
Preprint version is here.

Installation

Prerequisites

PIL
PyTorch
TorchVision
tqdm

In addition, you need OpenPose and Look Into Person (LIP) to get keypoints and segmentation of the human body.

Download repository

$ git clone https://github.com/shionhonda/viton-gan

Trained model

You can get trained model here.

Usage

VITON-GAN requires the keypoints from OpenPose and segmentation labels from Look Into Person.
First, prepare the following directories in viton-gan/viton_gan/data:

cloth
cloth mask
person
person-parse
pose

Second, prepare a file that makes pairs of clothing and human. For example, test_pairs.txt:

000001_0.jpg 001744_1.jpg
000010_0.jpg 004325_1.jpg
.
.
.

You can find more information here: https://github.com/sergeywong/cp-vton

After preparing the data and the list, run the following command:

$ python train_gmm.py
$ python run_gmm.py # warp clothing so that it fit on the body
$ python train_tom.py
$ python run_gmm.py # generate virtual try-on image

Cite

If you use this repository in your research, please include the paper in your references.

@inproceedings {p.20191043,
booktitle = {Eurographics 2019 - Posters},
editor = {Fusiello, Andrea and Bimber, Oliver},
title = {{VITON-GAN: Virtual Try-on Image Generator Trained with Adversarial Loss}},
author = {Honda, Shion},
year = {2019},
publisher = {The Eurographics Association},
ISSN = {1017-4656},
DOI = {10.2312/egp.20191043}
}

References

[1] BROCK A., DONAHUE J., SIMONYAN K.: Large scale GAN training for high fidelity natural image synthesis. In International Conference on Learning Representations (2019).
[2] CAO Z., SIMON T., WEI S.-E., SHEIKH Y.: Realtime multiperson 2d pose estimation using part affinity fields. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017).
[3] GONG K., LIANG X., ZHANG D., SHEN X., LIN L.: Look into person: Self-supervised structure-sensitive learning and a new benchmark for human parsing. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017).
[4] HAN X., WU Z., WU Z., YU R., DAVIS L. S.: Viton: An image-based virtual try-on network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2018).
[5] KARRAS T., LAINE S., AILA T.: A style-based generator architecture for generative adversarial networks. arXiv preprint arXiv:1812.04948 (2018).
[6] WANG B., ZHENG H., LIANG X., CHEN Y., LIN L., YANG M.: Toward characteristic-preserving image-based virtual try-on network. In Proceedings of the European Conference on Computer Vision (2018).

amalp12/viton-gan