This repository corresponds to the paper "Image-Based Virtual Try-On: A Survey". If you find our survey useful for your research, please cite the following paper:
@misc{Image_Based_Virtual_Try-On_A_Survey,
title={Image-Based Virtual Try-On: A Survey},
author={Dan Song and Xuanpu Zhang and Juan Zhou and Weizhi Nie and Ruofeng Tong and An-An Liu},
year={2023},
eprint={2311.04811},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
We have listed some of the most representative works in the field of virtual fitting from recent years.
To ensure a fair test for each model, we produced a high-resolution (1024x768) version of the VITON. dataset, following the data preprocessing method of VITON-HD.
- 14221 train
- images
- cloth
- segmentation
- densepose
- keypoints
- agnostic-person
- 2032 test
High-resolution Dataset.
We evaluated the models through two perspectives: visual results and quantitative metrics.
Visual results of VITON:
Visual results of VITON-HD:
More visual results from here(BaiduYunDownload:pdub).
Quantitative metrics of VITON:
SSIM:
FID:
LPIPS:
Semantic Score:
Quantitative metrics of VITON-HD:
SSIM:
FID:
LPIPS:
Semantic Score:
Papers
| model | Release Time | Paper | Code |
|---|---|---|---|
| CAGAN | 2017 | Paper | - |
Data Sets
| Data set | Release Time | Resolution | Quantity Train/Test | Link |
|---|---|---|---|---|
| VITON | 2018 | 256*192 | 14221/2032 | Link |
| MPV | 2019 | 256*192 | 52236/10544 | Collected by us(BaiduYunDownload:ipno) |
| DeepFashion | 2016 | 1101*750 | 52712/* | Link |
| VITON-HD | 2021 | 1024*768 | 11647/2032 | Link |
| ESF | 2022 | 512*512 | 170000/10000 | Link |
| DressCode | 2022 | 1024*768 | 48392/5400 | Link |
| VITON(After Processing) | 2022 | 1024*768 | 14221/2032 | BaiduYunDownload:mq5i |
We acknowledge the contributions of awesome-virtual-try-on to the community, which saved us time in collecting literature.