This project supports real-time virtual try-on without the need for any special sensors. All you need is a webcam and a PC with a GPU equivalent to an RTX 3060 or higher.
Unlike other image-based virtual try-on methods, our approach requires training a dedicated network for each garment in a low-barrier, accessible way (see this paper). Our method simply overlay the synthesized garment on the top of human body without removing the original garment to achieve real-time performace.
Check our tight garment demo and loose garment demo on YouTube.
git clone https://github.com/ZaiqiangWu/RTV.git
cd RTV
sudo apt install gcc g++ libxcb-xinerama0-dev libxcb-icccm4-dev libxcb-image0-dev libxcb-keysyms1-dev libxcb-randr0-dev libxcb-shape0-dev libxcb-sync-dev libxcb-xfixes0-dev libxcb-xkb-dev
sudo apt install qtbase5-dev qtbase5-dev-tools libqt5gui5 libqt5widgets5 libqt5multimedia5 libqt5multimediawidgets5 libqt5multimedia5-plugins libpulse-mainloop-glib0
conda create -n rtv python=3.9
conda activate rtv
pip install -r requirements.txt
pip install detectron2@git+https://github.com/facebookresearch/detectron2.git
pip install git+https://github.com/ZaiqiangWu/ROMP.git#subdirectory=simple_romp
Download our pretrained checkpoints by running the following command. Please refer to this page for instructions on training your own garment items.
sudo apt install git-lfs
git lfs install
git clone https://huggingface.co/wuzaiqiang/rtv_ckpts
Please make sure that there is a webcam connected to your PC before running the following command.
python rtl_demo.py
@article{10.1111:cgf.70272,
journal = {Computer Graphics Forum},
title = {{Real-Time Per-Garment Virtual Try-On with Temporal Consistency for Loose-Fitting Garments}},
author = {Wu, Zaiqiang and Shen, I-Chao and Igarashi, Takeo},
year = {2025},
publisher = {The Eurographics Association and John Wiley & Sons Ltd.},
ISSN = {1467-8659},
DOI = {10.1111/cgf.70272}
}
@misc{wu2025lowbarrier,
title={Low-Barrier Dataset Collection with Real Human Body for Interactive Per-Garment Virtual Try-On},
author={Zaiqiang Wu and Yechen Li and Jingyuan Liu and Yuki Shibata and Takayuki Hori and I-Chao Shen and Takeo Igarashi},
year={2025},
eprint={2506.10468},
archivePrefix={arXiv},
primaryClass={cs.GR}
}
This project is licensed under a custom Apache 2.0-based license with a Non-Commercial Use clause.
You may use, modify, and share the code freely for non-commercial purposes.
Commercial use requires prior written permission from the author.
See the LICENSE file for details.