phanxuanphucnd / OOTDiffusion-train

OOTDiffusion

This repository is the official implementation of OOTDiffusion

🤗 Try out OOTDiffusion (Thanks to ZeroGPU for providing A100 GPUs)

Or try our own demo on RTX 4090 GPUs

OOTDiffusion: Outfitting Fusion based Latent Diffusion for Controllable Virtual Try-on [arXiv paper]
Yuhao Xu, Tao Gu, Weifeng Chen, Chengcai Chen
Xiao-i Research

Our model checkpoints trained on VITON-HD (half-body) and Dress Code (full-body) have been released

• 🤗 Hugging Face link
• 📢📢 We support ONNX for humanparsing now. Most environmental issues should have been addressed : )
• Please download clip-vit-large-patch14 into checkpoints folder
• We've only tested our code and models on Linux (Ubuntu 22.04)

   

Installation

1. Clone the repository

git clone https://github.com/levihsu/OOTDiffusion

2. Create a conda environment and install the required packages

conda create -n ootd python==3.10
conda activate ootd
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
pip install -r requirements.txt

Inference

1. Half-body model

cd OOTDiffusion/run
python run_ootd.py --model_path <model-image-path> --cloth_path <cloth-image-path> --scale 2.0 --sample 4

2. Full-body model

Garment category must be paired: 0 = upperbody; 1 = lowerbody; 2 = dress

cd OOTDiffusion/run
python run_ootd.py --model_path <model-image-path> --cloth_path <cloth-image-path> --model_type dc --category 2 --scale 2.0 --sample 4

Train

accelerate launch ootd_train.py --load_height 512 --load_width 384 --dataset_list 'train_pairs.txt' --dataset_mode 'train' --batch_size 16 --train_batch_size 16 --num_train_epochs 200

Citation

@article{xu2024ootdiffusion,
  title={OOTDiffusion: Outfitting Fusion based Latent Diffusion for Controllable Virtual Try-on},
  author={Xu, Yuhao and Gu, Tao and Chen, Weifeng and Chen, Chengcai},
  journal={arXiv preprint arXiv:2403.01779},
  year={2024}
}

TODO List

• Paper
• Gradio demo
• Inference code
• Model weights
• Training code