Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/alibaba-yuanjing-aigclab/ViViD
ViViD: Video Virtual Try-on using Diffusion Models
https://github.com/alibaba-yuanjing-aigclab/ViViD
Last synced: 24 days ago
JSON representation
ViViD: Video Virtual Try-on using Diffusion Models
- Host: GitHub
- URL: https://github.com/alibaba-yuanjing-aigclab/ViViD
- Owner: alibaba-yuanjing-aigclab
- License: apache-2.0
- Created: 2024-05-20T02:09:51.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2024-06-21T04:16:28.000Z (7 months ago)
- Last Synced: 2024-10-19T20:54:14.681Z (3 months ago)
- Language: Python
- Size: 1.69 MB
- Stars: 440
- Watchers: 30
- Forks: 28
- Open Issues: 19
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-diffusion-categorized - [Code
README
# ViViD
ViViD: Video Virtual Try-on using Diffusion Models
[](https://arxiv.org/abs/2405.11794)
[](https://alibaba-yuanjing-aigclab.github.io/ViViD)
[](https://huggingface.co/alibaba-yuanjing-aigclab/ViViD)
## Dataset
Dataset released: [ViViD](https://huggingface.co/datasets/alibaba-yuanjing-aigclab/ViViD)
## Installation
```
git clone https://github.com/alibaba-yuanjing-aigclab/ViViD
cd ViViD
```
### Environment
```
conda create -n vivid python=3.10
conda activate vivid
pip install -r requirements.txt
```
### Weights
You can place the weights anywhere you like, for example, ```./ckpts```. If you put them somewhere else, you just need to update the path in ```./configs/prompts/*.yaml```.
#### Stable Diffusion Image Variations
```
cd ckpts
git lfs install
git clone https://huggingface.co/lambdalabs/sd-image-variations-diffusers
```
#### SD-VAE-ft-mse
```
git lfs install
git clone https://huggingface.co/stabilityai/sd-vae-ft-mse
```
#### Motion Module
Download [mm_sd_v15_v2](https://huggingface.co/guoyww/animatediff/blob/main/mm_sd_v15_v2.ckpt)
#### ViViD
```
git lfs install
git clone https://huggingface.co/alibaba-yuanjing-aigclab/ViViD
```
## Inference
We provide two demos in ```./configs/prompts/```, run the following commands to have a try😼.
```
python vivid.py --config ./configs/prompts/upper1.yaml
python vivid.py --config ./configs/prompts/lower1.yaml
```
## Data
As illustrated in ```./data```, the following data should be provided.
```text
./data/
|-- agnostic
| |-- video1.mp4
| |-- video2.mp4
| ...
|-- agnostic_mask
| |-- video1.mp4
| |-- video2.mp4
| ...
|-- cloth
| |-- cloth1.jpg
| |-- cloth2.jpg
| ...
|-- cloth_mask
| |-- cloth1.jpg
| |-- cloth2.jpg
| ...
|-- densepose
| |-- video1.mp4
| |-- video2.mp4
| ...
|-- videos
| |-- video1.mp4
| |-- video2.mp4
| ...
```
### Agnostic and agnostic_mask video
This part is a bit complex, you can obtain them through any of the following three ways:
1. Follow [OOTDiffusion](https://github.com/levihsu/OOTDiffusion) to extract them frame-by-frame.(recommended)
2. Use [SAM](https://github.com/facebookresearch/segment-anything) + Gaussian Blur.(see ```./tools/sam_agnostic.py``` for an example)
3. Mask editor tools.
Note that the shape and size of the agnostic area may affect the try-on results.
### Densepose video
See [vid2densepose](https://github.com/Flode-Labs/vid2densepose).(Thanks)
### Cloth mask
Any detection tool is ok for obtaining the mask, like [SAM](https://github.com/facebookresearch/segment-anything).
## BibTeX
```text
@misc{fang2024vivid,
title={ViViD: Video Virtual Try-on using Diffusion Models},
author={Zixun Fang and Wei Zhai and Aimin Su and Hongliang Song and Kai Zhu and Mao Wang and Yu Chen and Zhiheng Liu and Yang Cao and Zheng-Jun Zha},
year={2024},
eprint={2405.11794},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
## Contact Us
**Zixun Fang**: [[email protected]](mailto:[email protected])
**Yu Chen**: [[email protected]](mailto:[email protected])