KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#typescript #ai-agents #developer-tools #cli #codex #claude-code #react #mcp #open-source #deepseek-harness #dsh-plugin #ai

MagicFace

★ 62 Open GitHub ↗

Official implementation of MagicFace: High-Fidelity Facial Expression Editing with Action-Unit Control

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# MagicFace
Official implementation of MagicFace

> **MagicFace: High-Fidelity Facial Expression Editing with Action-Unit Control** [[arXiv paper](http://arxiv.org/abs/2501.02260)]<br>
> Mengting Wei, Tuomas Varanka, Xingxun Jiang, Huai-Qian Khor, Guoying Zhao<br>
> University of Oulu

## Introduction
We address the problem of facial expression editing
by controling the relative variation of facial action-unit (AU) from
the same person. This enables us to edit this specific person’s expression in a fine-grained, continuous and interpretable manner,
while preserving their identity, pose, background and detailed
facial attributes. By injecting AU variations
into a denoising UNet, our model can animate arbitrary identities
with various AU combinations, yielding superior results in high-fidelity expression editing compared to other facial expression
editing works.

### Dependencies

- Python 3.10
- Your computer should have a graphics card with approximately **8GB** to support running this test.

### Installation

You can first create a new Python 3.10 environment using `conda` and then install this package using `pip` from the PyPI hub:

```console
conda create -n magicface python=3.10
conda activate magicface
pip install torch==2.1.1 torchvision==0.16.1 torchaudio==2.1.1 --index-url https://download.pytorch.org/whl/cu118
pip install -r requirements.txt
```

### Download Models

You can directly retrieve our model from [HuggingFace](https://huggingface.co/mengtingwei/magicface/tree/main).

Or download using python script:

```python
# Download a specific file
from huggingface_hub import hf_hub_download
hf_hub_download(repo_id="mengtingwei/magicface", filename="79999_iter.pth", local_dir="./utils")
# Download all files 
from huggingface_hub import snapshot_download
snapshot_download(repo_id="mengtingwei/magicface", local_dir="./")
```

### Usage

#### Using our examples

You can test our model by editing the images we provided. Model inference needs an identity image
to edit, a background image for attribute condition and an AU condition.

> This script will directly download the model by the model cards of Huggingface, so the first time testing the download may take a lot of time.

* Test the model:

```--au_test``` The AUs you want to modify for the face. We provide 12 editable AUs here.
They are _AU1, AU2, AU4, AU5, AU6, AU9, AU12, AU15, AU17, AU20, AU25, AU26_. Only provide the AUs 
you want to modify here and split them by ``+``. For example, the following example shows how to 
edit AU1 and AU4. If only one AU is intended to change, just provide that one.

```--AU_variation```  Intensity integers you want to edit for each AU you specified. Also split them by ``+`` 
if multiple AUs are intended to change. We recommend to limit the intensity 
in the range of [-10, 10]. Integers outside this range may experience severe distortion.

```console
python inference.py --img_path './test_images/00381.png' --bg_path './test_images/00381_bg.png' --au_test 'AU4+AU1' --AU_variation '4+2'
```

#### Test your own images

If you want to edit your own images, you need to compute the 
background and pose for attribute condition.

1. Check if you have downloaded all required files and saved in the ``utils`` directory. It should be 
like this:

```
utils
    79999_iter.pth
    third_party
        ...
    third_party_files
        ...
    checkpoints
        ...
```

2. Crop your image into the resolution of 512 $\times$ 512. Please provide
the image including at least one face, otherwise it will result in an error.
```console
cd utils
python preprocess.py --img_path <your-image-path> --save_path <your-save-path>
```

3. Then parse the background and draw the contour from the cropped image.

```console
python retrieve_bg.py --img_path <your-cropped-path> --save_path <your-save-path>
```
4. Use the `inference.py` script introduced above to test your image.
### Issues or Questions?
If the issue is code-related, please open an issue here.

For questions, please also consider opening an issue as it may benefit future reader. 
Otherwise, email Mengting Wei at [mengting.wei@oulu.fi](mengting.wei@oulu.fi).

### Acknowledgements

This codebase was built upon and drew inspirations from [FineFace](https://github.com/tvaranka/fineface), [InsightFace](https://github.com/deepinsight/insightface),
[BiSeNet](https://github.com/zllrunning/face-parsing.PyTorch) and [Stable DIffusion](https://github.com/CompVis/stable-diffusion).

We thank the authors for making those repositories public.

MagicFace

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README