README.md

---
license: cc-by-nc-sa-4.0
extra_gated_prompt: |-
  Before gaining access to all the FaceDancer models and weights you must agree to the terms.
   The models is provided for open access under the CreativeCommons license: cc-by-nc-sa-4.0. is open access and available to all. There are a few key aspects that we need   
   to emphasise AND you must also comply with below terms  
   before gaining access:
   
     1. You can not use the model to deliberately produce nor share illegal or harmful outputs or content.
     2. You are accountable for their use and must not go against the license and the terms described in this agreement form.
     3. The models and weights provided can not be used commercially. This includes produced images using the model.
     4. If you use any of the provided models in the work you must cite this work:
     @inproceedings{Rosberg2023FaceDancer,
     title     = {FaceDancer: Pose- and Occlusion-Aware High Fidelity Face Swapping},
     author    = {F. Rosberg, E. Aksoy. C. Englund, F. Alonso-Fernandez},
     booktitle = {Proc. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
     year      = {2023}}
     5. You agree to above terms for any future models that is uploaded and/or published to this model card.
     
   By clicking on "Access repository" below, you accept that your *contact information* (email address and username) can be shared with the model authors as well. Make sure 
   to fill in the information about your intended use. Be clear and expect a matching email to the mentioned affiliation.
    
extra_gated_fields:
 Company / Affiliation: text
 Company / Affiliation Email: text
 Country: country
 I want to use this model for (A brief description): text
 I have read the License and the terms mentioned above, and agree with their terms: checkbox
extra_gated_description: "We process requests manually, it may take 2-7 days. Longer process time may occur."
---
# FaceDancer: Pose- and Occlusion-Aware High Fidelity Face Swapping

![demo_vid_0](133_to_4.gif)

Here you can find all the FaceDancer models from our work [FaceDancer: Pose- and Occlusion-Aware High Fidelity Face Swapping](https://arxiv.org/abs/2210.10473).
You can find the specification for most models in the paper. We may publish further models that is not specificed in the paper in the future. They will be detailed here in such case.
Currently there is one extra model, FaceDancer_config_C_HQ which uses same configuration as Config C detailed in the paper, but it is trained on high resolution images only.
This model only really produce high resolution faces, so using it on low resolution or distorted images will make the manipulation visible.

You can try out FaceDancer for images [here on Hugging Face](https://huggingface.co/spaces/felixrosberg/face-swap).

The source code can be found [here](https://github.com/felixrosberg/FaceDancer). Including script for face swapping for video.

![result_matrix](result_matrix.png)

## Citation
If you use this repository in your work, please cite us:
```
@InProceedings{Rosberg_2023_WACV,
    author    = {Rosberg, Felix and Aksoy, Eren Erdal and Alonso-Fernandez, Fernando and Englund, Cristofer},
    title     = {FaceDancer: Pose- and Occlusion-Aware High Fidelity Face Swapping},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2023},
    pages     = {3454-3463}
}
```