Face Anonymization Made Simple

Han-Wei Kung,Tuomas Varanka,Sanjay Saha,Terence Sim,Nicu Sebe

2024-11-02

Abstract:Current face anonymization techniques often depend on identity loss calculated by face recognition models, which can be inaccurate and unreliable. Additionally, many methods require supplementary data such as facial landmarks and masks to guide the synthesis process. In contrast, our approach uses diffusion models with only a reconstruction loss, eliminating the need for facial landmarks or masks while still producing images with intricate, fine-grained details. We validated our results on two public benchmarks through both quantitative and qualitative evaluations. Our model achieves state-of-the-art performance in three key areas: identity anonymization, facial attribute preservation, and image quality. Beyond its primary function of anonymization, our model can also perform face swapping tasks by incorporating an additional facial image as input, demonstrating its versatility and potential for diverse applications. Our code and models are available at <a class="link-external link-https" href="https://github.com/hanweikung/face_anon_simple" rel="external noopener nofollow">this https URL</a> .

Computer Vision and Pattern Recognition,Cryptography and Security

What problem does this paper attempt to address?

The paper attempts to address several major issues with current facial anonymization techniques:

1. **Dependence on facial recognition models**: Existing facial anonymization methods often rely on facial recognition models to compute identity loss, which can be inaccurate and unreliable.
2. **Need for additional data**: Many methods require additional data, such as facial landmarks and masks, to guide the image synthesis process, increasing complexity and dependency.
3. **Image quality degradation**: Traditional methods (such as blurring and pixelation) can protect privacy but significantly reduce image quality and are not flexible enough when dealing with key areas.
4. **Difficulty in retaining facial attributes**: Existing anonymization methods often struggle to retain important details such as facial expressions, head poses, and eye directions while protecting privacy.

To address these issues, the paper proposes a facial anonymization method based on a diffusion model. This method uses only reconstruction loss, does not require facial landmarks or masks, and can generate high-quality images with fine details, achieving state-of-the-art performance in the following three key areas:
- **Identity anonymization**
- **Facial attribute retention**
- **Image quality**

Additionally, the model has the capability to perform face swapping tasks, demonstrating its diverse application potential by introducing additional facial images as input.