Authors: Guanhua Zhao, Yu Gu, Xuhan Sheng, Yujie Hu, Jian Zhang

Published on: April 22, 2024

Impact Score: 7.4

Arxiv code: Arxiv:2404.14177

Summary

• What is new: Introduction of a comprehensive framework that not only detects facial retouching but also restores retouched images to their original state, including corrections for geometric deformations and color shifts.
• Why this is important: The challenge of authenticating facial photographs in critical areas like identity verification and social media, heightened by the easy creation of deceptive images through retouching tools.
• What the research proposes: A novel framework named Face2Face, which includes a facial retouching detector, an image restoration model (FaceR), and a color correction module (Hierarchical Adaptive Instance Normalization – H-AdaIN), to accurately restore retouched images.
• Results: Extensive experiments validate the effectiveness of the Face2Face framework and its components in accurately detecting and restoring retouched facial images.

Technical Details

Technological frameworks used: Face2Face

Models used: Facial retouching detector, FaceR, and Hierarchical Adaptive Instance Normalization (H-AdaIN)

Data used: nan

Potential Impact

Identity verification services, social media platforms, and digital forensic companies could benefit from the insights and technology developed in this paper.

Want to implement this idea in a business?

We have generated a startup concept here: Authentica.