New benchmark highlights face forgery detection robustness challenges

A new research paper introduces a benchmark for face forgery detection that accounts for realistic image degradations. The study evaluated six model families, including convolutional and transformer-based networks, as well as a frozen self-supervised DINOv3 backbone. Results indicate that models performing well on clean datasets often struggle with degraded images, with Xception achieving the best clean performance and frozen DINOv3 showing the most robustness under degradation. AI

IMPACT Highlights the need for more robust face forgery detection models that perform well under real-world image degradations.

RANK_REASON Research paper detailing a new benchmark and evaluation of existing models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• DINOv3
• Fourier
• MFFI
• RGB color model
• Xception

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →