Face-trace pipeline enables open-set synthetic face generator attribution

Researchers have developed a new pipeline called Face-trace for open-set synthetic face source attribution. This method can identify the generator of a synthetic face image, even when the generator is unknown or has not been seen during training. It combines known generator classification with energy-based out-of-distribution rejection and a clustering approach to discover new generators. Experiments on the WILD dataset show high accuracy in both closed-set and open-set scenarios, with the ability to progressively discover and categorize unknown generators over time. AI

IMPACT Enhances multimedia forensics by enabling the identification of synthetic faces from previously unseen generative models.

RANK_REASON Research paper detailing a new method for synthetic face generator attribution. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Claudio Schiavella
• Face-trace
• Forensic Self-Descriptors
• I-JEPA
• WILD dataset

• paper
• other

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