Kuaishou Tech
Apr 6, 2021 · Artificial Intelligence
Frequency-Aware Feature Learning with Single-Center Loss for Face Forgery Detection
Researchers from USTC and Kuaishou propose a frequency‑aware feature learning framework that combines a data‑driven adaptive frequency module with a novel single‑center loss, achieving state‑of‑the‑art performance on deepfake detection while addressing class‑distribution challenges.
AI securityDeepfake Detectioncomputer vision
0 likes · 7 min read