Video as Natural Augmentation: Towards Unified AI-Generated Image and Video Detection
Researchers have developed VINA, a new framework designed to detect both AI-generated images and videos. VINA addresses a key limitation where image detection models often fail when applied to video frames due to processing shifts like compression and resizing. By training on both image and video data and employing a cross-modal contrastive objective, VINA improves detection accuracy and robustness across various benchmarks. AI
IMPACT Enhances the ability to detect synthetic media, crucial for combating misinformation and ensuring authenticity.