MoE-FFD: Mixture of Experts for Generalized and Parameter-Efficient Face Forgery Detection

arXiv:2404.08452v1 Announce Type: new
Abstract: Deepfakes have recently raised significant trust issues and security concerns among the public. Compared to CNN face forgery detectors, ViT-based methods take advantage of the expressivity of transformers, achieving superior detection performance. However, these approaches still exhibit the following limitations: (1). Fully fine-tuning ViT-based models from ImageNet weights demands substantial computational and storage resources; (2). ViT-based methods struggle to capture local forgery clues, leading to model bias and limited generalizability. To tackle these challenges, this …

abstract arxiv cnn concerns cs.cv deepfakes detection detectors experts face fine-tuning forgery generalized however limitations mixture of experts moe performance public security security concerns transformers trust trust issues type vit