Two-Stream Manipulation Trace Network for Face Forgery Detection

In this paper, we propose a Two-Stream Manipulation Trace Network to learn deep fake methods for subtle manipulation traces of facial regions to address the lack of robust fake face detectors in complex scenes. It consists of two key modules: the spatial domain manipulation trace extraction module and the frequency domain manipulation trace extraction module. The spatial domain manipulation trace extraction module uses convolutional neural networks to learn subtle manipulation traces in the image space. The frequency domain manipulation trace extraction module learns the characteristics of high-frequency information in the image frequency domain. The reason is that facial forgery regions typically exist at the edges and details of the face, which corresponds to the high-frequency domain of the face frequency information. We also designed a feature fusion module combined with a self-attention mechanism. This module fuses the manipulation traces in the spatial and frequency domains to generate the final features to be classified. Experiments demonstrate that our model outperforms the state-of-the-art models and has better robustness and generalization ability on commonly used video datasets.