A Temporal Consistency Learning Framework for Face Forgery Detection

The proliferation of face forgery techniques, particularly through deepfake videos, poses significant challenges in terms of deceiving and manipulating viewers. To address this concern, we propose a novel framework that combines temporal consistency learning with improved dense trajectories to achieve more accurate and robust face forgery detection. Our approach aims to enhance the precision and resilience of existing video-level detection methods by leveraging temporal information and thus improving the representation of facial geometric features. Experimental results demonstrate that our method achieves competitive results compared to state-of-the-art methods. Moreover, the framework we propose has the potential to enhance the performance of other detection methods on challenging manipulation methods.