Method and system for detecting face forgery of double-stream video based on multiple clues

The invention provides a double-stream video face counterfeiting detection method and system based on multiple clues, and the method comprises the steps: inputting a to-be-detected video stream into a pre-trained multi-clue video counterfeiting detection model, and obtaining a face authenticity classification detection result; the detection model is obtained by carrying out parallel interactive fusion on the basis of an OfficientNet-B5 network and a Swin Transform network to form multiple clues and carrying out training on a forged video training data set. According to the invention, by using a combined clue of high-frequency information, low-level texture and optical flow information in a video image frame, and fusing a local feature extraction capability of an EfficentNet-B5 network and a global relationship perception capability of a Swin Transformer network, when distinguishing whether a face image in the video frame is true or false, a more excellent classification performance is embodied, and the face i