TVN: Detect Deepfakes Images using Texture Variation Network

Face manipulation technology is rapidly developing, making it impossible for human eyes to recognize fake face photos. Convolutional Neural Network (CNN) discriminators, on the other hand, can fast achieve high accuracy in distinguishing fake/real In this paper, we look at how CNN models discern between and real faces. forgery detection relies heavily Texture Variation (TVN) information, according our findings. We propose a new model, TVN, robust fraud detection, based Convolution pyramid pooling (PP), as result of aforesaid discovery. To produce stationary representation composition difference combines pixel intensity gradient information. Simultaneously, multi-scale information fusion PP prevent texture features from being destroyed. Our TVN beats previous techniques numerous databases, including Faceforensics++, DeeperForensics-1.0, Celeb-DF, DFDC. The more resistant image distortion, such JPEG compression blur, which critical wild.