DeepfakeNet, an Efficient Deepfake Detection Method

Different CNNs models do not perform well in deepfake detection cross datasets. This paper proposes a model called DeepfakeNet, which consists of 20 network layers. It refers to the stacking idea ResNet and split-transform-merge Inception design block structure, That is, structure ResNeXt. The study uses some data FaceForensics++, Kaggle TIMIT datasets, enhancement technology is used expand datasets for training testing models. experimental results show that, compared with current mainstream including VGG19, ResNet101, ResNeXt50, XceptionNet GoogleNet, same dataset preset parameters, proposed only has higher accuracy lower error rate detection, but also significant improvement performance.