Convolutional Neural Networks for Fake Images Classification

Abstract

An important generative model, known as the Generative Adversarial Network (GAN), can be found in a wide variety of fields. Recent studies have indicated that it is possible to obtain fake face images with a Based on this new model, the images are of high quality. If those fake faces are abused in image tampering, it would cause some potential moral, ethical and legal snags. Consequently, we first propose in this piece: machine learning algorithm that uses a Convolutional Neural Network (CNN) Fake portraits of people created using the most up-to-date technology [12] provide experimental evidences to show that the proposed method having an average accuracy of more than 90%, can produce acceptable results 99.4 percent of the time. Our comparison results are based on the following criteria: featuring a number of variations on the CNN architecture high pass filter, the number of the layer groups and the activation function, in order to further demonstrate the rationality of our approach.