Classification of Real images from DeepFakes

Abstract

The progress of Artificial Intelligence has brought along certain disadvantages primarily generation of Deepfakes that are being circulated through various channels of communication. These images are impossible to differentiate from real images due to their exceptional quality. This project is aimed at classifying real images from AI generated images (generated using diffusion models). This paper will draw a comparative study between various standard classifiers such as Gradient Boosting, K Nearest Neighbor, Random forest and Neural Network to find out which algorithm is best suited for this type of classification task.