CLASSIFICATION OF REAL AND FAKE AI-GENERATED IMAGES USING CNN

Abstract

The rapid development of artificial intelligence (AI) in producing realistic images has raised concerns over the use of such technology. In this study, we give a detailed explanation of real and AI-generated photos to develop a classification model that can distinguish between the two. Our primary objective is to create a strong method to identify AI-generated images, which can have tremendous implications for various applications, including content moderation and image forensics. We create a dataset of autistic images from diverse sources and AI-generated images created with stateof-the-art generative models. To extract their distinct features, we extract different features from these images, such as color histograms, texture features, and deep neural network features. We train and validate the convolutional neural network (CNN) machine learning model to classify the images. Our results affirm the effectiveness of the CNN model, with a 95.5% accuracy in detecting real and fake photos.