Classifying Real and AI-Generated Images Using Fine-Tuned ResNet50

Abstract

Due to advancements in technology, especially in Artificial Intelligence, it has become increasingly difficult for humans to distinguish between real and AI-generated images. Technologies like Generative Adversarial Networks (GANs) and Latent Diffusion Models (LDMs) can generate images so realistic that humans often cannot identify them as AI-generated. This paper addresses the challenge of identifying AI-generated and real images by presenting a Fine-Tuned ResNet50 model. The model is trained and tested on a dataset of 140K real and fake face images, demonstrating its effectiveness in image classification.