Performance of ResNet50 in Classifying Artificial Intelligence Generated Arts

The increasing prevalence of AI-generated art has sparked debates regarding its authenticity and impact on traditional art. This study evaluates the performance of the ResNet50 deep learning model in classifying images as either AI-generated or non-AI-generated (traditional art). A dataset collected from Kaggle was divided into two categories, and training was conducted using a learning rate of 0.001, a minibatch size of 32, and 20 epochs, with validation occurring every 50 iterations. The model achieved an overall accuracy of 95.45%, with class wise accuracies of 95.82% for Algenerated images and 95.03% for non-AI-generated images. Precision, recall, and F1 scores further confirmed the model's robust performance, with balanced metrics across both classes. The results highlight the effectiveness of ResNet50 in distinguishing between these two forms of art, demonstrating its potential as a tool for addressing controversies related to AI -generated content in art competitions and creative domains. Future research could focus on refining the model's performance by exploring more diverse datasets and advanced architectures. This study contributes to the growing body of research at the intersection of deep learning and creative industries, offering a reliable solution for classifying AI-generated and traditional art.