A Comparison of Image Classification Techniques on Real and Artificially Generated Art

Artificial image generation has become more common and sophisticated in recent years. These computer generated images cause issues for artists looking for reference material with search engines and increase the potential for misinformation spread on social media.  This case study created three image classifiers to distinguish between real and AI-generated artworks using convolutional neural networks (CNN), support vector machines (SVM), and k-nearest neighbors (KNN) for the machine learning models. The project objectives included creating three modeling pipelines with Python in Jupyter notebooks, assessing model performance with common classification metrics, creating visuals to help non-technical users interpret model performance, and recommending models based on varying business needs. The models included preprocessing techniques and feature detection methods found in research related to image classification problems.