Detecting AI-generated Artwork
Meien Li, Mark Stamp
TL;DR
This work tackles the problem of distinguishing AI-generated artwork from human-created art across Baroque, Cubism, and Expressionism by evaluating classic ML models (LR, SVM, MLP) on a 39-feature hand-crafted set and comparing them to CNNs trained on raw images. Binary detection achieves high accuracy up to 0.9758, while multiclass classification across six styles reaches 0.8208, with feature reduction via Recursive Feature Elimination identifying compact feature subsets that preserve performance. The results show that feature-based detectors, especially SVM and MLP, are highly effective for binary classification and competitive for multiclass tasks, while CNNs underperform relative to the best feature-based methods in this dataset. The findings support practical AI-art detection and suggest a potential two-stage system, plus avenues for extending to more styles, features, and models.
Abstract
The high efficiency and quality of artwork generated by Artificial Intelligence (AI) has created new concerns and challenges for human artists. In particular, recent improvements in generative AI have made it difficult for people to distinguish between human-generated and AI-generated art. In this research, we consider the potential utility of various types of Machine Learning (ML) and Deep Learning (DL) models in distinguishing AI-generated artwork from human-generated artwork. We focus on three challenging artistic styles, namely, baroque, cubism, and expressionism. The learning models we test are Logistic Regression (LR), Support Vector Machine (SVM), Multilayer Perceptron (MLP), and Convolutional Neural Network (CNN). Our best experimental results yield a multiclass accuracy of 0.8208 over six classes, and an impressive accuracy of 0.9758 for the binary classification problem of distinguishing AI-generated from human-generated art.