Enhancing Authorship Attribution with Synthetic Paintings
2026-03-04 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors looked at how to tell who painted a picture using computers, but there aren't many real paintings to train these computer models. They tried making fake paintings using a tool called Stable Diffusion to help train better models. Their tests showed that mixing real and fake paintings helped the models get better at guessing the artist. This approach could help identify art styles even when there isn't much real artwork available.
authorship attributionStable DiffusionDreamBoothsynthetic dataclassification modelsROC-AUCcomputer visionartwork authenticationgenerative modelsdiscriminative models
Authors
Clarissa Loures, Caio Hosken, Luan Oliveira, Gianlucca Zuin, Adriano Veloso
Abstract
Attributing authorship to paintings is a historically complex task, and one of its main challenges is the limited availability of real artworks for training computational models. This study investigates whether synthetic images, generated through DreamBooth fine-tuning of Stable Diffusion, can improve the performance of classification models in this context. We propose a hybrid approach that combines real and synthetic data to enhance model accuracy and generalization across similar artistic styles. Experimental results show that adding synthetic images leads to higher ROC-AUC and accuracy compared to using only real paintings. By integrating generative and discriminative methods, this work contributes to the development of computer vision techniques for artwork authentication in data-scarce scenarios.