Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ArtBrain: An Explainable end-to-end Toolkit for Classification and Attribution of AI-Generated Art and Style

Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, Golnaz Shahtahmassebi, Jordan J. Bird

TL;DR

The paper tackles the problem of authenticating AI-generated art and tracing its source diffusion model. It introduces the AI-ArtBench dataset and an AttentionConvNeXt-based detector, packaged in the ArtBrain web app, and validates its effectiveness through high attribution accuracy ($\approx 0.999$) and strong classification performance. An Artistic Turing Test shows the model markedly surpasses humans in identifying AI art, underscoring practical value for provenance and attribution. Collectively, the work provides a scalable, explainable toolkit for AI art detection, attribution, and user-facing verification with open data and code to support further research.

Abstract

Recently, the quality of artworks generated using Artificial Intelligence (AI) has increased significantly, resulting in growing difficulties in detecting synthetic artworks. However, limited studies have been conducted on identifying the authenticity of synthetic artworks and their source. This paper introduces AI-ArtBench, a dataset featuring 185,015 artistic images across 10 art styles. It includes 125,015 AI-generated images and 60,000 pieces of human-created artwork. This paper also outlines a method to accurately detect AI-generated images and trace them to their source model. This work proposes a novel Convolutional Neural Network model based on the ConvNeXt model called AttentionConvNeXt. AttentionConvNeXt was implemented and trained to differentiate between the source of the artwork and its style with an F1-Score of 0.869. The accuracy of attribution to the generative model reaches 0.999. To combine the scientific contributions arising from this study, a web-based application named ArtBrain was developed to enable both technical and non-technical users to interact with the model. Finally, this study presents the results of an Artistic Turing Test conducted with 50 participants. The findings reveal that humans could identify AI-generated images with an accuracy of approximately 58%, while the model itself achieved a significantly higher accuracy of around 99%.

ArtBrain: An Explainable end-to-end Toolkit for Classification and Attribution of AI-Generated Art and Style

TL;DR

The paper tackles the problem of authenticating AI-generated art and tracing its source diffusion model. It introduces the AI-ArtBench dataset and an AttentionConvNeXt-based detector, packaged in the ArtBrain web app, and validates its effectiveness through high attribution accuracy () and strong classification performance. An Artistic Turing Test shows the model markedly surpasses humans in identifying AI art, underscoring practical value for provenance and attribution. Collectively, the work provides a scalable, explainable toolkit for AI art detection, attribution, and user-facing verification with open data and code to support further research.

Abstract

Recently, the quality of artworks generated using Artificial Intelligence (AI) has increased significantly, resulting in growing difficulties in detecting synthetic artworks. However, limited studies have been conducted on identifying the authenticity of synthetic artworks and their source. This paper introduces AI-ArtBench, a dataset featuring 185,015 artistic images across 10 art styles. It includes 125,015 AI-generated images and 60,000 pieces of human-created artwork. This paper also outlines a method to accurately detect AI-generated images and trace them to their source model. This work proposes a novel Convolutional Neural Network model based on the ConvNeXt model called AttentionConvNeXt. AttentionConvNeXt was implemented and trained to differentiate between the source of the artwork and its style with an F1-Score of 0.869. The accuracy of attribution to the generative model reaches 0.999. To combine the scientific contributions arising from this study, a web-based application named ArtBrain was developed to enable both technical and non-technical users to interact with the model. Finally, this study presents the results of an Artistic Turing Test conducted with 50 participants. The findings reveal that humans could identify AI-generated images with an accuracy of approximately 58%, while the model itself achieved a significantly higher accuracy of around 99%.

Paper Structure

This paper contains 34 sections, 8 equations, 14 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Using AI to Analyse Art
  4. Can AI Generate art?
  5. Who owns AI Art?
  6. Theory and Design
  7. Data Generation using Diffusion Models
  8. Detection Model Design
  9. Attention Module
  10. Implementation
  11. Novel Dataset: AI-ArtBench
  12. Latent Diffusion
  13. Stable Diffusion
  14. CNN Model Development
  15. Preprocessing
  16. ...and 19 more sections

Figures (14)

  • Figure 1: 'AttentionConvNeXt’ model architecture design.
  • Figure 2: Attention Module Architecture.
  • Figure 3: Image counts in each class.
  • Figure 4: Sample of the dataset representing each style generated using each source. Rows (top to bottom): Latent Diffusion, Standard Diffusion, Human. Columns (left to right): Art Nouveau, Baroque, Expressionism, Impressionism, Post impressionism, Realism, Renaissance, Romanticism, Surrealism, Ukiyo-e.
  • Figure 5: Preprocessed image Vs Original image (Starry Night by Vincent-van-Gogh, 1889).
  • ...and 9 more figures