Artificial Intelligence for Geometry-Based Feature Extraction, Analysis and Synthesis in Artistic Images: A Survey

TL;DR

The paper surveys how incorporating geometric data improves AI tasks in artistic images, covering extraction, discriminative analysis, and synthesis. It categorizes methods by geometric feature types (objects, humans, segmentation, 3D) and representations (implicit, explicit, parametric), and evaluates their effectiveness across detection, style/scene classification, and perceptual studies. It also reviews synthesis tasks (style transfer, inpainting, novel-view, relighting, remodeling) and how geometry acts as a prior or conditioning signal to preserve structure and improve realism, with metrics and user studies. Limitations include dataset fragmentation and lack of standard benchmarks, while futures point to automatic annotations, cross-correspondences, controlled guidance, and geometry-aware object embeddings to enable robust, controllable AI for visual arts. Overall, geometry-guided AI has strong potential to enhance accuracy, realism, and adaptability in art analysis and generation, with meaningful applications in conservation, curation, and creative tooling.

Abstract

Artificial Intelligence significantly enhances the visual art industry by analyzing, identifying and generating digitized artistic images. This review highlights the substantial benefits of integrating geometric data into AI models, addressing challenges such as high inter-class variations, domain gaps, and the separation of style from content by incorporating geometric information. Models not only improve AI-generated graphics synthesis quality, but also effectively distinguish between style and content, utilizing inherent model biases and shared data traits. We explore methods like geometric data extraction from artistic images, the impact on human perception, and its use in discriminative tasks. The review also discusses the potential for improving data quality through innovative annotation techniques and the use of geometric data to enhance model adaptability and output refinement. Overall, incorporating geometric guidance boosts model performance in classification and synthesis tasks, providing crucial insights for future AI applications in the visual arts domain.

Figures (1)

• Figure 1: T-SNE visualization of the art domain as compared to the real-world images using the PACS dataset. The art modalities with paintings, cartoons and sketches showcase the clustering of art modalities close to photos, exaggerated geometries and no color or texture respectively. The dimensionality reduction uses a pre-trained VGG-19 model as a feature extractor and removes the fully connected head.