Abstract4D: A Large-Scale Dataset and Framework for Understanding the Visual Language of Abstract Art

• URL: http://arxiv.org/abs/2608.28339v1
• Date: Fri, 28 Aug 2026 13:51:38 GMT
• Title: Abstract4D: A Large-Scale Dataset and Framework for Understanding the Visual Language of Abstract Art
• Abstract summary: Abstract4D is the largest dataset of abstract paintings to date.<n>We analyze the semantic structure of abstract art through large-scale embedding visualization.<n>We establish benchmark tasks for classification, cross-modal retrieval, and text-to-image generation.
• Score: 9.446232705010916
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Artificial intelligence can classify artistic styles and synthesize images, but it still lacks a model of the visual language that gives art meaning. Abstract painting minimizes object semantics and foregrounds structural cues, making it an ideal testbed for computational perception. We introduce \textbf{Abstract4D}, the largest dataset of abstract paintings to date: more than 120,000 images paired with rich metadata and multi-dimensional prompts that capture each work's perceptual attributes---\textit{form, color, texture, and composition}. Annotations are produced by a hybrid human--VLM pipeline for quality and consistency. Using Abstract4D, we (i) analyze the semantic structure of abstract art through large-scale embedding visualization, uncovering how perceptual relationships organize artistic meaning, and (ii) establish benchmark tasks for classification, cross-modal retrieval, and text-to-image generation to evaluate how AI models perceive and reproduce abstract visual language. Together, these analyses demonstrate how Abstract4D enables both exploration and quantitative assessment of AI's ability to represent and interpret abstract art.

Related papers

• Conducting Stylistic Analysis of Paintings through an Art-History Agent [0.0]
  We present an AI framework that automates stylistic analysis of paintings.<n>By training a vision transformer (ViT) on a large corpus of paintings with metadata, our system encodes this art history-specific data as embeddings.<n>A large language model (LLM) then interprets each feature by retrieving associated artworks and their accompanying curator-written texts, and synthesizes them into descriptions.
  arXiv  Detail & Related papers  (2026-08-30T08:10:58Z)
• Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment [51.40989269202702]
  aesthetic quality assessment task is crucial for developing a human-aligned quantitative evaluation system for AIGC.<n>We propose ArtQuant, an aesthetics assessment framework for artistic images which couples isolated aesthetic dimensions through description generation.<n>Our approach achieves epoch state-of-the-art performance on several datasets while requiring only 33% of conventional trainings.
  arXiv  Detail & Related papers  (2025-12-29T12:18:26Z)
• TraitSpaces: Towards Interpretable Visual Creativity for Human-AI Co-Creation [0.0]
  Drawing on interviews with practicing artists and theories from psychology, we define 12 traits that capture affective, symbolic, cultural, and ethical dimensions of creativity.<n>Traits such as Environmental Dialogicity and Redemptive Arc are predicted with high reliability.<n>By linking cultural-aesthetic insights with computational modeling, our work aims not to reduce creativity to numbers, but to offer shared language and interpretable tools for artists, researchers, and AI systems to collaborate meaningfully.
  arXiv  Detail & Related papers  (2025-09-29T06:24:18Z)
• An Exploratory Study on Abstract Images and Visual Representations Learned from Them [27.912006389955945]
  We study the reasons behind this performance gap and investigate how much high-level semantic content can be captured at different abstraction levels.<n>We then train and evaluate conventional vision systems on HAID across various tasks including classification, segmentation, and object detection.<n>We also discuss if the abstract image can be considered as a potentially effective format for conveying visual semantic information and contributing to vision tasks.
  arXiv  Detail & Related papers  (2025-09-17T16:30:34Z)
• Seeing the Abstract: Translating the Abstract Language for Vision Language Models [13.065703240655973]
  We focus our investigation on the fashion domain, a highly-representative field with abstract expressions.<n>By analyzing recent large-scale multimodal fashion datasets, we find that abstract terms have a dominant presence.<n>We propose a training-free and model-agnostic method, Abstract-to-Concrete Translator (ACT), to shift abstract representations towards well-represented concrete ones.
  arXiv  Detail & Related papers  (2025-05-06T07:14:10Z)
• Compositional Entailment Learning for Hyperbolic Vision-Language Models [54.41927525264365]
  We show how to fully leverage the innate hierarchical nature of hyperbolic embeddings by looking beyond individual image-text pairs.<n>We propose Compositional Entailment Learning for hyperbolic vision-language models.<n> Empirical evaluation on a hyperbolic vision-language model trained with millions of image-text pairs shows that the proposed compositional learning approach outperforms conventional Euclidean CLIP learning.
  arXiv  Detail & Related papers  (2024-10-09T14:12:50Z)
• What Makes a Maze Look Like a Maze? [92.80800000328277]
  We introduce Deep Grounding (DSG), a framework that leverages explicit structured representations of visual abstractions for grounding and reasoning.<n>At the core of DSG are schemas--dependency graph descriptions of abstract concepts that decompose them into more primitive-level symbols.<n>We show that DSG significantly improves the abstract visual reasoning performance of vision-language models.
  arXiv  Detail & Related papers  (2024-09-12T16:41:47Z)
• GalleryGPT: Analyzing Paintings with Large Multimodal Models [64.98398357569765]
  Artwork analysis is important and fundamental skill for art appreciation, which could enrich personal aesthetic sensibility and facilitate the critical thinking ability. Previous works for automatically analyzing artworks mainly focus on classification, retrieval, and other simple tasks, which is far from the goal of AI. We introduce a superior large multimodal model for painting analysis composing, dubbed GalleryGPT, which is slightly modified and fine-tuned based on LLaVA architecture.
  arXiv  Detail & Related papers  (2024-08-01T11:52:56Z)
• AbsPyramid: Benchmarking the Abstraction Ability of Language Models with a Unified Entailment Graph [62.685920585838616]
  abstraction ability is essential in human intelligence, which remains under-explored in language models. We present AbsPyramid, a unified entailment graph of 221K textual descriptions of abstraction knowledge.
  arXiv  Detail & Related papers  (2023-11-15T18:11:23Z)
• Text-to-Image Generation for Abstract Concepts [76.32278151607763]
  We propose a framework of Text-to-Image generation for Abstract Concepts (TIAC) The abstract concept is clarified into a clear intent with a detailed definition to avoid ambiguity. The concept-dependent form is retrieved from an LLM-extracted form pattern set.
  arXiv  Detail & Related papers  (2023-09-26T02:22:39Z)
• Systematic Visual Reasoning through Object-Centric Relational Abstraction [5.914610036560008]
  We introduce OCRA, a model that extracts explicit representations of both objects and abstract relations. It achieves strong systematic generalizations in tasks involving complex visual displays.
  arXiv  Detail & Related papers  (2023-06-04T22:47:17Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.