CompoVista: A Composition-Graph-Based Visual Analytics System for Compositional Analysis of Traditional Chinese Paintings

CompoVista introduces a composition-graph approach for analyzing Traditional Chinese paintings, splitting each work into four layers: entities, relations, voids, and context. That structure feeds a canvas-based visual analytics system that lets users build cohorts with compositional queries instead of relying only on manual inspection.

The practical shift here is from one-off interpretation to collection-scale comparison. Art historians can inspect entity distributions, compare compositional differences across groups, and trace aggregate patterns back to individual paintings for evidence. The system was developed with two art historians and informed by a literature review, which gives it a stronger grounding than a generic visualization prototype.

The paper also reports two case studies, a user study, and expert interviews, all aimed at showing that the workflow helps discover and validate patterns across large TCP collections. While the domain is art history, the underlying problem will feel familiar to game developers working with concept art archives, reference libraries, or any visual asset set where metadata alone is too thin.

For teams building tools, the interesting takeaway is the combination of a domain-specific representation with interactive cohort analysis. That’s the kind of pattern that could translate to production art review, style consistency checks, or large-scale asset exploration if adapted carefully.