AffectiveArt Challenge 2026: Fine-Grained Emotion Understanding and Generation in Artistic Images

Hongxia Xie, Cheng Zhang, Wen-Huang Cheng, Ling Lo, Hong-Han Shuai, Jianlong Fu, Sicheng Zhao, Sanghoon Lee, Xing Huang

Keywords: Multimodal Foundation Models, Emotional Art Image Generation, Emotional Art Image Understanding

Abstract: Recent advances in AI-generated content (AIGC), driven by diffusion models and multimodal foundation models, have significantly improved visual realism. However, current systems remain limited in modeling the fine-grained emotional expression and abstract visual language inherent in artistic imagery. Unlike photorealistic generation, artistic creation requires coherent integration of style, brushwork, color composition, and affective intent. To address this gap, we propose the EmoArt 2026 Grand Challenge, built upon a large-scale emotion-aware art dataset containing 132,664 artworks across 56 artistic styles with structured affective annotations. The challenge consists of two complementary tracks: (1) Emotion-Aware Artistic Image Generation, focusing on emotionally and stylistically coherent synthesis; and (2) Multidimensional Art Emotion Understanding, targeting fine-grained affect recognition in artistic imagery. We introduce a rigorous evaluation protocol combining automatic metrics, including the proposed Attribute Alignment Score (AAS), and expert-based human aesthetic assessment. A hidden test set will be securely maintained to ensure fair benchmarking and prevent data leakage. The challenge will be co-located with the MUSE 2026 Workshop on Multimodal Understanding and Synthesis of Emotion in Art at ACM Multimedia 2026, fostering strong interaction between benchmark-driven evaluation and in-depth academic discussion. EmoArt 2026 aims to establish a sustainable benchmark for advancing affective computing and emotion-aware artistic AI.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 25