# Dimensional Distribution Emotion State: Leveraging Valence and Arousal as a Common Embedding Space for Visual Emotion Analysis

arXiv: https://arxiv.org/abs/2605.26262 · HTML: https://mlsanity.com/paper/2605.26262/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 13 of its 34 references that are in the map.
- First posted on arXiv 25 May 2026.

## Builds on (most cited first, 13 of 13)

- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [Dawn of the transformer era in speech emotion recognition: closing the valence gap](https://mlsanity.com/paper/2203.07378/) (2022)
- [ArtEmis: Affective Language for Visual Art](https://mlsanity.com/paper/2101.07396/) (2021)
- [EmoSet: A Large-scale Visual Emotion Dataset with Rich Attributes](https://mlsanity.com/paper/2307.07961/) (2023)
- [It is Okay to Not Be Okay: Overcoming Emotional Bias in Affective Image Captioning by Contrastive Data Collection](https://mlsanity.com/paper/2204.07660/) (2022)
- [Dimensional Emotion Detection from Categorical Emotion](https://mlsanity.com/paper/1911.02499/) (2019)
- [FindingEmo: An Image Dataset for Emotion Recognition in the Wild](https://mlsanity.com/paper/2402.01355/) (2024)
- [Affection: Learning Affective Explanations for Real-World Visual Data](https://mlsanity.com/paper/2210.01946/) (2022)
- [A Circular-Structured Representation for Visual Emotion Distribution Learning](https://mlsanity.com/paper/2106.12450/) (2021)
- [ArtELingo: A Million Emotion Annotations of WikiArt with Emphasis on Diversity over Language and Culture](https://mlsanity.com/paper/2211.10780/) (2022)
- [EEmo-Bench: A Benchmark for Multi-modal Large Language Models on Image Evoked Emotion Assessment](https://mlsanity.com/paper/2504.16405/) (2025)
- [EmoArt: A Multidimensional Dataset for Emotion-Aware Artistic Generation](https://mlsanity.com/paper/2506.03652/) (2025)
- [Training A Small Emotional Vision Language Model for Visual Art Comprehension](https://mlsanity.com/paper/2403.11150/) (2024)

## Abstract (arXiv)

Museums are important sites for the dissemination of culture and art. They are institutions rooted in history and tradition; their exhibitions are often designed to highlight these aspects. Recently, a new approach is being explored in the field: emotion-based exhibitions. These exhibitions are designed specifically to elicit emotions in the visitors, in order to maximize engagement, and as a way to democratize access to art and attract a wider, more diverse audience. To do so, the emotional content of the artworks must first be extracted, however, manually annotating the artworks by experts is a prohibitively labor-intensive process, and risks introducing the personal bias of curators. To assist the museum curators in their design of these exhibitions, we wish to develop a tool that can predict the emotional response evoked by a work of art. In this article, we leverage a continuous bi-dimensional emotion space to enhance emotion representations and the training process of deep learning models. Drawing inspiration from existing categorical and dimensional emotion representations, we introduce a new representation, Dimensional Distribution Emotion State (DDES), along with a pipeline for multi-dataset training. We show that DDES provides multiple advantages compared to widely used representations while exhibiting similar baseline performance.