Computer Science38th Conference on Neural Information Processing Systems (NeurIPS 2024)
To Err Like Human: Affective Bias-Inspired Measures for Visual Emotion Recognition Evaluation
C. Zhao, J. Shi, et al.
Accuracy treats all errors equally — but emotions don’t. This paper introduces a novel evaluation metric for visual emotion recognition based on Mikel’s emotion wheel: misclassifications are weighted by psychological distance so confusing 'excitement' with 'anger' is penalized more than with 'awe'. Semi-supervised experiments and a user study show the metric better aligns with human cognitive laws than plain accuracy. Code: https://github.com/ZhaoChenxi-nku/ECC. Research conducted by Chenxi Zhao, Jinglei Shi, Liqiang Nie, and Jufeng Yang.
Related Publications
Explore these studies to deepen your understanding
Adjacent work that informs or extends this paper's methodology and findings.
Computer Science
The paradox of creativity in generative AI: high performance, human-like bias, and limited differential evaluation
J. Desdevises
Computer Science
On Multimodal Emotion Recognition for Human-Chatbot Interaction in the Wild
N. Kovačević, M. Gross, et al.
Computer Science
The paradox of creativity in generative AI: high performance, human-like bias, and limited differential evaluation
J. Desdevises
Engineering and Technology
A star-nose-like tactile-olfactory bionic sensing array for robust object recognition in non-visual environments
M. Liu, Y. Zhang, et al.
