Measuring Emotion Intensity: Evaluating Resemblance in Neural Network Facial Animation Controllers and Facial Emotion Corpora

Sheldon Schiffer

2023

Abstract

Game developers must increasingly consider the degree to which animation emulates the realistic facial expressions found in cinema. Employing animators and actors to produce cinematic facial animation by mixing motion capture and hand-crafted animation is labour intensive and costly. Neural network controllers have shown promise toward autonomous animation that does not rely on pre-captured movement. Previous work in Computer Graphics and Affective Computing has shown the efficacy of deploying emotion AI in neural networks to animate the faces of autonomous agents. However, a method of evaluating resemblance of neural network behaviour in relation to a live-action human referent has yet to be developed. This paper proposes a combination of statistical methods to evaluate the behavioural resemblance of a neural network animation controller and the single-actor facial emotion corpora used to train it.

Paper Citation

in Harvard Style

Schiffer S. (2023). Measuring Emotion Intensity: Evaluating Resemblance in Neural Network Facial Animation Controllers and Facial Emotion Corpora. In Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2023) - Volume 2: HUCAPP; ISBN 978-989-758-634-7, SciTePress, pages 160-168. DOI: 10.5220/0011655300003417

in Bibtex Style

@conference{hucapp23,
author={Sheldon Schiffer},
title={Measuring Emotion Intensity: Evaluating Resemblance in Neural Network Facial Animation Controllers and Facial Emotion Corpora},
booktitle={Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2023) - Volume 2: HUCAPP},
year={2023},
pages={160-168},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011655300003417},
isbn={978-989-758-634-7},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2023) - Volume 2: HUCAPP
TI - Measuring Emotion Intensity: Evaluating Resemblance in Neural Network Facial Animation Controllers and Facial Emotion Corpora
SN - 978-989-758-634-7
AU - Schiffer S.
PY - 2023
SP - 160
EP - 168
DO - 10.5220/0011655300003417
PB - SciTePress