Consistency of Virtual Human Faces: Effect of Stylization and Expressiveness Intensity on Character Perception

No Thumbnail Available
Date
2023
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
The perception of virtual human faces is impacted by several interrelated factors, such as shapes, materials, textures, shading or animation. Previous research demonstrated that animation flaws and inconsistencies tend to be less accepted on realistic characters compared to stylized ones. To further investigate the relationship between stylization and facial expressions, we designed an online investigation to assess participants' perception of virtual human faces with regards to perceived realism, attractiveness and eeriness. We modeled a female and a male character with three levels of stylization (stylized, semi-realistic, realistic) and four expressiveness intensities (idle, low, moderate and exaggerated). 135 participants took part in the experiment which consisted of watching a set of 24 videos lasting 30 seconds each. Interaction effects were observed between the independent variables. Overall, results indicated that perceived realism was mainly driven by the stylization level. Visually realistic characters were perceived as more realistic even with exaggerated facial expressions. In terms of attractiveness, scores tended to decrease as the intensity of facial expressions increased. At the highest expressiveness intensity, the semi-realistic female character was preferred over the stylized and the realistic ones, while the realistic male character scored higher than the others. In line with previous studies, results indicated that perceived eeriness rises with the intensities of facial expressions. Taken as a whole, this study provides valuable guidelines for 3D artists and developers to design avatars and autonomous agents.
Description

CCS Concepts: Applied computing → Psychology; Computing methodologies → Perception

        
@inproceedings{
10.2312:egve.20231314
, booktitle = {
ICAT-EGVE 2023 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual Environments
}, editor = {
Jean-Marie Normand
 and
Maki Sugimoto
 and
Veronica Sundstedt
}, title = {{
Consistency of Virtual Human Faces: Effect of Stylization and Expressiveness Intensity on Character Perception
}}, author = {
Dubosc, Charlotte
 and
Gorisse, Geoffrey
 and
Christmann, Olivier
 and
Richir, Simon
}, year = {
2023
}, publisher = {
The Eurographics Association
}, ISSN = {
1727-530X
}, ISBN = {
978-3-03868-218-9
}, DOI = {
10.2312/egve.20231314
} }
Citation
Collections