Creating Realistic Human Avatars for Social Virtual Environments Using Photographic Inputs

Abstract

This paper presents the development and evaluation of realistic virtual reality avatars created with a Blender add-on called Facebuilder. In this process, a person's head is photographed from different angles. These photographs are used in subsequent steps to generate a realistic avatar face. To investigate the user experience of interacting with these avatars, a study was conducted in VR using the MyScore application. The study involved 22 participants who met in a virtual environment to discuss a topic of their choice. Statistical analyses including descriptive statistics, Wilcoxon Signed-Rank Test, and Friedman Test showed significant differences supporting all three hypotheses: users preferred communicating with realistic avatars, were more focused and engaged when interacting with them. The results indicate a significant preference for realistic avatars in educational use cases, primarily due to the perceived seriousness of the interactions and the resulting higher level of participant engagement. The suitability of realistic versus non-realistic avatars was found to be use-case dependent. Participants suggested that realistic avatars would be more appropriate for educational scenarios and non-realistic avatars for entertainment.

Keywords

• virtual reality
• education
• open-source software
• photogrammetry
• avatar