Perception Bias of Human and Machine Over Cross Cultural Facial Expression Images

Résumé fourni par la source

This paper explores the perception characteristics of human and machine in interpreting cross-cultural facial expressions. Through a comparative analysis of Thai and Japanese samples, this research investigates the differences in how humans from different cultures and Facial Expression Recognition (FER) systems perceive and interpret crosscultural facial expressions, identifying the cultural biases present in both. Our study included subjective evaluations from participants in Thailand and Japan, assessing facial expression images from both their own and another culture, followed by a comparison with machine-based assessments. Furthermore, the visual elements surrounding the faces (context) were also included in the subjective evaluations to confirm the influence on recognition capability. The findings reveal that human and machine perception exhibit remarkably similar biases, suggesting that FER systems may mirror the cultural tendencies of human observers.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.