Facial Ethnicity Identification by Collaborative Voting System

Ethnic groups share physical and genetic characteristics that allow them to be uniquely identified. The ethnic identification process is complex because it depends on several factors, such as cultural, geographic, economic, and social. A divergence between self-identification and external identification towards an individual could generate economic and social conflicts in minority ethnic groups, preventing them from exercising their right to self-determination. This research addresses ethnic identification using a set of facial photographs of the four majority ethnicities in Ecuador: Afro-Ecuadorian, Indigenous, European-descended, and Mestizos. From these photographs, a collaborative voting web system was developed using image retrieval information. With the data obtained from user interactions, a statistical analysis measures the divergence between self-perception and external perception of an individual’s ethnicity. The performance contrast of these two classification processes obtained an F1-score of 88.7% in the Afro-Ecuadorian ethnic group, unlike the indigenous ethnic group with 43.7%, thus showing that people classify certain ethnicities better based on a facial image. On the other hand, by ANOVA, it was determined that there is no significant difference in invariability in the perception of ethnic groups, providing a deeper understanding of the self-perception of Ecuadorian ethnicities.