Psychological research involving anthropometric measures or experimental manipulations of human faces has been used to test hypotheses related to the social perception of various attributes such as attractiveness, health, competence or trustworthiness based on the shape and appearance of a human face. A common and time-consuming step in such research is delineation, that is, the placement of fiducial markers on standardized and prominent anatomical features of a face. Traditionally, researchers have relied on human evaluators to perform this step. This paper examines the justification of using computer vision algorithms in delineating human faces. For this purpose, the standardized and research-validated Chicago Face Database was used, which, in addition to high-quality photographs of human faces of different ethnicities, also shows a number of anthropometric features measured by human evaluators in the validation data set. On the aforementioned database of photographs of human faces, delineation was performed using the Face++ computer vision service. The convergent validity of the computer algorithm’s estimates relative to human estimates was tested using a series of correlation coefficients. Results indicate that algorithmically implemented delineation was highly consistent with that performed by human evaluators, and the discussion argues for the applicability of these findings in psychological research practice.

human faces, computer vision, delineation, validation