Case Study

FaceReader is as good at recognizing facial expressions as humans.

In this study the FaceReader software was validated on two publicly available and objective datasets of human expressions of basic emotion (WSEFEP & ADFES). Facereader categorized 88% of the emotions accurately, which is similar to the human recognition scores (average of 85%). The table on the right shows the accuracy scores for each of the basic emotions.

Facereader even outperforms humans when categorising neutral faces, reducing a possible bias in the analysis. This study also indicated that action units from the Facial Action Coding System (FACS) can also be used with high accuracy (in the future action units will also be available in FaceReaderOnline).


Lewinski, P., den Uyl, T.M., Butler, C. (2014) Automated Facial Coding: Validation of Basic Emotions and FACS AUs in FaceReader. Journal of Neuroscience, Psychology, and Economics, 7(4), 227-236.

picture_as_pdfRead the Full Validation Paper