Two-Stage Recognition and beyond for Compound Facial Emotion Recognition

Facial emotion recognition is an inherently complex problem due to individual diversity in facial features and racial cultural differences. Moreover, expressions typically reflect the mixture of people’s emotional statuses, which can be expressed using compound emotions. Compound makes even more difficult because discrimination between dominant complementary emotions usually weak. We have created a database that includes 31,250 images with different 115 subjects whose gender distribution almost uniform address recognition. In addition, we organized competition based on proposed dataset, held at FG workshop 2020. This paper analyzes winner’s approach—a two-stage method (1st stage, coarse recognition; 2nd fine recognition), enhances classification symmetrical labels.