Most of the studies on emotion recognition problem are focused on single-channel recognition or multimodal approaches when the data is available for the whole dataset. However, in some practical cases data sources could be missed, noised or broken.

We presented the first machine learning competition on multimodal emotion recognition with missing data. The main goal of this challenge is to find approaches for a reliable recognition of emotional behavior when some data is unavailable.

The task consisted in predicting one of the six basic emotions (happiness, sadness, anger, disgust, fear and neutral state) based on the dataset of emotions acted by semi-professionals. Participants were presented with features for 4 modalities: audio, facial expressions, body-motion, and eye-tracking. They needed to beat the baseline solution based on naïve approach to compete for the prizes.

Competition Organizer and Sponsor — Neurodata Lab LLC, project company and R&D laboratory.