The 2nd Korean Emotion Recognition Challenge: Methods and Results

Songa Kim, Van Thong Huynh, Dung Tran Thi, Aran Oh, Guee-Sang Lee, Hyung-Jeong Yang, Soo-Hyung Kim

Abstract: The $$2^{nd}$$ Korean Emotion Recognition Challenge (KERC2020) is a global challenge to promote the emotion recognition technologies by using audio-visual data analysis, especially for the emotion of Korean people. KERC2020 comprise of 1236 videos with each length from two to four seconds based on Korean movies are dramas. Around 68 participating teams compete to achieve state-of-the-art in recognizing stress, arousal, valence from Korean video in the wild. This paper provides a summary of dataset, methods and results in the challenge.