Excited to announce MuSe2021!

Baselines + Winner

Baselines

See https://arxiv.org/pdf/2004.14858.pdf for detailed description. Code available on github.com.

MuSe-Wild: .2568

MuSe-Trust: .4128

Reporting Combined (0.5 · Arousal + 0.5 · Valence) for MuSe-Wild and Trustworthiness for MuSe-Trust , both using concordance correlation coefficient (CCC).

MuSe-Topic Target: 76.78 %

Reporting Combined (0.66 · F 1 + 0.34 · Unweighted Average Recall) for the topic predictions.

MuSe-Topic Emo.: 38.81 %

Reporting Combined (0.66 · F 1 + 0.34 · Unweighted Average Recall) for the 3-class valence and arousal predictions and the combined (mean) of valence and arousal.

Challenge Price 2020

Licai Sun, Zheng Lian, Jianhua Tao, Bin Liu, Mingyue Niu
University of the Chinese Academy of Sciences

"Multi-modal Continuous Dimensional Emotion Recognition Using Recurrent Neural Network and Self-Attention Mechanism"

Congratulations for the outstanding performance and the best models (of the 21 submitting teams).

Paper: https://dl.acm.org/doi/10.1145/3423327.3423672

Github: https://github.com/youcaiSUN/MuSe-Wild_2020

Arousal CCC: .473 / Valence CCC: .579 / Combined: .526 (results reproduced)

Second in the leaderboard:

Ruichen Li, Jinming Zhao, Jingwen Hu, Shuai Guo, "Multi-modal Fusion for Video Sentiment Analysis"

Paper: https://dl.acm.org/doi/10.1145/3423327.3423671

Github: https://github.com/FatDs-lrc/RucEmo-muse

Arousal CCC: .456 / Valence CCC: .583 / Combined: .5195 (results reproduced)

Page updated

Google Sites

Report abuse