Facial Expression Recognition based on Multi-head Cross Attention Network
This work addresses facial expression recognition for interactive computing domains, but it appears incremental as it builds on an existing model.
The paper tackled facial expression recognition in-the-wild by proposing an extended version of the DAN model, achieving preliminary results of 0.44 mean CCC for VA estimation and 0.33 average F1 score for expression classification.
Facial expression in-the-wild is essential for various interactive computing domains. In this paper, we proposed an extended version of DAN model to address the VA estimation and facial expression challenges introduced in ABAW 2022. Our method produced preliminary results of 0.44 of mean CCC value for the VA estimation task, and 0.33 of the average F1 score for the expression classification task.