Region Dual Attention-Based Video Emotion Recognition

Table 2

Evaluation of region dual attention mechanism on the MHED-F data set.

Method	Accuracy (%)
ResNet-50 + average aggregation	52.34
ResNet-50 + regional location attention mechanism	58.88
ResNet-50 + regional relationship attention mechanism	58.41
ResNet-50 + regional dual attention mechanism	59.81