Caption Generation Based on Emotions Using CSPDenseNet and BiLSTM with Self-Attention

Figure 3

The architecture of BiLSTM with self-attention mechanism.