Figure 4 | Scientific Reports

Figure 4

From: A convolutional recurrent neural network with attention for response prediction to repetitive transcranial magnetic stimulation in major depressive disorder

Figure 4

The architecture of the TL-BLSTM and TL-BLSTM-attention model was implemented in this study. Input video samples with the size of (5, 224, 224, 3) inserted to the model and spatial features with the size of (5, 128) extracted by TL models for each input sample. When attention is implemented, the second BLSTM layer should return a sequence of features with a size of (5, 64). Finally, the attention layer exports 64 most discriminative features to be classified.

Back to article page