Table 3 Detailed summary of the proposed CNN architecture, including layer configurations and parameters.
From: Spike train analysis in rehabilitation movement classification using deep learning approach
Proposed CNN architecture | |
|---|---|
Input: 224 \(\times\) 224 \(\times\) 3 images | |
1 | layers = intended layers |
2 | layers.add(zero padding 3 \(\times\) 3) |
3 | layers.add(convolutional 7 \(\times 7, 64\)) |
4 | layers.add(batch normalization) |
5 | layers.add(ReLu) |
6 | layers.add(maximum pooling stride of 2) |
7 | layers.add(convolutional block 1 \(\times 1, 64; 3\times 3, 64; 1\times \text{1,256}\)) |
8 | layers.add(2 identity blocks) |
9 | layers.add(convolutional block 1 \(\times 1, 128; 3\times 3, 256; 1\times \text{1,1024}\)) |
10 | layers.add(3 identity blocks) |
11 | layers.add(convolutional block 1 \(\times 1, 256; 3\times 3, 256; 1\times \text{1,1024}\)) |
12 | layers.add(5 identity blocks) |
13 | layers.add(convolutional block 1 \(\times 1, 512; 3\times 3, 512; 1\times \text{1,2048}\)) |
14 | layers.add(2 identity blocks) |
15 | layers.add(global average pooling 1 \(\times 1\times 2048\)) |
16 | layers.add(fully connected layers dropout probability = 0.5) |
17 | layers.add(fully connected layers dropout probability = 0.5) |
18 | layers.add(softmax = 5 classes) |
Output | |
