Table 1 Improved network parameters for each layer.
From: Research on the construction of weaponry indicator system and intelligent evaluation methods
Number of network layers | Network type | Convolution kernel size/pixel | Output size/pixels |
|---|---|---|---|
1 | Convolutional layer 1 | \(7 \times 7\) | \(72 \times 72 \times 16\) |
2 | Convolutional layer 2 | \(7 \times 7\) | \(66 \times 66 \times 32\) |
3 | Convolutional layer 3 | \(7 \times 7\) | \(60 \times 60 \times 32\) |
4 | Pooling layer 1 | \(2 \times 2\) | \(30 \times 30 \times 32\) |
5 | Convolutional layer 4 | \(3 \times 3\) | \(28 \times 28 \times 128\) |
6 | Convolutional layer 5 | \(3 \times 3\) | \(26 \times 26 \times 128\) |
7 | Convolutional layer 6 | \(3 \times 3\) | \(24 \times 24 \times 128\) |
8 | Pooling layer 2 | \(2 \times 2\) | \(12 \times 12 \times 128\) |
9 | Residual blocks 1 | \(3 \times 3\) \(3 \times 3\) | \(12 \times 12 \times 128\) |
10 | Convolutional layer 7 | \(3 \times 3\) | \(10 \times 10 \times 256\) |
11 | Convolutional layer 8 | \(3 \times 3\) | \(8 \times 8 \times 256\) |
12 | Convolutional layer 9 | \(3 \times 3\) | \(6 \times 6 \times 256\) |
13 | Pooling layer 3 | \(2 \times 2\) | \(3 \times 3 \times 256\) |
14 | Residual blocks 2 | \(3 \times 3\) \(3 \times 3\) | \(3 \times 3 \times 256\) |
15 | Fully connected layer | \(1 \times 1\) | \(1 \times 1 \times 2304\) |
16 | Fully connected layer | \(1 \times 1\) | \(1 \times 1 \times 1024\) |
