Table 6.
Layers and parameters of the proposed TrVGG16.
| Layer Name | Layer Type | Filter Size | Stride | Padding | Output Channel | Activation Function |
|---|---|---|---|---|---|---|
| conv1_1 | convolution 2d | 3,3 | 1,1 | 1 | 64 | relu |
| conv1_2 | convolution 2d | 3,3 | 1,1 | 1 | 64 | relu |
| pool1 | max pooling 2d | 2,2 | 2,2 | 0 | 64 | – |
| conv2_1 | convolution 2d | 3,3 | 1,1 | 1 | 128 | relu |
| conv2_2 | convolution 2d | 3,3 | 1,1 | 1 | 128 | relu |
| pool2 | max pooling 2d | 2,2 | 2,2 | 0 | 256 | – |
| conv3_1 | convolution 2d | 3,3 | 1,1 | 1 | 256 | relu |
| conv3_2 | convolution 2d | 3,3 | 1,1 | 1 | 256 | relu |
| conv3_3 | convolution 2d | 3,3 | 1,1 | 1 | 256 | relu |
| pool3 | max pooling 2d | 2,2 | 2,2 | 0 | 256 | – |
| conv4_1 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| conv4_2 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| conv4_3 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| pool4 | max pooling 2d | 2,2 | 2,2 | 0 | 512 | – |
| conv5_1 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| conv5_2 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| conv5_3 | convolution 2d | 3,3 | 1,1 | 1 | 512 | relu |
| pool5 | max pooling 2d | 2,2 | 2,2 | 0 | 512 | – |
| fc6 | fully connected | – | – | – | 4096 | relu |
| fc7 | fully connected | – | – | – | 4096 | relu |
| fc_optimized | fully connected | – | – | – | 4 | softmax |