Table 5.
Layer Type | Kernel Attribute | Number of Filters | Feature Map Size | |
---|---|---|---|---|
Image Input Layer | 64 × 64 × 1 | |||
Main Block 1 |
Convolutional Layer | 3 × 3 × 1, stride 1, padding = same | 32 | 64 × 64 × 32 |
Tanh Layer | 64 × 64 × 32 | |||
Max-Pooling Layer | 2 × 2, stride 2, no padding | 64 × 64 × 32 | ||
Main Block 2 |
Convolutional Layer | 3 × 3 × 32, stride 1, padding = same | 64 | 32 × 32 × 64 |
Tanh Layer | 32 × 32 × 64 | |||
Max-Pooling Layer | 2 × 2, stride 2, no padding | 16 × 16 × 64 | ||
Main Block 3 |
Convolutional Layer | 3 × 3 × 64, stride 1, padding = same | 64 | 16 × 16 × 64 |
Tanh Layer | 16 × 16 × 64 | |||
Max-Pooling Layer | 2 × 2, stride 2, no padding | 8 × 8 × 64 | ||
Main Block 4 |
Convolutional Layer | 3 × 3 × 64, stride 1, padding = same | 128 | 8 × 8 × 128 |
Tanh Layer | 8 × 8 × 128 | |||
Max-Pooling Layer | 2 × 2, stride 2, no padding | 4 × 4 × 128 | ||
Main Block 5 |
Convolutional Layer | 3 × 3 × 128, stride 1, padding = same | 256 | 4 × 4 × 256 |
Tanh Layer | 4 × 4 × 256 | |||
Max-Pooling Layer | 2 × 2, stride 2, no padding | 2 × 2 × 256 | ||
Classification Block |
Fully Connected Layer | 32 | ||
Tanh Layer | ||||
Dropout | ||||
Fully Connected Layer | 3 | |||
Softmax |