Table 4 Detailed information on CNN model employed for “C-3” mode
From: A hybrid deep CNN model for brain tumor image multi-classification
Layer Name | CNN Layer | Activations | Parameters (Trainable) | Total No. of Trainable Parameters |
|---|---|---|---|---|
Input | 227 × 227 × 3 | 227 × 227 × 3 | nil | 0 |
Convolutional | 128 (6 × 6 × 3), stride of (4,4), with (0 0 0 0) padding | 56 × 56 × 128 | 6 × (6 × 3) × 128 weights, 1 × 1 × 128 bias | 13,952 |
Activation layer | Activation layer-1 | 56 × 56 × 128 | nil | 0 |
Normalization | Normalization (cross-channel) | 56 × 56 × 128 | nil | 0 |
Max_pooling | (2 × 2) with stride of (2,2), and (0 0 0 0) padding | 28 × 28 × 128 | nil | 0 |
Convolutional | 96 (6 × 6 × 128), stride of (1,1), and (2 2 2 2) padding | 27 × 27 × 96 | 6 × (6 × 128) × 96 weights, 1 × 1 × 96 bias | 46,752 |
Activation layer | Activation layer-2 | 27 × 27 × 96 | nil | 0 |
Max_pooling | (2 × 2) with stride of (2,2), and (0 0 0 0) padding | 13 × 13 × 96 | nil | 0 |
Convolutional | 96 (2 × 2 × 96), stride of (1,1), and (2 2 2 2) padding | 16 × 16 × 96 | 2 × (2 × 96) × 96 weights, 1 × 1 × 96 bias | 36,864 |
Activation layer | Activation layer-3 | 8 × 8 × 96 | nil | 0 |
Max_pooling | (2 × 2) with stride of (2,2), and (0 0 0 0) padding | 6 × 6 × 256 | nil | 0 |
Fully_connected | 512 Fully_connected | 1 × 1 × 512 | 512 × 6144 weights, 512 × 1 bias | 3,146,240 |
Dropout | 30% | 1 × 1 × 512 | nil | 0 |
Fully_connected | 3 Fully_connected | 1 × 1 × 3 | 512 × 3 weights, 3 × 1 bias | 1539 |
Softmax | Softmax | 1 × 1 × 2 | nil | 0 |
Classification | G-II, G-III, G-IV | nil | nil | 0 |