Table 3.
The details of the proposed CNN model.
Name | Type | Activations | Learnables | |
---|---|---|---|---|
1 | ChestXrayCT | Image Input | 227 × 227 × 3 | – |
2 | conv1 | Convolution | 74 × 74 × 128 | Weights 9 × 9 × 3 × 128 Bias 1 × 1x128 |
3 | relu1 | ReLU | 74 × 74 × 128 | – |
4 | norm1 | Cross channel normalization | 74 × 74 × 128 | – |
5 | pool1 | Max Pooling | 37 × 37 × 128 | – |
6 | conv2 | Convolution | 19 × 19 × 256 | Weights 3 × 3 × 128 × 256 Bias 1 × 1 × 256 |
7 | relu2 | ReLU | 19 × 19 × 256 | – |
8 | norm2 | Cross channel normalization | 19 × 19 × 256 | – |
9 | pool2 | Max Pooling | 10 × 10 × 256 | – |
10 | conv3 | Convolution | 5 × 5 × 256 | Weights 3 × 3 × 256 × 256 Bias 1 × 1 × 256 |
11 | relu3 | ReLU | 5 × 5 × 256 | – |
12 | norm3 | Cross channel normalization | 5 × 5 × 256 | – |
13 | pool3 | Max Pooling | 3 × 3 × 256 | – |
14 | conv4 | Convolution | 2 × 2 × 512 | Weights 3 × 3 × 256 × 512 Bias 1 × 1 × 512 |
15 | relu4 | ReLU | 2 × 2 × 512 | – |
16 | norm4 | Cross channel normalization | 2 × 2 × 512 | – |
17 | pool4 | Max Pooling | 1 × 1 × 512 | – |
18 | conv5 | Convolution | 1 × 1 × 512 | Weights 3 × 3 × 512 × 512 Bias 1 × 1 × 512 |
19 | relu5 | ReLU | 1 × 1 × 512 | – |
20 | norm5 | Cross channel normalization | 1 × 1 × 512 | – |
21 | pool5 | Max Pooling | 1 × 1 × 512 | – |
22 | fc1 | Fully connected | 1 × 1 × 1024 | Weights 1024 × 512 Bias 1024 × 1 |
23 | drop1 | Dropout | 1 × 1 × 1024 | – |
24 | fc2 | Fully connected | 1 × 1 × 1024 | Weights 1024 × 1024 Bias 3 × 1 |
25 | relu6 | ReLU | 1 × 1 × 1024 | – |
26 | drop2 | Dropout | 1 × 1 × 1024 | – |
27 | fc3 | Fully connected | 1 × 1 × 3 | Weights 3 × 1024 Bias 3 × 1 |
28 | Softmax | Softmax | 1 × 1 × 3 | – |
29 | classoutput | Classification output | – | – |