Table 1.
CNN architecture details for Task 1
| Name | Type | Activations | Learnable Parameters | Total Learnables | |
|---|---|---|---|---|---|
| 1 | 227×227×3 images | Input image | 227×227×3 | - | 0 |
| 2 | 96 7×7×3 convolutions with stride [4 4] and padding [0 0 0 0] | Convolution | 56×56×96 | Weights: 7×7×3×96 Bias: 1×1×96 |
14,208 |
| 3 | ReLU-1 | ReLU | 56×56×96 | - | 0 |
| 4 | Cross channel normalization with 5 channels per element | Cross-channel normalization | 56×56×96 | - | 0 |
| 5 | 3×3 maxpooling with stride [2 2] and padding [0 0 0 0] | Maxpooling | 27×27×96 | - | 0 |
| 6 | 2 groups of 128 5×5×48 convolutions with stride [1 1] and padding [2 2 2 2] | Grouped convolution | 27×27×256 | Weights: 5×5×48×128×2 Bias: 1×1×12×2 |
307,456 |
| 7 | ReLU-2 | ReLU | 27×27×256 | - | 0 |
| 8 | 3×3 maxpooling with stride [2 2] and padding [0 0 0 0] | Maxpooling | 13×13×256 | - | 0 |
| 9 | 50% dropout | Dropout | 13×13×256 | - | 0 |
| 10 | 2 fully connected layer | Fully connected | 1×1×2 | Weights: 2×43,264 Bias: 2×1 |
86,530 |
| 11 | Softmax | Softmax | 1×1×2 | - | 0 |
| 12 | Crossentropyex with classes COVID-19(+) and COVID-19(−) |
classification output | - | - | 0 |