Algorithm 1: Proposed Modified MobileNetV2 Algorithm for COVID-19 Detection |
Input: 52,000 chest X-ray images. (80% train, 20% test data) |
Output: Result = COVID-19 positive or healthy |
Step 1: batch normalization, preprocessing, augmentation |
Step 2: Freeze the base layer and add proposed convolution layer with image size |
224, kernel size (3, 3), optimizer = RMSprop, activation: ReLU |
Step 3: Feed the first residual convolution layer with kernel size (2, 2), activation = |
ReLU, then average pooling, optimizer = RMSprop |
Step 4: Feed into the second residual convolution layer with kernel size (1, 1), stride = 2. |
Average pooling, dropout, optimizer = none |
Step 5: Feed into the third residual convolution layer with kernel size (2, 2), stride = 1. |
Max pooling, dropout, optimizer = RMSprop |
Step 6: Feed into the fourth and fifth residual convolution layer with kernel size (1, 1), |
stride = 2. Max pooling, dropout, optimizer = none |
Step 7: Feed into sixth residual convolution layer with kernel size (2, 2), stride = 2, no |
pooling, dropout, optimizer = RMSprop |
Step 8: Feed into seventh residual convolution layer with kernel size (2, 2), stride = 1, |
average pooling, dropout, optimizer = RMSprop |
Step 9: Apply proposed layer with image size 224 × 224, kernel size (3, 3), optimizer = |
RMSprop, activation: ReLu |
Step 10: Finding the accuracy, precision, f1 score, recall |