COVID-19 seems to be an extremely contagious disease and rapid human-to-human transition rate spreading quickly. It is also associated with high ICU admission resulting in an urgent need for development of fast and accurate detection and diagnosis. Identifying positive COVID-19 in early stages helps in isolation and breaking the infection chain.
In this project, we have created the classification model and the severity model for the detection of Covid-19 and evaluation of its severity from chest xrays.
Dice Loss: 0.0221 IOU: 0.9790 Recall: 0.9903 Precision: 0.9924
| Models | Accuracy |
|---|---|
| Conv-Capsule Network | 93.98% |
| Covid-Net | 93.3% |
| Xception | 92.85% |
| DarkNet | 87.02% |
| CoroNet(Xception) | 89.6% |
| Model | MAE (A, B, C, D, E, F) | MSE (A, B, C, D, E, F) |
|---|---|---|
| ResNet50 | 0.8509, 0.8513, 0.7953, 0.8434, 0.7910, 0.69 | 0.9678, 0.9804, 0.8938, 0.9164, 0.8575, 0.695 |
| DenseNet 201 | 0.4655, 0.7565, 0.7361, 0.8706, 0.4987, 0.6809 | 0.3351, 0.8612, 0.7463, 0.9684, 0.3874, 0.6442 |
| VGG16 | 0.9252, 0.9481, 0.9328, 0.9704, 0.8948, 0.8834 | 1.0761, 1.1396, 1.1211, 1.1391, 1.0231, 1.0268 |
Find other models at link