Table 5. Confusion matrices of models (A) ResNet64 (B) YOLOv5 (C) YOLOv8.
The highest correct classification rate in all three models belongs to the severe class. In these matrices, rows are actual values and columns are predicted values.
A | ||||
---|---|---|---|---|
ResNet64 | Mild | Moderate | Severe | Healthy |
Mild | 92 | 1 | 0 | 7 |
Moderate | 1 | 93 | 2 | 4 |
Severe | 0 | 1 | 99 | 0 |
Healthy | 7 | 5 | 0 | 89 |