Table 2.
Classification report of CNN models (data augmentation performed and obtained 1200 images, of which 960 were training images and 240 were validation images).
| Number of Epochs = 5 | |||||
|---|---|---|---|---|---|
| CNN Model | Grade | Precision | Recall | F1-Score | Accuracy |
| Model 1: EfficientNetV2L | Grade 0 | 0.78 | 0.74 | 0.76 | 0.71 |
| Grade 1 | 0.72 | 0.62 | 0.67 | ||
| Grade 2 | 0.66 | 0.79 | 0.72 | ||
| Model 2: ResNet152V2 | Grade 0 | 0.35 | 0.82 | 0.49 | 0.45 |
| Grade 1 | 0.66 | 0.24 | 0.36 | ||
| Grade 2 | 0.55 | 0.40 | 0.46 | ||
| Model 3: DenseNet201 | Grade 0 | 0.87 | 0.33 | 0.48 | 0.47 |
| Grade 1 | 0.73 | 0.12 | 0.20 | ||
| Grade 2 | 0.41 | 0.96 | 0.57 | ||
| Ensemble of CNN model (Model 1 + Model 2 + Model 3) | Grade 0 | 0.78 | 0.77 | 0.78 | 0.75 |
| Grade 1 | 0.67 | 0.82 | 0.74 | ||
| Grade 2 | 0.85 | 0.65 | 0.73 | ||
| Number of epochs = 10 | |||||
| CNN model | Grade | Precision | Recall | F1-Score | Accuracy |
| Model 1: EfficientNetV2L | Grade 0 | 0.70 | 0.87 | 0.77 | 0.76 |
| Grade 1 | 0.78 | 0.78 | 0.78 | ||
| Grade 2 | 0.80 | 0.67 | 0.73 | ||
| Model 2: ResNet152V2 | Grade 0 | 0.66 | 0.61 | 0.63 | 0.61 |
| Grade 1 | 0.65 | 0.53 | 0.58 | ||
| Grade 2 | 0.55 | 0.69 | 0.61 | ||
| Model 3: DenseNet201 | Grade 0 | 0.56 | 0.75 | 0.64 | 0.69 |
| Grade 1 | 0.77 | 0.57 | 0.66 | ||
| Grade 2 | 0.75 | 0.78 | 0.76 | ||
| Ensemble of CNN model (Model 1 + Model 2 + Model 3) | Grade 0 | 0.78 | 0.80 | 0.79 | 0.78 |
| Grade 1 | 0.74 | 0.84 | 0.79 | ||
| Grade 2 | 0.84 | 0.69 | 0.76 | ||
| Number of epochs = 15 | |||||
| CNN model | Grade | Precision | Recall | F1-Score | Accuracy |
| Model 1: EfficientNetV2L |
Grade 0 | 0.00 | 0.00 | 0.00 | 0.37 |
| Grade 1 | 0.38 | 0.83 | 0.52 | ||
| Grade 2 | 0.34 | 0.16 | 0.22 | ||
| Model 2: ResNet152V2 |
Grade 0 | 0.28 | 1.00 | 0.43 | 0.28 |
| Grade 1 | 0.00 | 0.00 | 0.00 | ||
| Grade 2 | 0.00 | 0.00 | 0.00 | ||
| Model 3: DenseNet201 |
Grade 0 | 0.00 | 0.00 | 0.00 | 0.34 |
| Grade 1 | 0.00 | 0.00 | 0.00 | ||
| Grade 2 | 0.34 | 0.99 | 0.51 | ||
| Ensemble of CNN model (Model 1 + Model 2 + Model 3) | Grade 0 | 0.82 | 0.76 | 0.79 | 0.78 |
| Grade 1 | 0.76 | 0.80 | 0.78 | ||
| Grade 2 | 0.78 | 0.78 | 0.78 | ||
| Number of epochs = 20 | |||||
| CNN model | Grade | Precision | Recall | F1-Score | Accuracy |
| Model 1: EfficientNetV2L |
Grade 0 | 0.74 | 0.83 | 0.79 | 0.82 |
| Grade 1 | 0.83 | 0.83 | 0.83 | ||
| Grade 2 | 0.87 | 0.79 | 0.83 | ||
| Model 2: ResNet152V2 |
Grade 0 | 0.64 | 0.55 | 0.59 | 0.62 |
| Grade 1 | 0.66 | 0.72 | 0.69 | ||
| Grade 2 | 0.55 | 0.56 | 0.55 | ||
| Model 3: DenseNet201 |
Grade 0 | 0.66 | 0.67 | 0.66 | 0.69 |
| Grade 1 | 0.66 | 0.80 | 0.72 | ||
| Grade 2 | 0.83 | 0.56 | 0.67 | ||
| Ensemble of CNN model (Model 1 + Model 2 + Model 3) | Grade 0 | 0.82 | 0.88 | 0.85 | 0.85 |
| Grade 1 | 0.85 | 0.87 | 0.86 | ||
| Grade 2 | 0.89 | 0.80 | 0.84 | ||