Table 4.
Evaluation based on three-way class classification.
| Model (Trainable/Total Parameters) |
Dataset | Accuracy | Sen | PPV | F1 |
|---|---|---|---|---|---|
| VGG-16 (165,379/14,880,835) |
Training | 96.68 | 96.81 | 96.88 | 96.84 |
| Validation | 97.33 | 97.30 | 97.31 | 97.30 | |
| Testing | 96.64 | 93.01 | 96.81 | 94.87 | |
| VGG-19 (264,195/20,288,579) |
Training | 84.19 | 82.94 | 84.86 | 83.89 |
| Validation | 89.20 | 88.92 | 89.68 | 89.30 | |
| Testing | 88.34 | 85.37 | 89.12 | 87.21 | |
| Inception-V3 (427,523/54,765,027) |
Training | 87.33 | 87.50 | 87.03 | 87.26 |
| Validation | 91.76 | 92.24 | 91.19 | 91.71 | |
| Testing | 89.60 | 89.94 | 89.46 | 89.70 | |
| ResNet-50 (13,141,507/36,729,987) |
Training | 87.11 | 85.07 | 84.02 | 84.54 |
| Validation | 87.78 | 86.36 | 88.37 | 87.35 | |
| Testing | 85.70 | 83.60 | 86.38 | 84.97 | |
| Proposed BNCNN (2,786,435/2,787,683) |
Training | 96.32 | 96.30 | 96.52 | 96.41 |
| Validation | 97.49 | 97.43 | 97.48 | 97.45 | |
| Testing | 96.84 | 93.06 | 97.40 | 95.18 |