Table 5 A comparison of our approach with previous deep learning methodologies for bacterial classification
Approach | Number of images | Augmentation | Data split | Loss | Accuracy | Precision | Recall | FScore | MCC |
|---|---|---|---|---|---|---|---|---|---|
Proposed model 1 | 660 | \(\checkmark\) | 7:2:1 | 0.2674 | 99.91 | 98.98 | 98.48 | 98.38 | 98.52 |
Proposed model 2 | 660 | \(\checkmark\) | 7:2:1 | 0.0431 | 99.82 | 97.98 | 96.97 | 96.77 | 97.04 |
ResNet-50 [14] | 660 | \(\checkmark\) | 7:2:1 | 0.0155 | 99.72 | – | 95.45 | 94.34 | – |
MobileNetV2 [17] | 660 | \(\checkmark\) | 7:2:1 | 3.0262 | 95.04 | – | 18.18 | 11.64 | – |
VGG16 [21] | 660 | \(\checkmark\) | 7:2:1 | 3.5460 | 94.58 | – | 10.61 | 5.82 | – |