Table 2.
Accuracy of 8 AI models in 4 comparative experiments on different test datasets.
| Model Database | ResNet50 | ResNet101 | VGG16 | VGG19 | GoogleNet | MobileNetV2 | AlexNet | DenseNet201 | |
|---|---|---|---|---|---|---|---|---|---|
| Using Transfer Learning | MET | 96.43% | 92.86% | 92.86% | 96.43% | 60.71% | 85.71% | 82.14% | 100.00% |
| MOP | 94.44% | 88.89% | 100.00% | 100.00% | 86.11% | 55.56% | 75.00% | 100.00% | |
| Using Transfer Learning(10% Gaussian noise) | MET | 96.43% | 96.43% | 96.43% | 85.71% | 64.29% | 64.29% | 71.43% | 85.71% |
| MOP | 94.44% | 86.11% | 91.67% | 94.44% | 88.89% | 75.00% | 69.44% | 83.33% | |
| Using Transfer Learning(20% Gaussian noise) | MET | 92.86% | 82.14% | 92.86% | 92.86% | 53.57% | 57.14% | 53.57% | 75.00% |
| MOP | 88.89% | 83.33% | 88.89% | 94.44% | 77.78% | 69.44% | 52.78% | 88.89% | |
| Using Transfer Learning(30% Gaussian noise) | MET | 82.14% | 78.57% | 89.29% | 71.43% | 53.57% | 67.86% | 57.14% | 64.29% |
| MOP | 86.11% | 80.56% | 41.67% | 86.11% | 58.33% | 61.11% | 41.67% | 80.56% | |
| No Transfer Learning | MET | 75.00% | 82.14% | 57.14% | 53.57% | 67.86% | 75.00% | 71.43% | 78.57% |
| MOP | 77.78% | 66.67% | 58.33% | 50.00% | 72.22% | 38.89% | 63.89% | 33.33% |