Table 5.
Classification results under different pre-training methods.
| TS | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Experiment 1 | Precision | Recall | F1 | Accuracy | AUC | ||||
| TSRNet | 100% | 99.59% | 99.79% | 99.80% | 1 | ||||
| DenseNet169 |
100% |
99.19% |
99.59% |
99.60% |
1 |
||||
| Only ImageNet Pretrained | |||||||||
| Experiment 2 |
Precision |
Recall |
F1 |
Accuracy |
AUC |
||||
| TSRNet | 100% | 99.18% | 99.59% | 99.59% | 0.9996 | ||||
| DenseNet169 |
99.59% |
98.78% |
99.18% |
99.19% |
0.9992 |
||||
| Random Initialization | |||||||||
| Experiment 3 |
Precision |
Recall |
F1 |
Accuracy |
AUC |
||||
| TSRNet | 100% | 97.96% | 98.97% | 98.99% | 0.9998 | ||||
| DenseNet169 | 97.98% | 98.37% | 98.17% | 98.19% | 0.9999 | ||||