Research Article
Pathological Myopia Image Recognition Strategy Based on Data Augmentation and Model Fusion
Table 3
Training results of VGG-16, AlexNet, GoogLeNet, and ResNet-50 on the filtered datasets.
| Primary learner | PALM-Training800-color | PALM-Training1600-overturning-cropping-deforming | PALM-Training3200-overturning-noise-color-cropping-deforming-dimming | PALM-Training1600-overturning-dimming-imgaug2 | Accuracy | Loss rate | Accuracy | Loss rate | Accuracy | Loss rate | Accuracy | Loss rate |
| AlexNet | 0.946083 | 0.162323 | 0.950833 | 0.160698 | 0.954667 | 0.197096 | 0.957583 | 0.157773 | GoogLeNet | 0.909 | 0.203461 | 0.9395 | 0.169759 | 0.962417 | 0.13617727 | 0.947917 | 0.170418 | ResNet-50 | 0.9395 | 0.219571 | 0.953583 | 0.151836 | 0.955917 | 0.156737 | 0.95125 | 0.166358 | VGG-16 | 0.9503 | 0.17019 | 0.95267 | 0.16523 | 0.9555 | 0.27185 | 0.95858 | 0.18674 |
|
|