Table 2.
Comparison of the performance among baseline encoders.
| Method | Accuracy |
|---|---|
| VGG16 | 0.5413 |
| Xception | 0.5502 |
| ResNet50 | 0.5506 |
| InceptionV3 | 0.5876 |
| DenseNet121 | 0.4591 |
| DenseNet169 | 0.5966 |
| DenseNet201 | 0.6020 |
All the models obtained the following four images as one data input: an initial image, spectral domain optical coherence tomography (SD-OCT) after the first injection, SD-OCT after the second injection, and SD-OCT after the third injection. The same fusion method, namely, attention, was employed.