Skip to main content
. 2022 Apr 6;12:5767. doi: 10.1038/s41598-022-09719-3

Table 5.

Modified architecture of the main network.

Layer type Output shape
Input—retinal fundus (600, 600, 3)
Pretrained network (InceptionV3, MobileNetV2 or EfficientNetB3) (1000)
Fully connected

(500) without metadata subnet

(750) with metadata subnet

Batch normalization
ReLU
Dropout (p = 0.2)
Output—fully connected (1)