Table 2.

Summary of the parameters used in training the deep learning models.

Parameter	Value
Weight Initialization	Native: Random initialization, Transfer learning: Pre-trained on ImageNet
Training set sample size/fold	2,006 participants, ~3,710 eyes, ~12,595 images
Validation set sample size/fold	97 participants, ~218 eyes, ~745 images
Testing set sample size/fold	226 participants, ~436 eyes, ~1,482 images
Input image size	224 × 224 pixels
Network output	Softmax probability of GON
Number of training iterations	50,000 (~10 epochs)
Learning rate	0.005 (iters. 0–100,00)