Table 3.

The selected hyper-parameter for the classification stage.

Model	Batch size	Epochs	Learning rate	Weight decay	F1-Score (%)
EfficientnetB5	64	30	0.001	0.00001	98.8
Resnet101	32	50	0.001	0.0001	98.7
Resnext101 32 × 8d	64	40	0.001	0.000001	98.7