TABLE 2. Optimal Values for the Parameters and Hyperparameters for the Custom and Pretrained Models Obtained Through Optimization Tools (M = Momentum, ILR = Initial Learning Rate, L2=L2-Weight Decay, and D = Dropout Ratio).

Models	Optimal values
	M	ILR	L2	D	Optimizer	Activation
Custom	–	–	–	0.5	SGD	ReLU
Pretrained	0.95	1e-3	1e-6	–	–	–