Table 5.

The pre-trained model performance in four categories on the difference image format.

Category = 4 (Healthy, Early, Mid, and Late PD)
Net	Image Format	Batchsize	MaxEpochs	Accuracy	Recall	Precision	F-Score	Kappa
AlexNet	Gray	80	40	0.825	0.753	0.874	0.809	0.725
	Color	162	81	0.622	0.504	0.701	0.587	0.397
GoogLeNet	Gray	14	7	0.687	0.673	0.728	0.700	0.518
	Color	34	17	0.586	0.492	0.672	0.568	0.368
VGG19	Gray	20	10	0.819	0.758	0.870	0.810	0.720
	Color	24	12	0.807	0.808	0.838	0.823	0.707
ResNet50	Gray	8	4	0.739	0.729	0.710	0.719	0.607
	Color	8	4	0.827	0.837	0.749	0.791	0.743
ResNet101	Gray	6	3	0.767	0.691	0.668	0.679	0.656
	Color	12	6	0.831	0.824	0.857	0.840	0.744
DenseNet	Gray	18	9	0.807	0.722	0.843	0.778	0.704
	Color	16	8	0.855	0.821	0.903	0.860	0.724