Table 4.

The recognition results of AlexNet, VGG-16, and GoogleNet models under the different training cases for each dataset.

Dataset	Recognition Model	$Val.Acc.(\%)$				$F1-Score(\%)$				$E_{\mathit{test}}(\%)$
		Case(A)	Case(B)	Case(C)	Case(D)	Case(A)	Case(B)	Case(C)	Case(D)	Case(A)	Case(B)	Case(C)	Case(D)
Latin	AlexNet	86.54	80.77	99.33	99.19	86.69	85	89.42	85.27	14.1	16.03	12.18	16.03
	Vgg16	82.69	84.62	99.36	98.59	87.24	83.66	92.20	85.50	14.74	18.59	8.33	16.03
	GoogleNet	65.38	69.23	98.05	97.24	73.96	74.29	86.98	86.31	27.56	28.21	14.1	14.47
Malay(Jawi-Arabic)	AlexNet	80.26	89.47	97.57	92.36	80.18	78.55	98.82	80.12	21.93	24.12	1.32	21.49
	Vgg16	72.37	76.32	96.79	92.61	76.18	79.08	89.93	84.85	27.63	23.25	14.04	16.76
	GoogleNet	43.42	47.37	91.78	86.89	39.46	38.4	82.60	72.52	60.53	64.91	18.86	30.26
Korean	AlexNet	85	90	97.24	96.80	89.83	88.05	100	85.80	11.76	12.92	0	15
	Vgg16	68.75	72.50	96.67	97.27	77.81	75.36	88.12	85.77	24.12	27.08	12.92	15.83
	GoogleNet	60	47.50	96.98	96.82	51.34	55.02	89.10	88.45	50	48.75	12.08	12.50
Sanskrit(old Indo-Aryan)	AlexNet	70.24	78.57	95.21	93.12	72.47	65.28	73.47	72.42	31.35	38.10	28.17	28.97
	Vgg16	54.76	61.90	94.82	91.56	55.54	52.51	77.01	70.43	49.60	53.17	24.60	30.95
	GoogleNet	29.76	30.95	91.50	87.73	29.24	27.37	75.11	74.45	74.60	75	26.59	27.78

Significant values are in [bold].