Table 18.

The result of the second model was initially trained using our dataset which is our W-Net model and then further trained using LISC public data

	NE. (%)	EO. (%)	BA. (%)	LY. (%)	MO. (%)
Fold-0	100	100	100	100	80
Fold-1	100	100	100	100	100
Fold-2	100	100	100	100	20
Fold-3	100	100	100	100	100
Fold-4	100	100	100	100	80
Fold-5	100	100	100	83	100
Fold-6	100	100	100	100	100
Fold-7	100	75	100	100	100
Fold-8	100	100	80	100	100
Fold-9	100	100	80	100	100
Avr. Acc.	100	98	96	98	88

The average accuracy for five classes is 96%