Fault-Level Grading of Photovoltaic Cells Employing Lightweight Deep Learning Models

Table 8

Multiclass classification results of the deep ANN using augmented data: (a) GLCM features and (b) LBP features.

		Actual class
		Healthy	Prob. 0.33	Medium faulty	Full faulty
(a) GLCM features
Predicted class	Healthy	148	13	3	20
	Prob. 0.33	27	27	02	9
	Medium faulty	12	2	9	15
	Full faulty	39	3	2	63
Classwise accuracy		65.5%	60%	56.2%	58.9%
Overall accuracy		62.7%
(b) LBP features
Predicted class	Healthy	182	7	2	25
	Less faulty	7	34	3	2
	Medium faulty	7	2	10	6
	Full faulty	30	2	1	74
Classwise accuracy		80.5%	75.5%	62.5%	69.1%
Overall accuracy		76.1%