Photovoltaic Module Fault Detection Based on Deep Learning Using Cloud Computing

Table 6

Comparison with different pretrained models.

Fault type	Pretrained models (%)		Trained model (%)
	VGG-16	ResNet-50	Proposed solution
Burn marks	49.99	77.42	100.00
Delamination	84.14	68.86	96.66
Discoloration	89.35	67.66	100.00
Glass breakage	87.51	72.54	100.00
Snail trail	53.39	78.66	96.66