SingleNet: A Lightweight Convolutional Neural Network for Safety Detection of an Industrial Control System

Table 3

Experimental results of the comparison algorithms on the public datasets.

Datasets	Algorithms	Accuracy	Training time	Model size
Oil depot dataset	SingleNet	98.4%	3 min 18 s	1.6 MB
	LeNet	72.6%	2 h 14 min 26 s	101 MB
	AlexNet	27.6%	1 h 6 min 38 s	22.4 MB
	SqueezeNet	94.5%	2 h 14 min 59 s	2.65 MB
	MobileNet	91.0%	2 h 16 min 49 s	12.4 MB
	ShuffleNet	91.6%	1 h 10 min 34 s	1.25 MB
Batadal dataset	SingleNet	98.4%	3 min 31 s	1.62 MB
	LeNet	84.5%	18 min 37 s	15.5 MB
	AlexNet	46.9%	19 min 24 s	3.15 MB
	SqueezeNet	2.77556e-17	26 min 28 s	2.02 MB
	MobileNet	97.7%	36 min 27 s	12.4 MB
	ShuffleNet	Not convergence	—	—
Mississippi dataset	SingleNet	98.2%	3 min 33 s	1.63 MB
	LeNet	86.4%	15 min 29 s	10.6 MB
	AlexNet	84.4%	8 min 2 s	3.16 MB
	SqueezeNet	98.9%	23 min 10 s	2.66 MB
	MobileNet	89.6%	31 min 10 s	12.4 MB
	ShuffleNet	Not convergence	—	—