Industrial Anomaly Detection and Attack Classification Method Based on Convolutional Neural Network

Table 3

Four architectures in this research.

No.	C1		S1		C3		S4		Accuracy (%)	Test Loss
	Kernel	Output	Window	Output	Kernel	Output	Window	Output
1	6  (3  3)	6  (24  24)	2  2	6  (12  12)	16  (3  3)	16  (10  10)	2  2	16  (5  5)	99.32	0.025
2	6  (3  3)	6  (24  24)	2  2	6  (12  12)	16  (5  5)	16  (8  8)	2  2	16  (4  4)	98.80	0.047
3	6  (5  5)	6  (22  22)	2  2	6  (11  11)	16  (3  3)	16  (9  9)	2  2	16  (4  4)	98.76	0.051
4	6  (5  5)	6  (22  22)	2  2	6  (11  11)	16  (5  5)	16  (7  7)	2  2	16  (3  3)	97.69	0.148