Detecting Gear Surface Defects Using Background-Weakening Method and Convolutional Neural Network

Table 3

The effect of different values on the clustering effect of datasets.

value	5
Accuracy	82.32%
Boxes	(47, 42), (68, 45), (40, 60), (37, 25), (31, 43)
Ratios	[0.67, 0.72, 1.12, 1.48, 1.51]
value	7
Accuracy	82.32%
Boxes	(72, 37), (37, 25), (46, 36), (39, 60), (66, 52), (42, 44), (50, 43)
Ratios	[0.65, 0.95, 1.16, 1.27, 1.28, 1.48, 1.95]
value	9
Accuracy	86.06%
Boxes	(39, 61), (58, 53), (54, 38), (33, 25), (43, 45), (41, 26), (31, 41), (46, 40.5), (73, 38)
Ratios	[0.64, 0.76, 0.96, 1.09, 1.14, 1.32, 1.42, 1.58, 1.92]