Weakly-supervised structural surface crack detection algorithm based on class activation map and superpixel segmentation

Advances in Bridge Engineering

Table 4 Bayesian optimization process

Iteration	Target	Region_size	Region_size (rounded)	threshold	Threshold (rounded)
1	0.0835	38.48	38	220.4	220
2	0.001419	23.13	23	246.2	246
3	0.3245	49.04	49	189.4	189
4	0.297	19.14	19	36.2	36
5	0.4203	89.92	90	142.1	142
6	0.5637	34.15	34	125	125
7	0.559	35.46	35	122.1	122
8	0.183	100	100	10	10
9	0.5203	10	10	112.5	113
10	0.5593	10	10	152.8	153
11	0.6669	43.94	44	144.3	144
12	0.4376	83.52	84	86.88	87
13	0.6058	59.3	59	134	134
14	0	100	100	250	250
15	0.5919	36.05	36	157	157
16	0.5294	57.2	57	152.9	153
17	0.4178	10.28	10	78.96	79
18	0.4138	49.59	50	85.22	85
19	0.4725	99.96	100	114	114
20	0.666	46.21	46	136.3	136
21	0.6292	38.38	38	142.3	142
22	0.4499	10.31	10	177.9	178
23	0.3538	99.52	100	63.76	64
24	0.6778	49.4	49	141	141
25	0.6698	47.11	47	140.9	141
26	0.1317	100	100	184.7	185
27	0.1907	49.31	49	10	10
28	0.5842	62.1	62	110.1	110
29	0.5682	10	10	133.6	134
30	0.6527	51.22	51	123.4	123
31	0.2081	11.04	11	12.78	13
32	0.3062	62.81	63	52.58	53
33	0.6648	50.95	51	132.3	132
34	0.4945	78.68	79	116.1	116
35	0.6528	47.02	47	148.1	148
36	0.5732	49.48	49	110.3	110
37	0.7003	51.84	52	138.4	138
38	0.3674	31.2	31	63.33	63
39	0.4653	25.52	26	97.22	97
40	0.4361	99.73	100	92.86	93