Table 6.

Crack detection results on AigleRN.

Methods	Tolerance Margin	Pr	Re	F1
Canny [35]	2	0.1989	0.6753	0.2881
Local thresholding [26]	2	0.5329	0.9345	0.667
FFA [43] 12	2	0.7688	0.6812	0.6817
MPS [42]	2	0.8263	0.841	0.8195
CrackForest [44]	2	0.8424	0.801	0.8233
CrackForest [44]	5	0.9028	0.8658	0.8839
Structed prediction [40]	2	0.9178	0.8812	0.8954
Method [67]	2	0.869	0.9304	0.8986
Ensemble network (threshold = 0.6) [57]	2	0.9302	0.9266	0.9238
Ensemble network (threshold = 0.5) [57]	2	0.9334	0.8879	0.9211
U-net [65]	2	0.9127	0.9076	0.91
U-net + HF	2	0.911	0.922	0.913
U-net + MDM	2	0.9138	0.9245	0.914
U-HDN	2	0.921	0.931	0.924