Table 4.

Results of different segmentation methods on the PROCESSED nuclear cladding coating surfaces crack dataset.

Method	MIoU(%)	P(%)	R(%)	F1-Score(%)
U-Net	73.31	79.74	80.68	80.20
DeepLabv3+	72.36	77.51	75.21	76.34
SETR	67.42	71.11	67.14	69.07
SegFormer	72.18	77.24	72.46	74.77
Mask2Former	75.95	80.52	79.06	79.78
CrackFormer-II	82.47	84.97	88.84	86.86
SAM	79.72	83.00	84.28	83.63
CrackCTFuse	84.51	87.20	93.17	90.08