Pavement Crack Segmentation Detection Integrating Multiple Attention Mechanisms

Model	IoU /%	Precision /%	Recall /%	F1 score /%
U-net	68.18	82.17	79.95	80.95
PSPnet	64.61	78.34	78.74	78.39
DeepLab v3+	69.37	75.56	89.25	81.77
DeepCrack	70.46	80.48	86.44	83.29
Swin-Unet	71.06	82.54	82.64	82.32
Ours	73.24	82.84	86.12	84.44

Table 2. Impact of different modules on model performance
View table
Table 2. Impact of different modules on model performance
Model IoU /% Precision /% Recall /% F1 score /%
Baseline 71.67 80.61 82.53 81.55
Baseline+L 72.24 82.22 83.17 82.69
Baseline+L+DAM 73.24 82.84 86.12 84.44

Table 3. Impact of different attention levels on model performance
View table
Table 3. Impact of different attention levels on model performance
Model IoU /% Precision /% Recall /% F1 score /%
+PAM 72.65 80.37 85.93 83.05
+CAM 72.85 82.42 85.13 83.75
+DAM 73.24 82.84 86.12 84.44

Table 4. Effects of different convolutions in encoder on model performance
View table
Table 4. Effects of different convolutions in encoder on model performance
Model IoU /% Precision /% Recall /% F1 score /%
CNN+Swin+decode 69.62 83.55 82.14 82.83
ResNet+Swin+decode 71.86 81.52 85.36 83.39
ResNest+Swin+decode 73.24 82.84 86.12 84.44

Tools

Get Citation

Copy Citation Text

Pengfei Gao, Liya Zhang, Yukun Wang, Lin Zhang. Pavement Crack Segmentation Detection Integrating Multiple Attention Mechanisms[J]. Laser & Optoelectronics Progress, 2025, 62(10): 1012002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites