Table 4 Summary of crack segmentation results from 8 networks on Crack500.
From: A novel convolutional neural network for enhancing the continuity of pavement crack detection
Model | Precision | Recall | F-score | mIoU | Params | FLOPS |
|---|---|---|---|---|---|---|
UNet | 99.00 | 55.20 | 70.88 | 63.43 | 24.89M | 211.72G |
DeepLabv3+ | 99.14 | 66.47 | 79.58 | 70.97 | 54.71M | 156.43G |
HRNet | 99.09 | 67.75 | 80.48 | 71.02 | 29.55M | 85.54G |
Segformer | 99.16 | 66.29 | 79.46 | 70.83 | 27.35M | 106.34G |
Crackseg | 99.12 | 65.74 | 79.05 | 70.50 | 44.02M | 266.43G |
Deepcrack | 99.20 | 65.88 | 79.18 | 70.61 | 14.72M | 151.00G |
CrackW-Net | 99.32 | 65.63 | 79.03 | 70.41 | 52.46M | 250.32G |
CPCDNet | 99.10 | 66.82 | 79.82 | 71.16 | 22.93M | 236.56G |
