Autonomous Identification of Bridge Concrete Cracks Using Unmanned Aircraft Images and Improved Lightweight Deep Convolutional Networks

Table 4

Model identification performance comparison of different algorithms.

Models	Precision (%)	Recall (%)	Accuracy (%)	IOU (%)
The developed method	97.62	97.23	97.42	93.25
UNET	94.23	94.35	94.29	93.12
FCN	93.89	92.95	93.21	92.86
PSPNET	92.78	92.57	92.82	91.38
DEEPLABV3	93.85	93.74	93.78	92.82