Deep-Learning-Based Bughole Detection for Concrete Surface Image

Table 3

Accuracy of the proposed DCNN on testing images.

Number of images	800
Number of pixels	52,428,800
Number of pixels with false-positive errors	592,445
Number of pixels with false-negative errors	1,279,263
Percentage of false-positive errors	1.13%
Percentage of false-negative errors	2.44%
Accuracy	96.43%