A Modified Fully Convolutional Network for Crack Damage Identification Compared with Conventional Methods

Table 2

Comparisons of different methods.

Method	Author	Size of image	Number of dataset for training and validation	Precision	Recall		mIoU
U-net	Liu et al. (2019)		10000	91%	91%	90%	61.28%
Sliding window; Method	Cha et al. (2017)		10000	85%	82%	83%	46.54%
FCN	Yang et al. (2018)		10000	82%	79%	80%	57.43%
Modified FCN (no data processing)	Meng et al.		10000	96.79%	92%	90%	71.12%
FCN (with data processing)	Meng et al.		10000	85%	81%	82%	59.36%
Modified FCN	Meng et al.		10000	97.92%	92%	91%	80.24%