Fine-Grained Point Cloud Semantic Segmentation of Complex Railway Bridge Scenes from UAVs Using Improved DGCNN

Table 6

Segmentation tests of point cloud scenes with different data sets.

Training scenes	Test scene	mIoU (%)	bACC (%)
(a)	(b)	25.40	42.15
	(c)	22.63	36.78
(b)	(a)	58.54	65.26
	(c)	42.99	49.65
(c)	(a)	62.49	71.35
	(b)	46.24	52.42
(a) and (b)	(c)	56.32	67.24
(a) and (c)	(b)	68.44	76.91
(b) and (c)	(a)	83.21	90.31
(a), (b), and (c)	(a)	98.65	99.03
	(b)	91.20	95.34
	(c)	77.87	86.06