Table 1.

Quantitative results for 3D semantic segmentation on the Sengupta labelled dataset. The bold fonts indicate the best results.

	Method	Road	Sidewalk	Building	Fence	Pole	Vegetation	Vehicle
Accuracy	Sengupta [34]	97.8	86.5	88.5	46.1	38.2	86.9	88.5
	Sengupta [27]	97.0	73.4	89.1	45.7	3.3	81.2	72.5
	Vineet [13]	98.7	91.8	97.2	47.8	51.4	94.1	94.1
	Yang [15]	98.7	93.8	98.2	84.7	66.3	98.7	95.5
	Ours	98.7	95.2	99.3	92.1	77.2	95.3	98.3
IoU	Sengupta [34]	96.3	68.4	83.8	45.2	28.9	74.3	63.5
	Sengupta [27]	87.8	49.1	73.8	43.7	1.9	65.2	55.8
	Vineet [13]	94.7	73.8	88.3	46.3	41.7	83.2	79.5
	Yang [15]	96.6	90.0	95.4	81.1	61.5	91.0	94.6
	Ours	96.7	91.3	98.4	83.0	59.4	93.4	96.0