Table 6 Comparison of various approaches. The values of each metric is reported in %. Bold highlight the best performing model for each evaluation metric.

From: Deep segmentation of 3+1D radar point cloud for real-time roadside traffic user detection

Method

F1

mAP.50

mAP.75

mAP

PointNet22

90.12

−

−

−

DBSCAN49 + PointNet22

−

72.60

67.70

66.90

PointNet22 + DBSCAN49

90.12

81.96

76.70

74.80

PointNet22 + Instance formation

90.12

84.60

79.80

78.68

This work

95.35

91.03

81.46

80.01