Annotated Lidar

Lidar Scans

The labelled pointclouds are released as pcd files. To read an pcd file into a numpy array, we recommend the package pypcd.

Sequence # scans Preview
ntu_day_01 6010 ntu_day_01
ntu_day_02 2273 ntu_day_02
ntu_day_10 3232 ntu_day_10
ntu_night_13 2323 ntu_night_13
kth_day_06 8893 kth_day_06
kth_day_09 7655 kth_day_09
kth_night_05 6638 kth_night_05
tuhh_day_02 4986 tuhh_day_02
tuhh_day_03 8376 tuhh_day_03
tuhh_night_08 7075 tuhh_night_08
tuhh_night_09 1832 tuhh_night_09

Semantic classes

The classes corresponding to the label value are as follows:

{0: 'barrier',
 1: 'bike',
 2: 'building',
 3: 'chair',
 4: 'cliff',
 5: 'container',
 6: 'curb',
 7: 'fence',
 8: 'hydrant',
 9: 'infosign',
 10: 'lanemarking',
 11: 'noise',
 12: 'other',
 13: 'parkinglot',
 14: 'pedestrian',
 15: 'pole',
 16: 'road',
 17: 'shelter',
 18: 'sidewalk',
 19: 'stairs',
 20: 'structure-other',
 21: 'traffic-cone',
 22: 'traffic-sign',
 23: 'trashbin',
 24: 'treetrunk',
 25: 'vegetation',
 26: 'vehicle-dynamic',
 27: 'vehicle-other',
 28: 'vehicle-static'}