boundingbox_label_readme.md

BoundingBox label

We generate 3D bounding boxes for both the SemanticKITTI dataset and KITTI-road dataset by first enclosing the clusters generated by the Euclidean clustering algorithm in PCL library, and then manually modifying erroneous cases.

KITTI-road dataset is partially manually labeled.

Instance label is saved in xxxxxx.npy file.

xxxxxx.npy --> box=["instance name","instance label","moving label",[x,y,z,l,w,h,theta]]
"instance name" --> "car","bus",....
"instance label"--> 1-------car
					2-------bycicle
					3-------bus
					4-------motorcycle
					5-------onrails
					6-------truck
					7-------othervehicle
					8-------person
					9-------bicyclist
					10-------motorcyclist
					12------other
					
"moving label"-->   0-------static
					1-------moving
[x,y,z,l,w,h,theta]-->center_x, center_y, center_z of bounding box
                     length,width,height
                     theta, the heading of bounding box ->[-90,90]

You can check bounding box label by python

import numpy as np

bounding_label=np.load(xxxxxx.npy,allow_pickle=True)
print(bounding_label)