This tool could annotate the KITTI Velodyne point cloud with the label. I found there is no such tool could do this. It's simple but I prefer there is some code available directly.
Time:
Trainig: Using 8 CPU core may take around whole night.
Testing: Using 8 CPU core may take around 30min.
If you have any questions, please let me know.
Some of the code is from: https://github.com/barrykui/kitti_object_vis
Download the data (calib, label_2, velodyne) from Kitti Object Detection Dataset
The folder structure is as following (you don't need to remove other folder, just make sure those folder exist):
training
calib
label_2
velodyne
testing
velodyne
Requirements
pip install laspy
pip install numpy
pip install pymp-pypi
Run
##Explaination
python label_kitti_point_cloud.py /input_folder/ /output_folder/ valid_distance has_label num_thread
##Example
####for training set
python kitti_label.py /volumes1/KITTI/training/ /volumes1/KITTI/las/ 50 True 8
####for testing set
python kitti_label.py /volumes1/KITTI/testing/ /volumes1/KITTI/las/ 50 False 8