• Maintainer status: maintained
• Author: Zhi Yan
• License: GPL-3.0
• Dataset: https://lcas.lincoln.ac.uk/wp/research/data-sets-software/l-cas-3d-point-cloud-people-dataset/

The tool provides a semi-automatic labeling function, means the 3D point cloud data (loaded from the PCD file) is first clustered to provide candidates for labeling, each candidate being a point cluster. Then, the user annotating the data, can label each object by indicating candidate's ID, category, and visibility. A flowchart of this process is shown below.

The quickest way to activate the optional steps is to modify the source code and recompile. 😱

Ubuntu 20.04.4 LTS (ROS Noetic)
Qt 5.12.8
VTK 7.1.1
PCL 1.10
LIBSVM

mkdir build
cd build
cmake ..
make
./cloud_annotation_tool

lcas_simple_data.zip contains 172 consecutive frames (in .pcd file) with 2 fully annotated pedestrians.

If you are considering using this tool and the data provided, please reference the following:

@article{yz19auro,
   author = {Zhi Yan and Tom Duckett and Nicola Bellotto},
   title = {Online learning for 3D LiDAR-based human detection: Experimental analysis of point cloud clustering and classification methods},
   journal = {Autonomous Robots},
   year = {2019}
}