Point Cloud Label Tool

This is a annotation tool for the 3D detection and tracking of point cloud based on open3d.

Requiments

open3d = 0.9.0 pillow numpy

Handling Instruction

The key bindings come from GLFW.

• ",/.": move prev frame or next frame
• "D/L": load the initial label from det or manual
• "S": save the label to the label_path
• "ENTER": switch the activated box to edit
• "Up/Down;Left/Right": move the activated box in the x-y plane
• "1/2;3/4;5/6": change the size of the activated box
• "R": rotate the activated box
• "PageUp/PageDown": change the track_id of the activated box
• "X": delete the activated bounding box
• "N": new a bounding box at the origin

Data Path

Edit the data_path, det_path, label_path of yourself.
The LiDAR data should be the '.pcd' file.
And the annotations are the ".txt", of which each row is (class,l,w,h,x,y,z,yaw,id).

To be Mentioned

The code is for open3d 0.9.0, while other version open3d can be used if you change some specific functions, such as the open3d.geometry.Geometry.rotate() in Line 188.