Convenient User Interface For Labelling 3D Point Cloud

• Main Author: Tan, Shi Hao
• Format: Final Year Project / Dissertation / Thesis
• Published: 2021
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://eprints.utar.edu.my/4212/1/1604655_FYP_Report_%2D_SHI_HAO_TAN.pdf; http://eprints.utar.edu.my/4212/

Nowadays, most labelling tools are able to label 2D images well. However, the same cannot be said for 3D point clouds. Thus, the 3D point cloud labelling tools are pretty expensive in the marketplace. This report reviews different types of 3D point cloud labelling tools, including automatic and manual 3D point cloud labelling tools. This project mainly focuses on manual open-source user interfaces for 3D point cloud labelling. A comparison between the opensource user interfaces was conducted. Objective evaluation of usability and labelling speed between the user interfaces was carried out. The user interfaces used for comparison are POINTS, 3D BAT, LATTE. Three of the user interfaces are screen-based and web-based tools, and the comparison is made using the same point cloud scene, taken from KITTI datasets. The usability of the user interfaces was evaluated in terms of instruction and shortcuts, projective views, editable in projective views, fast annotation feature, timer, and main view focus mode. The labelling speed of the user interfaces is evaluated in terms of time and errors. After the comparison, POINTS is the user interface with many useful features, which aids the users to label easily and faster. The labelling time of POINTS is the shortest among three user interfaces. POINTS can perform well as or better than the 3D point cloud labelling tool in the marketplace.