Annotation Tool and Urban Dataset for 3D Point Cloud Semantic Segmentation

Muhammad Ibrahim,Ajmal Mian,Michael Wise,Naveed Akhtar

doi:10.1109/access.2021.3062547

Muhammad Ibrahim, Ajmal Mian + Show 2 more

Open Access

https://doi.org/10.1109/access.2021.3062547

Copy DOI

Abstract

Accurate semantic segmentation of unstructured 3D point clouds requires large amount of annotated training data for deep learning. However, there is currently no free specialized software available that can efficiently annotate large 3D point clouds. We fill this gap by introducing PC-Annotate - a public annotation tool for 3D point cloud research. The proposed tool not only enables systematic annotation with a variety of fundamental volumetric shapes, but also provides useful functionalities of point cloud registration and the generation of volumetric samples that can be readily consumed by contemporary deep learning point cloud models. We also introduce a large outdoor public dataset for 3D semantic segmentation. The proposed dataset, PC-Urban is collected in a civic setup with Ouster LiDAR and labeled with PC-Annotate. It has over 4.3 billion points covering 66K frames and 25 annotated classes. Finally, we provide baseline semantic segmentation results on PC-Urban for popular recent techniques.

Highlights

P OINT CLOUD semantic segmentation is a central problem in the real-world scene understanding
In Table 4, we summarise the results of our experiments for point cloud semantic segmentation
There is no effective public tool to annotate large point cloud datasets. This paper filled this gap by introducing PCAnnotate - a user friendly comprehensive public annotation tool for 3D point clouds

Summary

Introduction

P OINT CLOUD semantic segmentation is a central problem in the real-world scene understanding. A point cloud encodes objects and scenes using their surface coordinates and enables accurate analysis based on their 3D shapes [1].point cloud semantic segmentation finds many important applications in emerging technologies such as self-driving vehicles, human-machine interaction, automatic surgery, and robot navigation. Deep neural networks have been shown to learn accurate computational models for this task. They can only do so with the availability of large amount of annotated training data. The ground breaking performance of deep learning methods for point cloud semantic segmentation calls for easy access of 3D data annotation tools that can further this research direction by enabling efficient labeling of large training datasets. There is no specialized annotation tool available that can be deployed on local machines for efficient annotation of large point clouds. The research community generally resorts to commercial [8], [9] or online [10], [11] annotation tools, which are cost prohibitive, and raise data privacy concerns

Methods

Results

Conclusion