PoleDetection: A Comprehensive Dataset for Pole Detection and Localization Using LiDAR Imaging

Published: 5 November 2024| Version 1 | DOI: 10.17632/tt6rbx7s3h.1
Contributors:
Durga Prasad Bavirisetti,
,
,
,
,

Description

The PoleDetection dataset is a comprehensive collection of labeled LiDAR images specifically designed for pole detection in road environments. It was collected using a high-resolution OS2-128 LiDAR sensor and a GNSS system mounted on an autonomous vehicle, covering diverse environments such as mountainous, open, and forested areas. This dataset supports applications in computer vision and autonomous navigation, with a particular focus on pole detection and geospatial localization. The OS2-128 LiDAR sensor captured 360-degree images at the test location across four modalities: Near-IR, Signal, Reflectivity, and Range. To enhance usability, color images were generated by assigning the first three modalities (Near-IR, Signal, and Reflectivity) to the blue, green, and red channels, respectively, excluding the Range modality. Initial labeling was conducted using Roboflow, with further refinement in CVAT, resulting in high-quality annotations. The dataset comprises a total of 1,954 manually labeled images, divided into 1,564 training images and 390 validation images, following an 80/20 split. Since the images across all modalities are pixel-aligned, the labels for the color images are also applicable to each modality individually. This structure allows researchers to use the dataset directly for pole detection tasks, whether focusing on color or individual LiDAR modalities.

Files

Institutions

Norges teknisk-naturvitenskapelige universitet

Categories

Computer Vision, Image Processing, Geographic Information System, Object Detection, Autonomous Driving, Reflectivity, Lidar, Distance Measurement, Infrared Imaging, Range Image Processing

Funding

The Research Council of Norway

333875

Licence