Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Panoptic-CUDAL: Rural Australia Point Cloud Dataset in Rainy Conditions

Tzu-Yun Tseng, Alexey Nekrasov, Malcolm Burdorf, Bastian Leibe, Julie Stephany Berrio, Mao Shan, Zhenxing Ming, Stewart Worrall

TL;DR

Panoptic-CUDAL fills a critical gap by providing a rural, rainy autonomous-driving dataset with high-resolution $128$-channel LiDAR, surround-view cameras, and precise pose data, all annotated for panoptic segmentation in the SemanticKITTI format. The authors establish baselines for semantic segmentation, panoptic segmentation, and 3D occupancy prediction using a mix of projection-based, transformer-based, and hybrid methods, revealing notable challenges in rural rain scenarios and rare classes. The dataset enables robust evaluation and development of perception systems that must operate under adverse weather and non-urban settings, with concrete findings on model performance and the benefits of LiDAR-camera fusion. Overall, Panoptic-CUDAL advances rural-weather perception research and provides a valuable resource for benchmarking and method development in this underexplored domain.

Abstract

Existing autonomous driving datasets are predominantly oriented towards well-structured urban settings and favourable weather conditions, leaving the complexities of rural environments and adverse weather conditions largely unaddressed. Although some datasets encompass variations in weather and lighting, bad weather scenarios do not appear often. Rainfall can significantly impair sensor functionality, introducing noise and reflections in LiDAR and camera data and reducing the system's capabilities for reliable environmental perception and safe navigation. This paper introduces the Panoptic-CUDAL dataset, a novel dataset purpose-built for panoptic segmentation in rural areas subject to rain. By recording high-resolution LiDAR, camera, and pose data, Panoptic-CUDAL offers a diverse, information-rich dataset in a challenging scenario. We present the analysis of the recorded data and provide baseline results for panoptic, semantic segmentation, and 3D occupancy prediction methods on LiDAR point clouds. The dataset can be found here: https://robotics.sydney.edu.au/our-research/intelligent-transportation-systems, https://vision.rwth-aachen.de/panoptic-cudal

Panoptic-CUDAL: Rural Australia Point Cloud Dataset in Rainy Conditions

TL;DR

Panoptic-CUDAL fills a critical gap by providing a rural, rainy autonomous-driving dataset with high-resolution -channel LiDAR, surround-view cameras, and precise pose data, all annotated for panoptic segmentation in the SemanticKITTI format. The authors establish baselines for semantic segmentation, panoptic segmentation, and 3D occupancy prediction using a mix of projection-based, transformer-based, and hybrid methods, revealing notable challenges in rural rain scenarios and rare classes. The dataset enables robust evaluation and development of perception systems that must operate under adverse weather and non-urban settings, with concrete findings on model performance and the benefits of LiDAR-camera fusion. Overall, Panoptic-CUDAL advances rural-weather perception research and provides a valuable resource for benchmarking and method development in this underexplored domain.

Abstract

Existing autonomous driving datasets are predominantly oriented towards well-structured urban settings and favourable weather conditions, leaving the complexities of rural environments and adverse weather conditions largely unaddressed. Although some datasets encompass variations in weather and lighting, bad weather scenarios do not appear often. Rainfall can significantly impair sensor functionality, introducing noise and reflections in LiDAR and camera data and reducing the system's capabilities for reliable environmental perception and safe navigation. This paper introduces the Panoptic-CUDAL dataset, a novel dataset purpose-built for panoptic segmentation in rural areas subject to rain. By recording high-resolution LiDAR, camera, and pose data, Panoptic-CUDAL offers a diverse, information-rich dataset in a challenging scenario. We present the analysis of the recorded data and provide baseline results for panoptic, semantic segmentation, and 3D occupancy prediction methods on LiDAR point clouds. The dataset can be found here: https://robotics.sydney.edu.au/our-research/intelligent-transportation-systems, https://vision.rwth-aachen.de/panoptic-cudal

Paper Structure

This paper contains 12 sections, 10 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Dataset
  4. Data Collection Vehicle.
  5. Location.
  6. Data Post-processing.
  7. Data Annotation
  8. Experiments and Results
  9. Task and Metrics
  10. Baseline Approaches
  11. Experimental Results
  12. Conclusion

Figures (10)

  • Figure 1: Front-view image from the Panoptic-CUDAL dataset, with projected LiDAR point clouds.
  • Figure 2: The LiDAR is located on the vehicle's roof used to collect the data.
  • Figure 3: Surround-view images Our vehicle is equipped with 8 cameras ensuring 360$^{\circ}$ view of the environment.
  • Figure 4: Camera configuration Our vehicle is equipped with cameras ensuring 360$^{\circ}$ view of the environment. The blue and green colour represents the 60$^{\circ}$ and 120$^{\circ}$ FOV of the cameras, respectively.
  • Figure 5: Perception challenges for rural areas. The recorded data contains multiple examples of challenging conditions, rarely observed in other datasets.
  • ...and 5 more figures