The GOOSE Dataset for Perception in Unstructured Environments
Authors:
Peter Mortimer,
Raphael Hagmanns,
Miguel Granero,
Thorsten Luettel,
Janko Petereit,
Hans-Joachim Wuensche
Abstract:
The potential for deploying autonomous systems can be significantly increased by improving the perception and interpretation of the environment. However, the development of deep learning-based techniques for autonomous systems in unstructured outdoor environments poses challenges due to limited data availability for training and testing. To address this gap, we present the German Outdoor and Offro…
▽ More
The potential for deploying autonomous systems can be significantly increased by improving the perception and interpretation of the environment. However, the development of deep learning-based techniques for autonomous systems in unstructured outdoor environments poses challenges due to limited data availability for training and testing. To address this gap, we present the German Outdoor and Offroad Dataset (GOOSE), a comprehensive dataset specifically designed for unstructured outdoor environments. The GOOSE dataset incorporates 10 000 labeled pairs of images and point clouds, which are utilized to train a range of state-of-the-art segmentation models on both image and point cloud data. We open source the dataset, along with an ontology for unstructured terrain, as well as dataset standards and guidelines. This initiative aims to establish a common framework, enabling the seamless inclusion of existing datasets and a fast way to enhance the perception capabilities of various robots operating in unstructured environments. The dataset, pre-trained models for offroad perception, and additional documentation can be found at https://goose-dataset.de/.
△ Less
Submitted 25 October, 2023;
originally announced October 2023.
Efficient Global Occupancy Map** for Mobile Robots using OpenVDB
Authors:
Raphael Hagmanns,
Thomas Emter,
Marvin Grosse-Besselmann,
Jürgen Beyerer
Abstract:
In this work we present a fast occupancy map building approach based on the VDB datastructure. Existing log-odds based occupancy map** systems are often not able to keep up with the high point densities and framerates of modern sensors. Therefore, we suggest a highly optimized approach based on a modern datastructure coming from a computer graphic background. A multithreaded insertion scheme all…
▽ More
In this work we present a fast occupancy map building approach based on the VDB datastructure. Existing log-odds based occupancy map** systems are often not able to keep up with the high point densities and framerates of modern sensors. Therefore, we suggest a highly optimized approach based on a modern datastructure coming from a computer graphic background. A multithreaded insertion scheme allows occupancy map building at unprecedented speed. Multiple optimizations allow for a customizable tradeoff between runtime and map quality. We first demonstrate the effectiveness of the approach quantitatively on a set of ablation studies and typical benchmark sets, before we practically demonstrate the system using a legged robot and a UAV.
△ Less
Submitted 8 November, 2022;
originally announced November 2022.
