Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OpenSatMap: A Fine-grained High-resolution Satellite Dataset for Large-scale Map Construction

Hongbo Zhao, Lue Fan, Yuntao Chen, Haochen Wang, yuran Yang, Xiaojuan Jin, Yixin Zhang, Gaofeng Meng, Zhaoxiang Zhang

TL;DR

By publishing and maintaining the dataset, this work provides a high-quality benchmark for satellite-based map construction and downstream tasks like autonomous driving, and covers and aligns with the popular nuScenes dataset and Argoverse 2 dataset.

Abstract

In this paper, we propose OpenSatMap, a fine-grained, high-resolution satellite dataset for large-scale map construction. Map construction is one of the foundations of the transportation industry, such as navigation and autonomous driving. Extracting road structures from satellite images is an efficient way to construct large-scale maps. However, existing satellite datasets provide only coarse semantic-level labels with a relatively low resolution (up to level 19), impeding the advancement of this field. In contrast, the proposed OpenSatMap (1) has fine-grained instance-level annotations; (2) consists of high-resolution images (level 20); (3) is currently the largest one of its kind; (4) collects data with high diversity. Moreover, OpenSatMap covers and aligns with the popular nuScenes dataset and Argoverse 2 dataset to potentially advance autonomous driving technologies. By publishing and maintaining the dataset, we provide a high-quality benchmark for satellite-based map construction and downstream tasks like autonomous driving.

OpenSatMap: A Fine-grained High-resolution Satellite Dataset for Large-scale Map Construction

TL;DR

By publishing and maintaining the dataset, this work provides a high-quality benchmark for satellite-based map construction and downstream tasks like autonomous driving, and covers and aligns with the popular nuScenes dataset and Argoverse 2 dataset.

Abstract

In this paper, we propose OpenSatMap, a fine-grained, high-resolution satellite dataset for large-scale map construction. Map construction is one of the foundations of the transportation industry, such as navigation and autonomous driving. Extracting road structures from satellite images is an efficient way to construct large-scale maps. However, existing satellite datasets provide only coarse semantic-level labels with a relatively low resolution (up to level 19), impeding the advancement of this field. In contrast, the proposed OpenSatMap (1) has fine-grained instance-level annotations; (2) consists of high-resolution images (level 20); (3) is currently the largest one of its kind; (4) collects data with high diversity. Moreover, OpenSatMap covers and aligns with the popular nuScenes dataset and Argoverse 2 dataset to potentially advance autonomous driving technologies. By publishing and maintaining the dataset, we provide a high-quality benchmark for satellite-based map construction and downstream tasks like autonomous driving.

Paper Structure

This paper contains 41 sections, 2 equations, 13 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. OpenSatMap Dataset
  4. Data Collection
  5. Geographic Diversity
  6. Special Road Structures
  7. Traffic Feature
  8. Alignment with Driving Benchmark
  9. Annotation
  10. Line representation.
  11. Line categories.
  12. Where should it be labeled?
  13. Fine-grained line attributes.
  14. Instances definition.
  15. Image-level tags.
  16. ...and 26 more sections

Figures (13)

  • Figure 1: Demonstrations of OpenSatMap dataset. It contains high-resolution satellite images with fine-grained annotations, covering diverse geographic locations and popular driving datasets caesar2020nusceneschang2019argoverse.
  • Figure 2: Alignment with driving benchmark.
  • Figure 3: Examples of categories and attributes.
  • Figure 4: Definition of instances (zoom in for best viewing). In (a), a change of line type results in two instances sharing the same point. In (b), a line should be divided into three instances when it is forked or merged.
  • Figure 5: Number of instances in each image in OpenSatMap19 (left) and OpenSatMap20 (right).
  • ...and 8 more figures