awesome-satellite-imagery-datasets
A curated list of satellite and aerial imagery datasets with annotations for computer vision and deep learning tasks.
What is awesome-satellite-imagery-datasets?
Awesome Satellite Imagery Datasets is a curated, categorized list of publicly available datasets for training computer vision and deep learning models on satellite and aerial imagery. It addresses the problem of fragmented and hard-to-discover geospatial training data by providing a single, structured resource with detailed metadata for each dataset.
Researchers, data scientists, and machine learning engineers working on geospatial computer vision projects, such as object detection in satellite imagery, land cover classification, or disaster damage assessment.
It saves significant time in dataset discovery and evaluation by aggregating and organizing datasets from various sources (academia, competitions, government) with consistent, practical metadata, enabling faster prototyping and benchmarking in geospatial AI.
Overview
🛰️ List of satellite image training datasets with annotations for computer vision and deep learning
Use Cases
Best For
- Finding annotated satellite imagery datasets for object detection tasks like building or ship identification
- Sourcing training data for semantic segmentation of land cover or urban features
- Discovering datasets for academic research in remote sensing and computer vision
- Identifying benchmark datasets from competitions like SpaceNet or xView for model comparison
- Exploring multi-temporal or multi-sensor datasets for change detection or time-series analysis
- Locating specialized datasets for agriculture, forestry, or disaster response applications
Not Ideal For
- Teams needing datasets released after 2021, as the list is archived and may lack recent additions.
- Projects requiring datasets with guaranteed open commercial licenses, since licensing details aren't consistently provided.
- Developers looking for integrated tools or APIs for direct data downloading and preprocessing, as it's a static reference list.
Pros & Cons
Pros
Datasets are categorized by computer vision tasks like instance segmentation and object detection, as shown in the README sections, making it easy to find relevant data for specific ML problems.
Each entry includes key statistics, sensor details, and links to papers, providing essential context to assess dataset suitability, such as image counts and resolutions for models.
Covers diverse use cases from disaster assessment to agricultural monitoring, evidenced by datasets like FloodNet and PASTIS listed in the categories.
Part of the 'awesome' list ecosystem with community contributions, ensuring vetted and practical entries for researchers, as indicated by the structured updates and pointers.
Cons
The README explicitly states the list is archived with pointers to newer resources, meaning it doesn't include datasets published after its last update, reducing current relevance.
It only indexes datasets; users must navigate external sources for downloads, which can involve complex registration or lack of straightforward access, adding overhead.
While metadata is detailed, licensing terms for commercial use are often missing, requiring additional research on sources like iSAID's academic-only restrictions.
Frequently Asked Questions
Related Projects
awesome-public-datasetsA topic-centric list of HQ open datasets.
awesome-semantic-segmentation:metal: awesome-semantic-segmentation
Awesome GIS😎Awesome GIS is a collection of geospatial related sources, including cartographic tools, geoanalysis tools, developer tools, data, conference & communities, news, massive open online course, some amazing map sites, and more.
awesome-geospatialLong list of geospatial tools and resources
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.