satellite-image-deep-learning/datasets

Datasets for deep learning with satellite & aerial imagery

What it solves

Finding the right dataset for a satellite or aerial imagery deep learning project can be time-consuming. This repository gathers many of these datasets in one place, so researchers and developers can quickly locate what they need without scouring multiple sources.

How it works

The project maintains a hand-curated list divided into sections: general dataset collections, remote sensing data hubs (like AWS Open Data and Google Earth Engine), and then dedicated lists for Sentinel-1 (radar) and Sentinel-2 (optical) imagery. Each entry includes a direct link and, where available, a short description of what the dataset contains or the task it addresses (e.g., change detection, super-resolution, classification). You can search the page with your browser's find function (Control+F) to locate a specific paper name or topic.

Who it's for

Researchers, data scientists, and machine learning engineers working on problems involving satellite or aerial imagery, such as land cover mapping, environmental monitoring, agriculture, disaster response, and object detection. It is especially useful for those who need to find benchmarks or specialized datasets quickly.

Highlights

  • Covers both Sentinel-1 (SAR) and Sentinel-2 (optical) imagery, plus general remote sensing resources.
  • Links to major data hubs like AWS Registry of Open Data and Google Earth Engine.
  • Includes datasets for many tasks: flood detection, crop classification, cloud removal, ship detection, forest monitoring, and more.
  • Provides direct links to papers and repositories for further context.
  • Simple searchable format—just use Control+F to find a specific dataset or paper.

Related

  • Project
    Azure/AzurePublicDataset

    A collection of public, production-scale traces from Microsoft Azure that provides researchers with real-world data on cloud workloads, serverless computing, AI inference, and agentic coding sessions.

  • Project
    mlmed/torchxrayvision

    TorchXRayVision is an open‑source Python library that standardizes access to many public chest‑X‑ray datasets and supplies pretrained deep‑learning models (DenseNet‑121, ResNet‑50, auto‑encoders, segmentation nets). It lets researchers load images, metadata, and pathology labels with a uniform API, apply built‑in preprocessing, and obtain pathology probabilities instantly, facilitating rapid prototyping, benchmarking, and few‑shot learning in medical imaging.

  • Project
    compdemocracy/polis

    Polis is an AI-powered sentiment gathering platform that lets participants vote and comment on statements, then uses a math engine to map opinion spaces. It's more organic than surveys and less effort than focus groups. It can be self-hosted via Docker, is free for nonprofits and government on the main site, and is licensed under AGPLv3 with additional permission.

  • Project
    SJTU-ViSYS/M2DGR

    A large-scale, multi-modal dataset for ground robot SLAM that provides rich sensor data across diverse, challenging real-world environments to help researchers test and improve navigation algorithms.