Skip to main content

A labelling tool for satellite imagery

Project description

Annotate satellite imagery for training AI models.

Pip Version

SCANEO is a labelling tool for annotating satellite images. Its primary goal is to improve the efficiency of the creation of training datasets for AI models. To that end, SCANEO features an intuitive labelling interface and assisted labelling capabilities. Current capabilities include:

  • Create labelling campaigns from local or remote data
  • Annotate images for classification, detection and segmentation tasks
  • Assisted labelling with AI models
  • Export annotations in GeoJSON format
  • Integration with EOTDL.

Why SCANEO?

The development and deployment of AI models for Earth Observation (EO) applications faces significant challenges due to the scarcity of high-quality, well-annotated training datasets. This data bottleneck stems from several factors: the complex nature of satellite imagery, which requires domain expertise for accurate labeling; the high costs associated with data acquisition and manual annotation; and the technical challenges in preprocessing and standardizing multi-source EO data. These limitations have created a substantial gap between the potential applications of AI in Earth Observation and its current implementation, particularly when compared to more mature AI applications in other domains where training data is more readily available and easier to process.

To address some of these challenges, SCANEO provides a web-based annotation platform that combines artificial intelligence and active learning techniques to accelerate the creation of training datasets from satellite imagery.

Installation

We recommend using uv to install SCANEO.

uv init
uv add scaneo

Is is recommended to upgrade the package regularly, in order to get the latest changes.

uv add scaneo --upgrade

Usage

You can launch scaneo with a single command:

scaneo

This will launch the UI, which will be accessible on your localhost:8000.

To learn more about the CLI and options, run:

scaneo --help

Documentation and tutorials

To view the documentation, launch the UI and go to localhost:8000/docs.

Contributing

See CONTRIBUTING.md for more information.

Development

This repository contains the source code for SCANEO.

  • scaneo: includes the CLI and the API.
  • ui: includes the web application UI.
  • inference: includes a simple inference API.

First, clone the repository and install the dependencies:

git clone https://github.com/earthpulse/scaneo.git
cd scaneo
uv sync

Run the API with the CLI:

make dev

Then, run the UI:

cd ui
bun run dev

Remember to create the .env file from .env.example cp .env.example .env

Build

In the UI, create .env.production with an empty PUBLIC_API_URL.

Then, build the Python package:

make build v=0.1.0

The v parameter is the version of the package, change it to the desired version.

Inference

You can run the sample inference API to test the assisted labelling capabilities.

First, get the models and samples from EOTDL:

uv run eotdl models get SCANEO -p eotdl -a -f

Then, move the models and samples to the inference folder:

mv eotdl/SCANEO/models inference/models
mv eotdl/SCANEO/samples inference/samples

Finally, start the inference API:

make inference

You can test the models in the inference notebook.

Notes

Do not add scaneo/ui to gitignore since the build process will fail (missing entry folder).

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

scaneo-2025.6.23.tar.gz (19.1 MB view details)

Uploaded Source

Built Distribution

scaneo-2025.6.23-py3-none-any.whl (6.7 MB view details)

Uploaded Python 3

File details

Details for the file scaneo-2025.6.23.tar.gz.

File metadata

  • Download URL: scaneo-2025.6.23.tar.gz
  • Upload date:
  • Size: 19.1 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.5.6

File hashes

Hashes for scaneo-2025.6.23.tar.gz
Algorithm Hash digest
SHA256 3b66f575015bf32dccd3fee240d0e2d85e36e7ef2a57b9386d4a15d8911891bf
MD5 be76c96b6bb6274a2aa8c5caf8b97d32
BLAKE2b-256 02267f1688759ffa748bf1c6de2b5c928e9b3990d1e65b39dc46d46e3d2dc8c9

See more details on using hashes here.

File details

Details for the file scaneo-2025.6.23-py3-none-any.whl.

File metadata

File hashes

Hashes for scaneo-2025.6.23-py3-none-any.whl
Algorithm Hash digest
SHA256 866507cc8fde3232b36276e42f52762f16a20e0adb94baef9e8895f0d0136a00
MD5 c74c426b560bbda2bec8128e0ae72309
BLAKE2b-256 e24f42b063c92f2b04b4f42ccffc71a9759e4f8e79e8bed8b11b755c18aafcd0

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page