Open Call / 2026 HOTOSM

Map your city with fAIr.

1000+

Training samples (target)

Open public outputs

City, mapped openly

The idea

You know what geographic data is missing to make better decisions. We support you to fill those gaps using AI-assisted mapping on drone imagery, with all results published as open data.

Selected projects receive funding depending on the proposal scope, plus full technical support from HOT. The AI side is on us. You define the feature, run the mapathon, and validate the output.

The pipeline

Step 01

Define the feature

Pick anything visible from drone imagery. Open spaces, wastebins, rooftops, solar panels, informal structures, street furniture, anything you can detect and want mapped.

Step 02

Feasibility study

Confirm enough samples exist (around 1000+ is a good target) and decide on imagery: fly your own drone or use existing drone imagery for the area.

Step 03

Run a Locate mapathon

Set up a Locate Objects project on MapSwipe and mobilise volunteers to label your feature. We help you scope and run the session.

MapSwipe / Locate Objects

Step 04

Use the model output

HOT supports you to train a fAIr model on your samples and shares the inference output. If results are weak, you can iterate with the team and collect more data.

Model training runs on HOT infrastructure

Step 05

Publish openly

Validate the model outputs and publish them as open mapping data. OpenStreetMap is the preferred destination. Where features need conflation (rooftops, for example), HDX or another public platform is accepted.

OSM preferred

What you get, what we expect

Project funding covering mapathon costs, validation effort, and team time.

AI support. We help you train the object detection model with fAIr on your data and share the inference outputs.

Technical guidance on MapSwipe Locate setup, sample sizing, validation, and OSM contribution.

Possible drone or expert support.Subject to availability. We may be able to send a team member or connect you with mentors.

A clear feature target visible in drone imagery, with around 1000+ expected samples (smaller cases considered).

Imagery access. Fly your own drone (or seek support from HOT Team) or use existing drone imagery covering the project area.

A local team or community able to run a Locate mapathon and validate the results.

Willingness to iterate with the HOT team if the model needs more data to perform well.

Open data commitment. All outputs published openly. Mapping features go to OpenStreetMap where feasible, or to HDX or another public platform when conflation or other constraints apply.

Final deliverables

Deliverable 01

Completed MapSwipe project

The Locate project results from your mapathon, finalised and contributed publicly through MapSwipe.

Deliverable 02

Open mapping data

Validated features published as open data. OpenStreetMap is the preferred destination. Where conflation or other constraints make OSM contribution unfeasible (rooftops, for example), HDX or another open public platform is accepted.

Sample Locate projects

Freetown

Locate Solar Panels - IV

A live MapSwipe Locate Objects project mapping solar panels from drone imagery.

Open project →

Freetown

Locate Trees - V

A live MapSwipe Locate Objects project mapping trees from drone imagery.

Open project →

FAQ

What is MapSwipe?

MapSwipe is a free mobile app that lets volunteers contribute to humanitarian mapping by swiping through satellite or drone imagery. You tap to mark where features appear, building a dataset that guides more detailed mapping work.

Why would I use MapSwipe in my project?

MapSwipe helps you annotate your drone imagery and create the training data used to train the fAIr model.

What is a Locate Objects project?

Locate Objects is a project type within MapSwipe. Instead of simply flagging broad areas, it asks volunteers to point to the exact position of a specific feature in each image tile, producing precise coordinates suitable for training an AI model.

What is fAIr?

fAIr is HOT's open-source AI-assisted mapping tool. It trains object detection models on community-labelled data and runs inference over drone imagery, producing feature predictions that mappers then validate and publish as open data.

Looking for AI model development instead?

This call is about running Locate mapathons and publishing open data. If you want to build the Earth Observation / GeoAI models themselves, see the open call for Earth Observation GeoAI models.

Can I use any drone imagery, including my own private footage?

No. Imagery used for the project must be licensed CC BY 4.0. The resulting mapped data follows the same open standard, published under ODbL.

Does the mapathon need to cover the entire city?

No. Partial coverage is fine, as long as the area is large enough to produce enough samples of your target feature.

Do I need to map the entire area of interest, or just the training area?

The entire area of interest must be fully mapped as your deliverable, regardless of how much area you used to collect training samples.

Is the 1000+ samples figure a hard target?

No. It can be adjusted depending on the use case, whether your feature is a generic, easy-to-detect case or a harder one.

Do I need one feature type, or can I map several?

Multiple feature types are possible, typically run as 2 to 3 separate MapSwipe projects. In that case, distribute your sample target evenly across the feature types.

When does the project need to be completed?

By November 2026 at the latest.

Are the features we can map restricted to buildings only?

No. You can be creative. Any feature that can be seen and classified from drone imagery is a valid target.

Send us your proposal.

Who Community mappers, urban researchers, students, civic tech teams, NGOs.

Budget $3,000 to $5,000 per selected project, depending on the proposal scope.

Deadline 30 September 2026Applications are reviewed on a rolling basis. The call may close early if applications exceed our review capacity, so apply sooner rather than later.

Apply Apply via Google Form

Contact fair@hotosm.org