User bases¶

People meet fAIr at three levels. Each one draws from the same model collection and feeds something back, so the work keeps improving.

Mapper¶

"There is a model out there. Let me see how it works in my area."

A mapper runs an existing model and reviews what it finds.

1. Sees a model. Browses the collection for one that fits: buildings, trees, building damage, solid waste, and more.
2. Runs it. Points the model at their own area and lets it detect features.
3. Gets results. Predictions come back as GeoJSON, ready for the map.
4. Validates. Checks each result.

Validation can happen by hand, in the field, or at scale; see How it works.

Need a model for a different feature? Ask the team in #fair-coord on Slack.

Advanced user¶

"This is good, but tuned to my area it would work better."

An advanced user is a mapper who retrains a model for their own area and publishes the local model. Training works through the API today; the screens for it are planned.

1. Picks a model. Starts from one already in the collection.
2. Creates a training area. Sets up an area and maps example features there.
3. Maps it and retrains. Feeds the local examples back into the model.
4. Publishes. The local model returns to the collection for the next mapper.

The API steps for training and publishing are in ML pipeline.

Developer¶

"I built a model that maps a new feature. I want to share it."

A developer writes a new geo-AI model and adds it to fAIr.

1. Writes the model. Builds and trains a new model, for example one that maps a feature not yet covered.
2. Adds it to fAIr. Publishes the model to the collection with a clear name and the features it detects.
3. Available to everyone. From that point, any mapper or advanced user can run it.

See Contributing a model.