The crowdsourcing platform for collecting training data and initiating the ML process within the active learning framework.

This version evolved from the original version called DIYlandcover (Estes et al, 2016), that was designed to connect to Amazon's Mechanical Turk workforce. It was re-engineered into mapper, a standalone version with its own worker interface and (simple) management system. labeller is a lightweight version of mapper, renamed to better describe its purpose.

Documentation is in the process of being updated. Here are a few pointers to some of it.

• A description of labeller's components, as originally designed.

• A description of labeller's database.

• How labeller and learner interact in the active learning process.

• Image acquisition and processing

• Segmentation

• Building a labeller instance from scratch

• Cloning an existing labeller instance and using that to create a new instance