View on GitHub

GuideLines For Labelling Agricultural Fields

Our project seeks to improve the quality of cropland maps, which requires high quality data to train and validate the models used to create the maps.

These guidelines are for the people who will be doing the labelling, which is undertaken through labeller, our informal name for the crowdsourced mapping platform we have developed. labeller provides a set of digitizing tools, a target region to label, and a view of the imagery that the algorithm will be processing.

Labellers draw boundaries around the fields that overlap the target, and which are visible within the imagery that is going to be processed.

To undertake labelling assignments, you have to receive an invitation email by a labeller employer or administrator, and then pass a brief qualification test (8 sites you have to map successfully) before you are eligible to undertake normal mapping assignments.

Please follow links to learn more about the mapping interface, mapping rules, etc.

The Labelling Interface

Labelling Rules