At the moment the rs extract handlers are responsible to walk over OpenStreetMap and collect and extract GeoJSON features for the training dataset. Every handler has to start from scratch with OpenStreetMap data validation e.g. making sure polygons are closed ways, or that ways have more than a single node *.
We should write a base handler which already provides reasonable guarantees for the input data to work on.
At the moment the
rs extract
handlers are responsible to walk over OpenStreetMap and collect and extract GeoJSON features for the training dataset. Every handler has to start from scratch with OpenStreetMap data validation e.g. making sure polygons are closed ways, or that ways have more than a single node *.We should write a base handler which already provides reasonable guarantees for the input data to work on.
This base handler could also provide functionality to batch-write features; see https://github.com/mapbox/robosat/issues/120.
* Which is not always the case, e.g. see https://www.openstreetmap.org/way/5699705