For now, the postprocessing operation produce vectorized version of model predictions that erase the building type (hence one currently considers that every building is of ID 1, which is of course wrong).
The postprocessing code must take that point in charge, and produces .geojson files with proper building types.
For now, the postprocessing operation produce vectorized version of model predictions that erase the building type (hence one currently considers that every building is of ID 1, which is of course wrong).
The postprocessing code must take that point in charge, and produces
.geojsonfiles with proper building types.