Missing geospatial information after patched data

[nilsleh]

Hi @MWieland , thank you for this nice and interesting work. I noticed that in the patch split script the resulting tiff files do not contain the lat/lon or time coordinates anymore, and was wondering, whether it would be possible to adapt the script to include those values in the resulting tiff patches as well.

[MWieland]

Hi @nilsleh , thanks for your interest in our work and dataset. Here we decided to use plain numpy for the tiling as it is fast and simple and in most cases one does not require the geocoordinates during model training. You could try with rasterio windowed read, which should be straight forward to implement and would enable you to keep the geocoding of the tiles. See here for a possible solution (haven't tried it myself though). Alternatively, also gdal_retile should do the trick.

[nilsleh]

Thank you, will do so. Maybe as an additional question. The current script patches the S1 and S2 data separately (one has to run one after the other) but does that also guarantee that the corpus between the sensors is aligned?