[BUG]: restrict_to_aoi not implemented

restrict_by_aoi is defined as a boolean argument in the raster tiler but doesn’t appear to be implemented. I was thinking of implementing this but instead of making it a boolean, the portion of the raster that falls within the aoi would remain and the portion of the tile that falls outside the aoi would be filled with the mean of the training dataset. I think this type of filling would work since labels are not present outside of the aoi_boundary (by definition of how aoi_boundary should be used). Does this sound good @nrweir ?

[yangsiyu007]

I just wanted to say filling in with custom values (0 in particular) would be great to have as one of the options at least. Computing the mean can be time-consuming over a large area, and being able to mask the label raster with 0 where the imagery tile has no data would be useful (i.e. to not have these contribute to the loss during training).

[nrweir]

Ideally, we enable both: allow users to specify a fill value (which they could derive from the full-dataset mean) or solaris could derive the mean within the image itself. Would essentially need to provide an argument specifying either a value to fill with, or allow users to specifiy 'mean', in which case solaris will get the image’s mean value and use that. e.g. a new argument nodata_fill=None where the user can provide, for example, nodata_fill=50 (or whatever other specific value they want) or nodata_fill='mean' in which case solaris will get it from the image for them.