ne1s0n
commented
3 days ago This would probably be done easier if we introduce a multi-sensor image class, which just reads everything and takes care of alignment and different resolutions. It can easily be extended in the .ini as:
orthomosaic_1=path/to/file1.tif
channels_1=red,green,blue
orthomosaic_2=path/to/file2.tif
channels_2=blue,green,red,rededge,NIR
with some check on conflicting channel names
Very interesting options will open if we are able to combine multiple sources in a single computation. However, the current implementation in drone2report.py goes single-dataset -> single-task.
Moreover, this kind of task are bound to be sensor-specific. Should we add a target field in the new multi-sensor task?
Datasets will also need to be reported at the same size to be able to do a pixel-by-pixel comparison