Modify the resampling method in load

[bcerripromethee]

Hi,

Is there a way to replace the resampling method, which is nearest, to bilinear (for example) in the load function ?

Actually when i use the function :
prod.load(band, window=window, resolution=10, resampling=Resampling.bilinear)

By doing this, i'm getting this error :
eoreader.utils.read() got multiple values for keyword argument 'resampling'

My first idea is to reach out your function in your code and modify it, but it's not really "clean", and i would have to reproduce it at each update of eoreader.

Do you have an idea ?
Thanks,
Bastien

[remi-braun]

Hi Bastien,

Are you sure the resampling method is set to nearest by default ?
This shouldn't be the case (see the _read_band method of the satellite you are using)

For the moment I am managing this by forcing a method (so I should document it) but the philosophy is:

• for bands that should be continuous, loaded in float (like spectral bands) : using bilinear method
• for masks and other classified rasters : nearest

[bcerripromethee]

Honestly, this is a quite difficult question to answer. Perhaps a good idea would be to add a parameter with a default value (=your current rounding). With this parameter, it will be easier for users to set it as they wish. No ?

[remi-braun]

The issue with this is I don't want to complicate EOReader too much (and to many parameters are hard to test).
I will leave it as it is, and wait for other people that may need similar parameters before acting (if it's fine with you)

[bcerripromethee]

Yeah, no problem with it, i totally understand !
When I see the Discussion topic, I can see that I spammed you a little bit, sorry for this 😅

[remi-braun]

In fact, I like those conversations because it always offers another perspective to my work and to other usecases 😉
I created an issue related to this rounding topic, don't hesitate to write a comment in it. (#105)
I won't do anything about that right now, even if I think this should be done (the reprojection shouldn't have values with another precision than the initial one!)