Martin Raspaud

Threading might work, but I would recommend going all the way and using multiprocessing instead. Memory can be shared in threads, but in principle not with separate processes.

At the time, the hardest part was to find a minimal example to reproduce the error... It would be great if that could be done, so we can investigate further!

@yukaribbba that's really interesting! so that means that the image saving is the leaking?

very interesting! I have nothing against using GDAL directly as long the functionality is preserved. However I'm wondering if we are doing something wrong in riosave, or if the issue...

Interesting question! I you have sufficient information in your images to transform the physical values of reflectance pyspectral outputs to image (0-255) values, then in principle, in will be doable....

@guidocioni I don't think we are going to implement this ourselves: there is no need for it from our side, and on top of that, at the rgb workshop that...

@Berhinj thanks a lot for performing this investigation! this is indeed a bit worrying. I’m all for simplifying the code, because at the time we wrote this I don’t think...

Great investigative work @strandgren ! I'm in favour of replacing sunz correction wi the solar pathlength. However, replacing what "sunz_corrected" goes a bit too far imo, because I can think...

I agree with @simonrp84 in that data that is not for visualisation should not use the effective pathlength for normalisation. I’m just wondering how we can accommodate the two use...

> [@mraspaud](https://github.com/mraspaud) what do you mean with interactive use? Like this `scn.load(dataset, modifiers=["sunz_corrected"]`? yes