Allow using image parameters for preprocessing #14
Conversation
|
@borondics, This now does something (labels are changed), and keeps image structure, so it is probably mostly OK. See:
This was this PR + Quasars/pySNOM/pull/19 |
markotoplak
force-pushed
the
use-image-params
branch
from
68b1e73 to
44b9465
Compare
|
This is done from my point of view. I also adapted two preprocessors to the new interface for demo. Please test and merge if OK. |
| def get_ndim_hyperspec(data, attrs): | ||
| # mostly copied from orangecontrib.spectroscopy.utils, | ||
| # but returns the indices too | ||
| # also avoid Orange domains as much as possible |
There was a problem hiding this comment.
I'm curious about this, is it faster? Should we just/also change the orangecontrib.spectroscopy.utils implementation? I think it used domains because get_column() didn't exist at that time.
There was a problem hiding this comment.
I did it because it seemed nicer to avoid domains where there are not needed. I do not expect any performance speedups, and yes, we could also do it in Spectroscopy.
|
@markotoplak thanks a lot! It mostly seems to work, but the BG Polyfit crashes. I sent the error report through Quasar. Could you have a look? Should we merge at this stage already? |
|
Also, there is a problem with the chunks, I think. Have a look at the picture below. I think what happens is that the median which we use for the linelevel correction gets computed for the chunks not the whole image. |
|
Merging for now. I don't see a nice way to chunk this now, if operations work like this then chunking is probably a no-go. |
No description provided.