Display QPI image data #146
Conversation
|
Progress so-far: I just need to generalise the methods and then this should be ready for review. (I'm sure I'll need some tests also)
|
There was a problem hiding this comment.
@paulmueller when you get a chance, pleae have a look at the changes :)
I tested that it works on normal rtdc files, and on rtdc files with qpi data.
Perhaps I need to add some tests?
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #146 +/- ##
==========================================
- Coverage 78.93% 78.71% -0.22%
==========================================
Files 63 63
Lines 6768 6807 +39
==========================================
+ Hits 5342 5358 +16
- Misses 1426 1449 +23 ☔ View full report in Codecov by Sentry. |
There was a problem hiding this comment.
The main issue that I see is that the current way contours are plotted is outdated and also does not work with phase data (changing pixels in an RGB image). So as far as I can see, there is no way around rewriting displaying the contour for phase and amplitude data.
Also, I believe there should be some way of setting the range for (at least) phase values. And phase values should always be centered at zero with the corresponding colormap showing white or no color. "coolwarm" from matplotlib is a good way of displaying phase, but you would have to copy-paste it to work with pyqtgraph, because I think it is not (yet) in pyqtgraph. "viridis" got ported to pyqtgraph, so it should be straight-forward.
shapeout2/gui/quick_view/qv_main.py
Outdated
| return cellimg, imkw | ||
|
|
||
| @staticmethod | ||
| def display_contour(ds, event, state, cellimg): |
There was a problem hiding this comment.
In the future, we might want to change how we compute/display the contour, i.e. we are planning to use non-discrete contours (@SaraKaliman). So it might make sense to implement two functions add_contour_to_image (for "image" data) and draw_contour_on_plot (for float32 qpi data) which could then also be used later for non-discrete contours.
There was a problem hiding this comment.
Practically I want to use shapeout to visualise the image data so Kyoo can look at the data in one place. For that, what I have built right now is enough. The colormaps and contour issues will just hold that up. Could we put all that in a separate PR?
There was a problem hiding this comment.
If it is just for Kyoo, then you could just create a custom executable for him:
cd build_recipes
pip install -r win_build_requirements.txt
pyinstaller -y --log-level=WARN win_ShapeOut2.spec
There was a problem hiding this comment.
I would not like to merge an incomplete feature. Moreover, the grayscale colormap leads to ambiguities. This should not be used by the general crowd.
There was a problem hiding this comment.
If you think it isn't a complete feature then I understand. I'll try to implement the colorbar etc soon.
Thanks for the custom executable, much better... I was just gonna do the dev version. :)
There was a problem hiding this comment.
Phase is back!
Phase values are inverted here (as they are saved inverted in the example file).
I will impement the contour utility methods that you mentioned above now. Contour looks a bit crazy right now!
With contours:
There was a problem hiding this comment.
Also, I believe there should be some way of setting the range for (at least) phase values.
Do you envisage something like like:
Or on a per-image basis? For per image, we can use the built-in histogram. However, I don't like this.
|
A really fun side-effect of having floating-point values in the image display is that we get a pinkish contour for the phase. :) I guess we have to fix this :')
EDIT: I (unfortunately) fixed the colour issue. I also confirmed that the colours are, in fact, the same! |
| cellimg[cont, 1] = int(imkw["levels"][0]) | ||
| cellimg[cont, 2] = int(imkw["levels"][0]) | ||
| else: | ||
| # just 2D images (not RGB) |
There was a problem hiding this comment.
This is a work in progress (probably going to follow your idea of having specific methods for contour handling).
|
|
||
| try: | ||
| # if we have qpi data, image might be a different shape | ||
| if "qpi_pha" in ds: |
There was a problem hiding this comment.
Maybe there should be a separate class entirely that handles image plotting? Generalization is not easy due to the different data types and requirements for visualization.
I think a simple method works for our use-case for image display. Could even make this into a method, seeing as we use it twice.
As described in (create issue) I want to see qpi data non-scalar features in shapeout2.