[garrettwrong]

Reconciles some errors in the first Python implementation relative to the MATLAB.

• incorrect taper call
• model bounds
• adjusted defaults for number of tapers and psd size

Refactoring:

• Various optimizations.
• Adds BFGS (replacement for repeated gd loop, left for repro).
• Extends to any steerable basis (FFB and FLE under test, others are implemented via ctf tests extension).
• Implements optional 1D and 2D plots
• Moves most redundant function args to instance variables
• Migrates IO into helper function(s)
• Created main estimate method.
• Cleanup CLI and main method calls.
• Parameterize several internal hard-coded params/tols/stop conditions.

Still WIP, some todos.

• Need to meet with Ayelet to answer a few questions.
• Need to run against full experimental datasets to get results and times
• Try with K2 datasets and attempt to implement alt model for K2 if still needed now.

I will probably prefer make a separate PR handling any source integration changes to support a "from micrograph" preprocessing example. Currently this will produce the STAR files on disk which the sources were designed to consume. That design isn't very good IMO, but that's another matter... If source changes bleed into this it will make the PR a bear.

[codecov]

Codecov Report

❌ Patch coverage is 88.31522% with 43 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.44%. Comparing base (e4a05cc) to head (3da7678).

Files with missing lines	Patch %	Lines
src/aspire/ctf/ctf_estimator.py	89.50%	38 Missing ⚠️
src/aspire/commands/estimate_ctf.py	0.00%	5 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1416      +/-   ##
===========================================
- Coverage    90.65%   90.44%   -0.21%     
===========================================
  Files          135      135              
  Lines        14932    15044     +112     
===========================================
+ Hits         13537    13607      +70     
- Misses        1395     1437      +42     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[garrettwrong]

Ran against several full datasets. Times are great; defocus and astigmatism values look good. Angle appears to be impacted by a transpose somewhere. I'll see if I can sniff that out in the code, but is also easy enough to fix after computed as well. I used the logic below to adjust

-ang - 90 # = (90 - ang) -180 # Transpose axes before arctan - 180

With the various fixes and improvements in place the code appears to work for the majority of one of the problem K2 datasets now, but I will still look at making an alternative model for that case then reach out to Ayelet after.

[garrettwrong]

Angle appears to be impacted by a transpose somewhere.

Angle has been resolved in 43a7017.

I'm scheduling a follow up with Ayelet to cover the few questions I had remaining and discuss what I see regarding some of micrographs with poor solutions.... but it won't be for a little bit... Probably this can be reviewed and merged in by then; soI think it would be most efficient to review this larger PR now and just have a smaller one with any followup tweaks or changes for the problem K2 micrographs.

[j-c-c]

Looks good! Found a few typos (per my contractual obligation), but also a couple small potential bugs.

[j-c-c]

Looks good!

[garrettwrong]

Cool, thanks. @janden this one should be ready for you now.

[garrettwrong]

reminder

[garrettwrong]

Replace this with Image.load or Image._load_raw to handle the warnings from published data with junk headers...