Adding near field ptychography

[allevitan]

I'm starting some work to get a near-field ptychography model working. I thought about a few different ways to do it, and eventually I settled on the idea of making it just an option in fancy_ptycho. I think this has the main downside of making Fancy_ptycho more unwieldy, but it has many upsides:

• It actually turned out to be quite easy to add, the only differences were adjusting the propagator, the pixel size calculation, and the initialization
• All the same options (background, position refinement, etc) are immediately applicable
• A lot of near-field and far-field ptycho is taken in situations where either model could be applicable, so there are genuinely many situations where one might want to simply test turning on the "near-field" switch.

The way that it works at present is that near-field can be turned on in the FancyPtycho.from_dataset() factory function by setting the near_field argument to True. In that case, the code will either set up:

(1) Genuine near-field propagation with a 2-step Fresnel propagator if "propagation_distance" is set to 0
(2) Fresnel-scaling-theorem propagation with a 2-step Fresnel propagator if "propagation_distance" is finite

This meshes with "propagation_distance" being the initial guess of the illumination's propagation distance from focus. Essentially, if one has a reconstruction working in Far-field with a moderate "propagation_distance" argument, there's a chance you can just try turning on near_field=True and the reconstruction will still work (even if far field is still the appropriate method).

Right now, I've tested the Fresnel-scaling theorem part on some internal datasets that I unfortunately can't share here.

Todos

• Test Fresnel scaling theorem version internally, including oversampling
• Test genuine near-field internally, including oversampling
• Find test datasets for both cases that can be included in the CDTools repo (meeting size constraints and appropriately licensed)
• Write example scripts for near field
• Add test covereage for near-field mode

[allevitan]

I've now tested the near-field code on some near field data from a paper by Wenhui Xu. We are discussing the possibility of including this data in the repo and licensing it properly - if so, then I can add an example script using the data and that would leave only the test coverage

[allevitan]

I have added tests for near-field ptycho using the data here. This data comes from an exeriment I was helping with last month, and we got everyone's approval to release it with the proper CC license so we could use it as test near-field data.

If anyone has time to give this a review, you get cookies the next time I see you in person! Then we can stop saying "cdtools kiiiind of has near-field ptycho, but..."!!!

[yoshikisd]

We should also include the example near-field ptycho dataset, as it is currently missing from examples/example_data.

[allevitan]

We should also include the example near-field ptycho dataset, as it is currently missing from examples/example_data.

Whoops, I forgot to track the file in git. It is now added, thanks for the catch.

[yoshikisd]

Looks good for the most part. There's one suggested change and one question I have at the moment.

[allevitan]

Responded to the comments and made a commit. lmk how it looks :)

[yoshikisd]

This looks good to me!