search

CosmoStat / wf-psf

Data-driven wavefront-based PSF modelling framework.
MIT License
19 stars 9 forks source link

Applying WaveDiff to Experimental Data with Custom Parameters #141

Open CAClaveau opened 1 day ago

CAClaveau commented 1 day ago

Hello,

Thank you for sharing your WaveDiff tool! The performance reported in your papers is very impressive!

I would like to use it for one of my projects on the Keck telescopes, both for PSF reconstruction and phase retrieval. I would have a few questions, please:

  1. Is wf-diff at a stage where it is ready to be applied to any experimental data? I’m happy to test and adapt the code as needed, fine-tuning the hyperparameters to ensure it works effectively with my specific data.

  2. Sorry if I missed it, but do you have a notebook or tutorial that, in addition to the "configuration" section of your documentation, provides a step-by-step example of applying wf-diff to real data?

  3. Unless I'm mistaken, I couldn't find in the config section where I can modify and adapt certain key parameters specific to my system, such as pupil shape/obscuration, focal length/aperture diameter/pixel size (which I think are also used internally to determine the set of wavelengths "allowed" and satisfy the conditions of your zero-padding/resampling strategy, etc.) I believe these parameters are given in the SimPSFToolkit class and are currently set to represent the Euclid optical system.

  4. Is it possible in any version of wf-psf to utilize information obtained via phase diversity (from out-of-focus PSFs)?

  5. Is your new strategy for improved phase retrieval presented in one of your posters (link](https://drive.google.com/file/d/1WJ7qFlLHuAHlyDDU4vFAc-6MYIzsDMGA/view?usp=share_link)) available in any of the published versions of wf-psf?

Thank you for your time and consideration!

tobias-liaudat commented 1 day ago

Hi @CAClaveau

Thanks for your interest in the method. I have provided some fast answers to your questions in the text below. However, if you're interested in going more into the details of the model and how to use/adapt it, we could program a Zoom call (send me an email to tobias.liaudat@cea.fr). I could assist you in adapting the code for your needs.

  1. We are currently working with real Euclid data, so the answer is yes. However, we are still analysing the data and adding more features to the model to improve its performance. We have not tested the model with ground-based observations, so I'm very interested to see how it performs!

  2. Unfortunately, we do not have one notebook like that. Nevertheless, if there's interest, we could build one!

  3. We have not exposed those parameters to the config files as we were focusing on Euclid (so these parameters were fixed). In any case, there is nothing stopping us from exposing them and adapting them for another telescope.

  4. Yes, this is one of the updates that are not yet published. The idea is to use wavefront information (obtained from simulations or via a phase diversity study) as a prior in the model and take it into account while optimising the model. This information should follow the wavediff format. The model that handles a wavefront prior is the poly_physical model.

  5. Yes, the strategy for phase retrieval is implemented. A paper describing that method in detail is being prepared. However, those updates have not yet been merged into the main, but are available at the branch case_study_psf_decontamination.

Note that several new features of the model are not described in a paper as we are focused on processing Euclid data, which was the motivation for the development of the WaveDiff model.