saugatkandel / ptychoSampling

Code for ptychography with tensorflow

ptychoSampling

Contains:

1. Contains demo code for the papers:

• Using automatic differentiation as a general framework for ptychographic reconstruction

and

• Efficient ptychographic phase retrieval via a matrix-free Levenberg-Marquardt algorithm

The optimization part of second-order optimization approach described in the latter paper is contained in https://github.com/saugatkandel/sopt as more general opitmization code, whereas this repository contains the wrappers for the ptycohgraphy applications.

2. Tutorials for simple ptychography reconstruction applications with tensorflow (contained in tensorflow_tutorials). It is safe to ignore the python package setup procedure.
3. A generic, modular tensorflow-based simulation and reconstruction framework in ptychoSampling. The documentation, however, is quit e sparse and sometimes unchanged from that for older versions of the code.
4. Simulation and reconstruction examples for far-field cases.

Warning:

The documentation is completely out-of-date.

Notes:

1. For ease of application, the forward model simulation code uses numpy. Only the reconstruction code uses Tensorflow.

2. Uses Tensorflow 1.14 for now. In the future, I am planning on switching away from the static computational graphs to a dynamic framework (Tensorflow 2.0, Pytorch, Autograd, Jax, etc) for ease of usage.