Phaser: BCDI Phase retrieval in Python

Created by: Siddharth Maddali

Argonne National Laboratory

MOST RECENT UPDATES

• Genetic algorithms now properly parallelize on the GPU with MPI. Look at new script Guided_GPU.py for GPU usage of genetic algorithms.

Full changelog

1. Core.ImageRestart now automatically fftshifts input arrays.
2. Now suppressing Tensorflow messages in the command line (comment os.environ statement in Phaser.py to undo this).
3. GPUModule_Core.ImageRestart now shadows functionality of Core.ImageRestart.
4. Slight tweak to Guided.py logging.
5. New file Guided_GPU.py runs genetic algorithms using GPU + mpi4py.
6. ASCII-art of new logo in Phaser.py
7. PostProcessing.centerObject now removes phase ramps.

Introduction

• Basic Python tutorial of module Phaser for BCDI phase retrieval.

• Contains diffraction geometry modules for the 34-ID-C setup at the Advanced Photon Source.

  • Can be easily adapted to other geometries, please open an issue as a feature request if you need this done for your beamline.

• Modular, much simpler to use and modify than existing Matlab legacy code.

• Current dependencies (as determined by pipreqs)

mmatplotlib==3.1.3
scikit_image==0.16.2
tensorflow==2.2.0
mpi4py==3.0.3
tqdm==4.42.1
numpy==1.18.1
scipy==1.4.1
pyfftw==0.12.0
pyvista==0.32.0
pyvistaqt==0.5.0
skimage==0.0
vtk==9.0.3

• These modules are based on my current Python environment.
• All modules can be installed in the usual way: pip install <module>.
• The tensorflow 1.x-compatible library is available on branch tensorflow-1.x of this repo.

Quick start

A full tutorial on using Phaser to reconstruct your BCDI data is available here.