smpr3d 0.0.3

smpr3d is a toolkit for 3D reconstruction from scanning diffraction data

Welcome to Smpr3D!

Smpr3d (pronounced 'semper 3D', latin for 'always 3D', short for S-Matrix Phase Retrieval & 3D imaging) simplifies recovering 3D phase-contrast information from scanning diffraction measurements, such as those collected in 4D-Scanning Transmission Electron Microscopy (4D-STEM) experiments

🚨 Smpr3D is a research project at an early stage of development. Expect monstrous bugs and razor-sharp edges! A beta release is planned for M&M 2021.

Installing

TODO - this project is not on pip or conda yet - install with python setup.py develop --user

You can use Smpr3d without any installation by using Google Colab. In fact, every page of this documentation is also available as an interactive notebook - click "Open in colab" at the top of any page to open it (be sure to change the Colab runtime to "GPU"!).

You can install Smpr3d on your own machines with conda (highly recommended - not working yet :)). If you're using Anaconda then run:

conda install -c smpr3d -c pytorch -c anaconda smpr3d 

To install with pip, use: pip install smpr3d. If you install with pip, you should install PyTorch first by following the PyTorch installation instructions.

Hackathon - How to use on the nesap cluster

git clone [e-mail]:s-matrix/smpr3d.git

module purge

module load cgpu

module load pytorch/1.8.0-gpu

cd smpr3d

python setup.py develop --user

cd examples

sbatch slurm.sh

About Smpr3d

A fabulous idea

Acknowledgements

Hamish Brown (former NCEM) - theory and first demonstrations

Colin Ophus (NCEM) - theory and first demonstrations

Alex Rakowski (NCEM) - hpc

Jim Ciston (NCEM) - first demonstrations

Mary Scott (NCEM) - first demonstrations

Scott Findlay (Monash) - theory

Pierre Carrier (HPEnterprise) - performance profiling

Daniel Margala (NERSC) - performance profiling

References

Pelz, P. M. et al. Phase-contrast imaging of multiply-scattering extended objects at atomic resolution by reconstruction of the scattering matrix. (2021).

Brown, H. G. et al. A three-dimensional reconstruction algorithm for scanning transmission electron microscopy data from thick samples. doi <http://arxiv.org/abs/2011.07652>

How to contribute

Before committing, run

nbdev_build_lib && nbdev_clean_nbs && nbdev_build_docs

to compile the notebook into script files, clean the notebooks, and build the documentation.