/AutoPhaseNN

Primary LanguageJupyter NotebookOtherNOASSERTION

AutoPhaseNN: Unsupervised Physics-aware Deep Learning of 3D Nanoscale Coherent Imaging

AutoPhaseNN is DL-based approach which gives direct inversion of the 3D BCDI data from the far-field measurement to the real space image. By incoorporating the forward physical model of the coherent diffraction imaging process, the model is trained with only measured diffraction patterns without needing real space images. https://www.nature.com/articles/s41524-022-00803-w

Requires

Tensorflow 2.x version

Trained model

The model trained on simulated data can be downloaded with the link: https://anl.box.com/s/qtdy2zl054i5zvoz7mg6hwsf0sqcmdnz The trained model is obtained with TF2.4.1 version. Please make sure using the same TF version to successfully load the model.

License

Copyright (c) UChicago Argonne, LLC. All rights reserved. See LICENSE file.