This repository is the official implementation of PID3Net: a deep learning approach for single-shot coherent X-ray diffraction imaging of dynamic phenomena.
Please cite us as
@article{Vu2025,
author = {Vu, Tien-Sinh and Ha, Minh-Quyet and Bachtiar, Adam Mukharil and Dao, Duc-Anh and Tran, Truyen and Kino, Hiori and Takazawa, Shuntaro and Ishiguro, Nozomu and Sasaki, Yuhei and Abe, Masaki and Uematsu, Hideshi and Okawa, Naru and Ozaki, Kyosuke and Kobayashi, Kazuo and Honjo, Yoshiaki and Nishino, Haruki and Joti, Yasumasa and Hatsui, Takaki and Takahashi, Yukio and Dam, Hieu-Chi},
doi = {10.1038/s41524-025-01549-x},
issn = {2057-3960},
journal = {npj Computational Materials},
number = {1},
pages = {66},
title = {{PID3Net: a deep learning approach for single-shot coherent X-ray diffraction imaging of dynamic phenomena}},
url = {https://doi.org/10.1038/s41524-025-01549-x},
volume = {11},
year = {2025}
}
We developed a Physics-Informed Deep learning Network for Dynamic Diffraction imaging (PID3Net) that takes advantage of a neural network to reconstruct the phase image of objects in Coherent x-ray diffraction imaging (CXDI) experiment.
PID3Net leverages established physical principles and utilizes experimental conditions to guide the network’s optimization. By introducing physics-based priors via the measurement-informed refinement block (RB) and the loss functions, PID3Net ensure that our reconstructions remain consistent with the underlying diffraction physics rather than relying solely on learned statistical patterns.
PID3Net is a self-supervised learning approach, where the network learns directly from the measured diffraction data without relying on external reference images or human-provided labels.
Figure 1 shows the overall schematic of the model.
Firstly, create a conda environment to install the package, for example:
conda create -n test python==3.9
source activate test
For hardwares that have CUDA support, the tensorflow version with gpu options should be installed. Please follow the installation from https://www.tensorflow.org/install for more details.
Tensorflow can also be installed from conda for simplification settings:
conda install -c conda-forge tensorflow-gpu
You can install the lastes development version of PID3Net from this repo and install using:
git clone https://github.com/sinhvt3421/PID3Net
cd PID3Net
python -m pip install -e .
The settings for experiments specific is placed in the folder configs
We provide an implementation for the Moving Ta test chart [1], gold nanoparticles AuNP [2] experiments.
[1] Tien-Sinh Vu, T., Takazawa & Chi, D. H. Diffraction intensity for phase retrieval of Ta test chart and AuNP dynamics motion using single-shot coherrent X-ray diffraction imaging. https://doi.org/10.5281/zenodo.12144981 (2024)
[2] Takazawa, S. et al. Coupling x-ray photon correlation spectroscopy and dynamic coherent x-ray diffraction imaging: Particle motion analysis from nano-to-micrometer scale. Phys. Rev. Res. 5, L042019 (2023) https://doi.org/10.1103/PhysRevResearch.5.L042019.