schreiber-lab / mlgixd

Deep learning-based feature detection for GIXD data

mlgixd

Deep learning-based feature detection for GIXD data

The package provides the source code used in the following scientific publication:

Starostin, V., Munteanu, V., Greco, A. et al. Tracking perovskite crystallization via deep learning-based feature detection on 2D X-ray scattering data. npj Comput Mater 8, 101 (2022). https://doi.org/10.1038/s41524-022-00778-8

Installation

Requirements

Python 3.6 or above and CUDA 10.2 or above are required. It is also recommended installing PyTorch with torchvision by following the instructions from the official website: https://pytorch.org/get-started/locally/ before installing the mlgixd package.

Other python dependencies can be installed automatically during the next step:

• numpy
• scipy
• tqdm
• PyYAML
• scikit-image

Install the package

To install the repository locally, execute the following command in the terminal to clone the repository and install it via pip:

git clone [email protected]:schreiber-lab/mlgixd.git && cd mlgixd && pip install .

or if pip is not available:

git clone [email protected]:schreiber-lab/mlgixd.git && cd mlgixd && python setup.py install

To test that the package is installed correctly, execute in the terminal from the package folder:

mlgixd config/test_config.yaml

or

python -m mlgixd config/test_config.yaml

The command should start short model training, testing and saving to a file without errors.

Train & test the models

To train and test one of the implemented models, use the corresponding configuration file provided with the package (for instance, config/our_model.yaml):

mlgixd config/our_model.yaml

The results will be saved to a file in saved_models folder specified in the configuration file and can be opened in python:

import torch

results = torch.load('saved_models/our_model.pt')

Authors

• Vladimir Starostin (Institute of Applied Physics, University of Tübingen)
• Valentin Munteanu (Institute of Applied Physics, University of Tübingen)
• Alessandro Greco (Institute of Applied Physics, University of Tübingen)
• Ekaterina Kneschaurek (Institute of Applied Physics, University of Tübingen)
• Alina Pleli (Institute of Applied Physics, University of Tübingen)
• Florian Bertram (Deutsches Elektronen-Synchrotron DESY)
• Alexander Gerlach (Institute of Applied Physics, University of Tübingen)
• Alexander Hinderhofer (Institute of Applied Physics, University of Tübingen)
• Frank Schreiber (Institute of Applied Physics, University of Tübingen)