radionets
Imaging radio interferometric data with Neural Networks.
Science Score: 77.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
✓CITATION.cff file
Found CITATION.cff file -
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
✓DOI references
Found 1 DOI reference(s) in README -
✓Academic publication links
Links to: zenodo.org -
✓Committers with academic emails
8 of 14 committers (57.1%) from academic institutions -
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (18.0%) to scientific vocabulary
Keywords
Repository
Imaging radio interferometric data with Neural Networks.
Basic Info
- Host: GitHub
- Owner: radionets-project
- License: mit
- Language: Python
- Default Branch: main
- Homepage: https://radionets.readthedocs.io/en/latest/
- Size: 178 MB
Statistics
- Stars: 17
- Watchers: 1
- Forks: 4
- Open Issues: 10
- Releases: 10
Topics
Metadata Files
README.md
radionets 
Imaging Radio Interferometric Data with Neural Networks
Deep-learning framework for the simulation and analysis of radio interferometric data in Python. The goal is to reconstruct calibrated observations with convolutional Neural Networks to create high-resolution images. For further information, please have a look at our paper.
Analysis strategies leading to reproducible processing and evaluation of data recorded by radio interferometers:
* Simulation of datasets (see also the radiosim repository)
* Simulation of radio interferometer observations (see also the pyvisgen repository)
* Training of deep learning models
* Reconstruction of radio interferometric data
Installation
This repository is built as a python package. We recommend creating a mamba environment to handle the dependencies of all packages.
You can create one by running the following command in this repository:
$ mamba env create -f environment.yml
Depending on your cuda version you have to specify the cudatoolkit version used by pytorch. If you are working on machines
with cuda versions < 10.2, please change the version number in the environment.yml file. Since the package pre-commit is used, you need to execute
$ pre-commit install
after the installation.
Usage
For each task, executables are installed to your PATH. Each takes toml configuration files as input to manage data paths and options.
Simulated data is saved in hdf5; trained models are saved as pickle files.
radionets_simulations <...>This script is used to simulate radio interferometric data sets for the training of deep learning models.radionets_training <...>This script is used to train a model on events with known truth values for the target variable, usually Monte Carlo simulations.radionets_evaluation <...>This script is used to evaluate the performance of the trained deep-learning models.
Default configuration files can be found in the examples directory. The examples directory contains jupyter notebooks, which show an example
analysis pipeline and the corresponding commands. (need a rework)
Structure of the Repository
dl_framework
The used deep learning framework is based on pytorch and fastai. An introduction to Neural Networks and an overview of the use of fastai to train deep learning models can be found in Practical Deep Learning for Coders, v3 and fastbook.
dl_training
Functions for handling the different training options. Currently, there are the training, the learning rate finder, and the loss plotting mode available.
simulations (further developed in pyvisgen repository)
Functions to simulate and illustrate radio interferometric observations. At the moment simulations based on the MNIST dataset and
simulations of Gaussian sources are possible. We are currently working on simulating visibilities directly in Fourier space.
For more information, visit our corresponding repository pyvisgen. In the future, the simulations will be created
using the pyvisgen repository, while the radionets repository contains the training and evaluation methods.
evaluation
Functions for the evaluation of the training sessions. The available options reach from single, exemplary plots in (u, v) space and image space to methods computing characteristic values on large test datasets. In detail:
- Amplitude and phase for the prediction and the truth. Example image below includes the difference between prediction and truth.
- Reconstructed source images with additional features, such as MS-SSIM values or the viewing angle. Example image below.
- Histogram of differences between predicted and true viewing angles. The image includes a comparison with wsclean.
- Histogram of the ratio between predicted and true source areas. The image includes a comparison with wsclean.
- Histogram of flux difference in the core component. The image includes a comparison with wsclean.
- Included, but not yet fully operational
- Histogram of differences between predicted and true MS-SSIM values on a dedicated test dataset
- Histogram of differences between predicted and true dynamic range values on a dedicated test dataset
All histograms are created on a dedicated test dataset.
Contributors
- Kevin Schmidt @Kevin2
- Felix Geyer @FeGeyer
- Arne Poggenpohl @ArnePoggenpohl
- Stefan Fröse @StFroese
- Paul-Simon Blomenkamp @PBlomenkamp
- Olivia Locke @olivialocke
- Kevin Laudamus @K-Lauda
- Emiliano Miranda @emilianozm24
- Maximilian Büchel @MaxBue
- Rune Dominik @RuneDominik
Versions used and tested
- Python >= 3.8
- pyTorch >= 1.11.0
- torchvision >= 0.12.0
- cudatoolkit >= 11.3
Owner
- Name: Radionets-Project
- Login: radionets-project
- Kind: organization
- Email: kevin3.schmidt@tu-dortmund.de
- Location: Dortmund, Germany
- Repositories: 3
- Profile: https://github.com/radionets-project
Python packages for Machine Learning in radio interferometry.
Citation (CITATION.cff)
# This CITATION.cff file was generated with cffinit.
# Visit https://bit.ly/cffinit to generate yours today!
cff-version: 1.2.0
title: >-
radionets: Imaging Radio Interferometric Data with Neural
Networks.
message: 'If you use this software, please cite it as below.'
type: software
authors:
- family-names: Schmitz
given-names: Kevin
affiliation: 'TU Dortmund University, Germany'
orcid: 'https://orcid.org/0000-0002-9883-4454'
- family-names: Knierim
given-names: Anno
affiliation: 'TU Dortmund University, Germany'
orcid: 'https://orcid.org/0009-0004-8728-2879'
- family-names: Blomenkamp
given-names: Paul-Simon Wilhelm
affiliation: 'Ruhr University Bochum, Germany'
orcid: 'https://orcid.org/0000-0002-5032-5896'
- family-names: Fröse
given-names: Stefan
affiliation: 'Academia Sinica, Taiwan'
orcid: 'https://orcid.org/0000-0003-1832-4129'
- family-names: Geyer
given-names: Felix
affiliation: 'TU Dortmund University, Germany'
orcid: 'https://orcid.org/0000-0002-5615-2498'
- family-names: Locke
given-names: Olivia
affiliation: 'McGill University, Canada'
- family-names: Poggenpohl
given-names: Arne
affiliation: 'TU Dortmund University, Germany'
orcid: 'https://orcid.org/0000-0002-0746-4735'
- family-names: Zaldivar
given-names: Emiliano
affiliation: 'University of Hamburg,, Germany'
identifiers:
- type: doi
value: 10.1051/0004-6361/202142113
repository-code: 'https://github.com/radionets-project/radionets'
url: 'https://radionets.readthedocs.io/en/latest/'
keywords:
- Astronomy
- Data analysis
- Deep Learning
- Machine Learning
- Computer Vision
license: MIT
GitHub Events
Total
- Create event: 15
- Release event: 1
- Issues event: 3
- Watch event: 3
- Delete event: 12
- Issue comment event: 1
- Push event: 92
- Pull request review comment event: 1
- Pull request review event: 15
- Pull request event: 28
Last Year
- Create event: 15
- Release event: 1
- Issues event: 3
- Watch event: 3
- Delete event: 12
- Issue comment event: 1
- Push event: 92
- Pull request review comment event: 1
- Pull request review event: 15
- Pull request event: 28
Committers
Last synced: about 2 years ago
Top Committers
| Name | Commits | |
|---|---|---|
| Kevin Schmidt | k****t@t****e | 1,043 |
| Felix Geyer | f****r@t****e | 1,018 |
| Kevin Schmidt | k****t@u****u | 53 |
| Kevin Laudamus Kevin.Laudamus@studium.uni-hamburg.de | p****0@l****r | 38 |
| Kevin Schmidt | K****2 | 15 |
| Rune Michael Dominik | r****k@t****e | 12 |
| Arne Poggenpohl | a****e@a****d | 8 |
| Stefan Fröse | s****e@t****e | 8 |
| Maximilian Büchel | m****l@t****e | 3 |
| Emiliano Zaldivar Miranda emiliano.zaldivar.miranda@uni-hamburg.de | e****a@g****m | 2 |
| Emiliano Zaldivar Miranda emiliano.zaldivar.miranda@uni-hamburg.de | p****6@l****r | 2 |
| Arne Poggenpohl | a****l@t****e | 1 |
| Maximilian Büchel | m****l@t****e | 1 |
| Arne Poggenpohl | 3****l | 1 |
Committer Domains (Top 20 + Academic)
Issues and Pull Requests
Last synced: 6 months ago
All Time
- Total issues: 24
- Total pull requests: 119
- Average time to close issues: 6 months
- Average time to close pull requests: 8 days
- Total issue authors: 4
- Total pull request authors: 8
- Average comments per issue: 0.79
- Average comments per pull request: 0.67
- Merged pull requests: 98
- Bot issues: 0
- Bot pull requests: 5
Past Year
- Issues: 4
- Pull requests: 18
- Average time to close issues: 10 days
- Average time to close pull requests: 7 days
- Issue authors: 2
- Pull request authors: 4
- Average comments per issue: 0.0
- Average comments per pull request: 0.0
- Merged pull requests: 12
- Bot issues: 0
- Bot pull requests: 5
Top Authors
Issue Authors
- Kevin2 (13)
- FeGeyer (9)
- ArnePoggenpohl (1)
- aknierim (1)
Pull Request Authors
- FeGeyer (74)
- Kevin2 (22)
- aknierim (12)
- pre-commit-ci[bot] (2)
- RuneDominik (2)
- dependabot[bot] (2)
- ArnePoggenpohl (2)
- StFroese (1)
Top Labels
Issue Labels
Pull Request Labels
Packages
- Total packages: 1
-
Total downloads:
- pypi 220 last-month
- Total dependent packages: 0
- Total dependent repositories: 0
- Total versions: 5
- Total maintainers: 2
pypi.org: radionets
Imaging radio interferometric data with neural networks
- Homepage: https://github.com/radionets-project
- Documentation: https://radionets.readthedocs.io/
- License: MIT
-
Latest release: 0.4.1
published 7 months ago
Rankings
Dependencies
- actions/checkout v3 composite
- andstor/file-existence-action v2 composite
- actions/checkout v3 composite
- actions/setup-python v3 composite
- actions/setup-python v4 composite
- codecov/codecov-action v3 composite
- astropy *
- click *
- comet_ml *
- fastai *
- h5py *
- jupyter *
- kornia *
- numba *
- numpy *
- pandas *
- pre-commit *
- pytest *
- pytest-cov *
- pytest-order *
- pytorch-msssim *
- scikit-image *
- toml *
- tqdm *
- cartopy
- cudatoolkit
- numba
- numpy
- pip
- python
- pytorch