 |
TOmographic MOdel-BAsed Reconstruction software PAPER (CT Meeting 2020)
ToMoBAR is a Python library of fast direct and model-based regularised iterative algorithms with a plug-and-play capability for reconstruction of parallel-beam geometry data. ToMoBAR offers you a selection of various data models and regularisers resulting in complex objectives for tomographic reconstruction. ToMoBAR can handle multi-GPU parallel reconstruction in Python and also device-to-device methods operating on CuPy arrays. It is currently used in production at Diamond Light Source as a part of the HTTomo framework for big-data processing and reconstruction. |
| Master | Anaconda binaries |
|---|---|
 |
    |
-
$\sf\color{red}!$ To better communicate breaking changes, ToMoBAR is moving from calendar versioning to semantic versioning. The 2026.1.0.0 release is based on the 2025.12 version. We keep the year in order to make the sorting work, so we'd have the following structure: year.major.minor.patch. - There are
$\sf\color{red}BREAKING$ $\sf\color{red}CHANGES$ from ToMoBAR$\sf\color{red}v.2025.08$ , see CHANGELOG for all detailed changes.
Check what ToMoBAR can do. Please also see Tutorials and Demos. ToMoBAR
Please check the detailed installation guide where all software dependencies are listed.
- Wrappers around ASTRA-toolbox to simplify access to various reconstruction methods available in ASTRA-Toolbox
- Optimised CUDA/CuPy implementation of the fast Log-Polar (Fourier-based) direct reconstruction method.
- Regularised iterative ordered-subsets FISTA reconstruction algorithm with linear and non-linear data fidelities
- Regularised iterative ADMM reconstruction algorithm
- CuPy driven forward/backward projectors to enable faster device-to-device operations and all-in-GPU memory prototyping of algorithms.
- Demos to reconstruct synthetic and also real data [4-6]
