astra-toolbox/astra-toolbox

What It Is

The ASTRA Toolbox is a free (open-source) software package that helps scientists and engineers reconstruct 3D images from 2D X-ray measurements — a technique called tomography. It is available as a library for Python and MATLAB.

What It Does

Supports multiple imaging setups: Parallel-beam and fan-beam for 2D scans; parallel-beam and cone-beam for 3D scans. It allows flexible positioning of the X-ray source and detector.
Provides ready-made reconstruction algorithms: FBP, SIRT, SART, and CGLS.
Uses GPU acceleration for the core forward projection (simulating what a scan would measure) and backward projection (building an image from measurements) — making these operations fast.
Lets you build custom algorithms: Because the projections are directly callable from Python and MATLAB, researchers can write their own reconstruction methods on top of these building blocks.

How You Get It

Several installation paths are offered:

Python via conda (recommended for Windows/Linux, 64-bit Python 3.9–3.13): conda install -c astra-toolbox -c nvidia astra-toolbox
Python via conda-forge (includes Linux ARM and Apple Silicon builds, CPU-only option available)
Python via pip (Linux only): pip install astra-toolbox, depends on NVIDIA CUDA packages from PyPI
Windows binary: Download a ZIP from the website, add its folders to your MATLAB path, or install the included Python wheel. Requires Microsoft Visual Studio 2017 redistributable.
From source (Linux/macOS/Windows): Detailed instructions are given for building with CUDA, MATLAB, Python, or as a standalone C++ library.

Testing Your Installation

Python: run import astra then astra.test()
MATLAB: run astra_test

License & Reference

The toolbox is open source under the GPLv3 license. The project asks researchers to cite two key papers (linked in the README) if they use it in publications.

The software is maintained by the Vision Lab at the University of Antwerp and CWI (Amsterdam), with contact via astra@astra-toolbox.com.