Nederlands

Tomosipo: Pythonic high-performance GPU primitives for 3D tomography with complex geometries

Tomosipo is a pythonic interface for ASTRA Toolbox focusing on high-performance GPU primitives for 3D tomography. It provides a convenient and powerful tool for modeling and visualizing complex data acquisition geometries and prototyping high-performance reconstruction algorithms.

Tomosipo is a pythonic wrapper for the ASTRA Toolbox focusing on high-performance GPU primitives for 3D tomography.

The aim of this library is to:

  • Expose a user-friendly API for high-performance 3D tomography, while allowing strict control over resource usage
  • Enable easy manipulation and visualization of complex 3D geometries
  • Provide easy integration with GPU-accelerated algebra and Deep Learning libraries, such as PyTorch and CuPy.

This allows the user to easily prototype reconstruction algorithms for various experimental setups by utilizing building blocks based on different computational frameworks without sacrificing performance.