Speaker
Description
HoToPy is an open-source Python-based toolbox for X-ray holographic and tomographic data reconstruction. It features state-of-the-art phase retrieval algorithms for both the holographic and direct contrast imaging regimes, including nonlinear approaches and extended options for regularization and constraint sets. By combining automatic differentiation with provided fully differentiable forward operators, HoToPy enables straightforward extensions and modifications to the phase retrieval problem. Furthermore, the toolbox includes auxiliary functions for (iterative) tomographic alignment, image processing, and simulation of imaging experiments. Through its modular design and user-friendly interface, HoToPy can be easily integrated into existing pipelines, used as a building block for new applications, and applied in classroom teaching. Implemented using the PyTorch framework, HoToPy supports cross-platform, high-performance computations on both GPU and CPU, and enables seamless integration into machine learning workflows. We demonstrate HoToPy’s capabilities and discuss current developments and challenges in data reconstruction using the nano-tomography instrument ‘GINIX’ at the P10 beamline of the PETRA III storage ring (DESY, Hamburg).
