A modular software framework for the design and implementation of ptychography algorithms.

Area of Science:

• Computational microscopy
• Coherent diffractive imaging
• Image reconstruction algorithms

Background:

• Advanced microscopy techniques like ptychography rely heavily on computational methods.
• Developing and implementing new algorithms for these techniques is often time-consuming due to low-level programming requirements for performance.
• A need exists for accessible software frameworks that facilitate the design and testing of novel ptychography algorithms.

Purpose of the Study:

• To introduce SciComPty, an open-source software framework for simulating ptychography data.
• To enable efficient testing and development of existing and novel ptychography reconstruction algorithms.
• To leverage GPU acceleration for accelerating computational microscopy research.

Main Methods:

• Development of SciComPty, a Python-based software framework utilizing PyTorch and CUDA for GPU acceleration.
• Implementation of ptychography dataset simulation capabilities.
• Integration and adaptation of reconstruction algorithms, including an improved position refinement method (Adam) and a modified rPIE algorithm for partial coherence.

Main Results:

• Demonstration of SciComPty's utility in simulating ptychography datasets.
• Validation of implemented algorithms on both synthetic and real-world experimental data.
• Successful adaptation of algorithms for partial coherence scenarios.

Conclusions:

• SciComPty provides a user-friendly and efficient platform for advancing ptychography research.
• The framework's GPU acceleration significantly speeds up the development and testing cycle for new algorithms.
• Open-source availability promotes wider adoption and collaborative development in computational microscopy.