The 6th International Conference on X-ray and Neutron Phase Imaging with Gratings was successfully held from April 8th to 12th, 2024, at the Shenzhen Institute of Advanced Technology. The conference covered hot topics, frontiers, and challenges in the fields of grating fabrication, detectors and radiation sources, signal recovery and algorithms, life sciences and biomedical/clinical applications, methods and instrumentation, as well as materials science and industrial applications. It facilitated in-depth sharing and exchange, effectively promoting academic communication among international peers and the development of the field.
Xinyu Pan, a PhD student from the High Energy Photon Source (HEPS) beamline software group at the Institute of High Energy Physics, Chinese Academy of Sciences, attended the conference and presented a poster titled "Efficient Ptychography Reconstruction Strategy by Fine-tuning Pre-trained Large Models." The poster introduced the process of learning the mapping from coherent diffraction signals to real-space object distributions through convolutional neural networks. By optimizing the neural network architecture design, better reconstruction performance was achieved, and unsupervised fine-tuning techniques embedded with physical processes were used to enable neural network models for different experimental scenarios and data. Xinyu Pan introduced the entire process and answered questions from the audience (15 minutes). The presentation introduced the recent progress of the research group in six-dimensional tensor diffraction imaging methodology, the application of the methodology in the characterization of biomaterials, and looked forward to the application planning of the new methodology at new light sources. It actively sought cooperation with domestic peers, arousing widespread interest and attention from the conference experts.
