[17p-S2_204-1~18] 23.1 Joint Session N "Informatics"

5:30 PM - 5:45 PM JST(8:30 AM - 8:45 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[17p-S2_204-15] Deep Learning Models for High-Accuracy Estimation in Low-Electron-Dose Electron Microscopy

〇Naomu Sekiguchi¹, Jiaxin Chen¹, Qing Wang¹, Yusuke Shimada¹, Satoshi Iikubo¹ (1.Kyushu Univ.)

Keywords:

electron microscope,organic-inorganic perovskite,deep learning

For perovskite materials that require low-electron-dose conditions in electron microscopy, we developed a deep-learning reconstruction model based on WGAN-GP to accurately recover lattice images from weak-signal, noise-prone data. As a result, atomic column positions were clarified in low-dose simulated images of FASnI3 and in experimental images of CsPbBr3, confirming that the structural information can be effectively restored.