Deep-learning segmentation on organic particle assembly
발표자
()
초록
내용
Directly observing dynamic solution-state particles is very significant to understand their mutual interaction despite their fast Brownian motion and unpredictable interaction with solvent. Real-time transmission electron microscopy (TEM) imaging is on the rise as a robust methodology for analyzing their dynamic movement. In this work, I introduced U-net which has been commonly used deep learning model for image-processing to binarize images which was extracted from in-situ TEM video. With binarized images, the particle movement can be processed with ImageJ to get the trajectory of the nanoparticles. This research of imaging the small molecule by utilizing deep learning has the potential to advance the visualization of these systems, particularly through its ability to provide a 4D spatio-temporal representation and distinguish contrasts in a 3D manner and also furnish a framework for the imaging of small molecule interactions and effects on other systems, especially in liquid phase.