Electron Microscopy Image Segmentation for Nanoparticle Size Distribution Recognition

This paper presents an innovative approach for nanoparticle size measurement and shape classification using the Segment Anything Model (SAM). Nanoparticles have very important properties that have emerged as crucial elements in fields such as medicine, construction, and mechanical en-gineering. The study focuses on enhancing the accuracy of nanoparticle segmentation, particularly in challenging scenarios involving significant particle overlap. The methodology involves applying SAM to Transmission Electron Microscopy (TEM) images, using carefully tuned hyperparameters to optimize the model's performance. Our results demonstrate that SAM outper-forms traditional models like YOLOv8, particularly in accurately segmenting overlapping nanoparticles, leading to more reliable size and shape measurements.