Structural analysis and classification of silver–graphene oxide (Ag-GO)nanocomposites using SAM-based segmentation and machine learning models

Abstract

Nanocomposites have acquired significant interest in antimicrobial, catalytic, and electronic work due to their synergistic properties. The morphology and structural arrangements of the nanocomposites at the micro and nano scale are also very strong determinants of these functional properties. This study addresses the Structural Activity Relationship (SAR) of Silver-Graphene Oxide (Ag-GO) nanocomposites by employing a methodology to segment the nanocomposites and classify them. Scanning Electron Microscopy (SEM) images were used to segment nanocomposites using the Segment Anything Model (SAM) and extract structural features such as area, perimeter, aspect ratio, eccentricity, circularity to enable the classification of nanocomposites using various machine learning models. We have used supervised machine learning techniques, namely Random Forest, Logistic Regression, Decision Tree, K-Nearest Neighbors (KNN), and XGBoost. with these features in order to classify the nanocomposites. Thus, this study addresses a critical gap by incorporating automated segmentation and robust classification using a transparent performance evaluation framework, where XGBoost demonstrated the highest classification accuracy of 97% for shape and 95% for size, outperforming other models in identifying and categorizing different morphological patterns within Ag–GO nanocomposites.