Machine Learning‐Aided Spatial Adaptation for Improved Digital Image Correlation Analysis of Complex Geometries

Digital image correlation (DIC) is a widely used experimental technique for measuring full‐field deformation, but its application to complex scenarios involving large deformations, discontinuities, or intricate geometries is often hampered by the need for manual region of interest (ROI) definition. This limitation is particularly acute for incremental tracking strategies, where frequent ROI updates create a significant bottleneck and require substantial user intervention. To overcome this challenge, we present a machine learning‐aided workflow that automates and accelerates the analysis pipeline. Our approach leverages the Segment Anything Model 2 (SAM 2) for rapid initial mask generation, followed by novel spatial and temporal smoothing filters to denoise and refine the ROI sequence. These high‐quality masks are then seamlessly integrated into our SpatioTemporally Adaptive Quadtree‐mesh DIC (STAQ‐DIC) method, which performs automated adaptive meshing and subset splitting near complex boundaries. Through multiple challenging case studies, we demonstrate that our method reduces processing time by one to two orders of magnitude compared to manual methods and shows superior scalability over other automated techniques. The workflow enables robust and accurate full‐field measurements in scenarios that were previously intractable or required extensive manual effort. To facilitate broad adoption and benefit the experimental mechanics community, we have integrated the entire pipeline into a user‐friendly, open‐source code package with graphical user interfaces (GUIs).