Accessing topological feature of polycrystalline microstructure using object detection technique

Faces-classes of grains, often referred to as topological features, largely dictate the evolution polycrystalline microstructures during grain growth. Realising these features is generally an arduous task, demanding sophisticated techniques. In present work, a distinct machine-learning algorithm extended for first time comprehend distribution grains constituting continuum. This regression-based object-detection approach, besides significantly reducing human-efforts and ensuring computational efficiency, predicts face-class by introducing appropriate ‘bounding boxes’. After sufficient training validation, over 500 epochs, current model exhibits remarkable overlap with ground truth that encompasses manually realised microstructures. Accuracy this treatment further substantiated relevant statistical studies including precision–recall analysis. The exposed unknown test dataset its performance assessed comparing predictions labelled Reflecting accuracy, strong agreement between algorithm-predictions noticeable in comparative involving systems varying number grains.