From microstructure to performance optimization: Innovative applications of computer vision in materials science

This review article explores how rapid progress in computer vision (CV), especially deep learning (DL)-based methods, has revolutionized materials science by shifting from traditional experimental approaches to more efficient, automated analysis of material microstructures. It covers four main application areas:

• Microstructure-based performance prediction Using convolutional neural networks (CNNs) and multimodal models to directly predict material properties (e.g., tensile strength, ionic conductivity) from microstructure images, achieving high accuracy and significant speedups over conventional methods.
• Microstructure information generation Employing generative models like GANs, VAEs, and diffusion models to create realistic synthetic microstructure images or simulate evolution under different conditions, helping overcome data scarcity.
• Microstructure defect detection DL frameworks (e.g., Mask R-CNN, Faster R-CNN) for automated, real-time identification and tracking of defects like voids or dislocations in microscopy images, improving quality control and reliability.
• Crystal structure-based property prediction Integration with graph neural networks (GNNs) like CGCNN or ALIGNN to forecast properties from atomic-level data, accelerating high-throughput material screening and design.

The paper highlights how combining microstructural and crystal-level CV/DL predictions bridges scales, reduces experimental costs, and drives innovation in next-generation materials. It also discusses challenges (e.g., dataset quality, model interpretability) and future directions, such as physics-informed models and greater interdisciplinary collaboration for sustainable materials development.