From Classic Computer Vision to Large Vision Models: MIPAR's Integrated Solution for Materials and Manufacturing Imaging Challenges

Advancing materials and manufacturing image analysis, MIPAR has released Spotlight, an AI-powered extension that leverages zero-shot generalization for enhanced feature detection and segmentation. Building on our technical foundation since 2017, Spotlight integrates with our comprehensive technology stack, combining deep learning and traditional computer vision capabilities. This framework enhances micrograph analysis while eliminating manual labeling requirements. A key innovation in our workflow is using deep learning model outputs to generate region-specific prompts for Spotlight—effectively prompting AI with AI. In this approach, one AI model identifies regions of interest that serve as visual prompts directing Spotlight where to perform segmentation, creating a powerful analytical pipeline. Our presentation will demonstrate this technical integration through materials manufacturing applications, showing how this AI-to-AI prompting improves grain size measurement precision, enhances defect detection sensitivity, and increases accuracy in dimensional analysis of microstructural features. These examples will illustrate how our complete technology stack, from classical computer vision to next-gen large vision models, maximizes the potential for solving and automating common materials characterization challenges.