[Customer Success Story] How Korea Land & Housing Corporation Uses VLMs to Classify 18,000+ Housing Defect Combinations from a Single Image

Korea Land & Housing Corporation (LH), the country’s largest public housing provider, partnered with Superb AI to automate residential defect intake and classification using a Vision-Language Model that interprets photos and complaint text together. Facing roughly 300,000 defect images monthly, inconsistent classifications driven by varying agent experience, and a five-dimensional taxonomy producing more than 18,000 possible categories, LH adopted a single-model system that classifies space, materials, defect type, construction trade, and work type in real time. The system identifies issues such as cracks, leaks, mold, damage, and staining, provides natural-language explanations for its classifications, and was designed to meet targets of at least 90% detection accuracy, under 10% false positives and negatives, and processing in less than one second per image. Beyond reducing reliance on manual review and improving repair routing, the initiative creates structured defect data that LH can use to identify recurring problems, plan maintenance and budgets, and move toward preventive asset management, while offering a potential model for other public services handling large volumes of unstructured image-based reports.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.