Indoor Navigation Using AI + 3D Mapping
The Problem
Customers in large grocery stores waste significant time searching for products. Staff are constantly interrupted with "Where is X?" questions, pulling them away from other tasks. Traditional store directories are static and quickly become outdated as products move. The client wanted a digital solution that could guide customers to any product instantly, without requiring manual data entry for thousands of SKUs or expensive hardware installation. The challenge was finding a way to automatically build and maintain an accurate product location database from existing store infrastructure.
The Solution
We leveraged the store's existing Matterport 3D scan — a virtual walkthrough already used for marketing purposes. Our pipeline extracts still images from specific viewpoints throughout the 3D model, then runs OCR (using Google Document AI and Tesseract) to detect and read product names visible on shelves. Each detected product is mapped to its corresponding 3D coordinates within the Matterport space. We built a search interface where customers type a product name, and the system displays an interactive 3D path from their current location to the product — complete with turn-by-turn visual directions similar to Google Maps navigation.
The Result
The system successfully identified and mapped over 2,000 products from a single store scan. Customer wayfinding time dropped dramatically — products that previously took minutes to find were located in seconds. Staff interruptions for location questions decreased significantly. The solution requires no additional hardware, works on any smartphone browser, and updates automatically when the store is rescanned. This experiment proved that combining existing 3D capture technology with AI-powered OCR creates a practical, scalable indoor navigation solution without the cost of traditional beacon or sensor-based systems.