Monitoring grocery store shelf inventory with computer vision

A computer vision-based product identification pipeline

OneSix designed and implemented a robust computer vision-based product identification pipeline capable of analyzing high-resolution shelf images and delivering product-level insights at scale. The solution featured:

• Custom-trained computer vision models to detect, segment, and classify products in complex retail environments.
• High-speed vector search across tens of millions of product variations to rapidly match images to known SKUs.
• Rich metadata output, including product IDs and associated SKU-level details.
• Scalable production infrastructure, incorporating optimized models, graph-based search, and dynamic supporting databases.
• Comprehensive tooling, enabling internal teams and external customers to deploy product ID endpoints, create and manage product sets, and support new retailer onboarding.
• Advanced classification methods, including open-set and few-shot learning to recognize novel or rare products and identify out-of-set items for refinement.

Results

Faster, smarter grocery store shelf analytics

The end-to-end solution delivered measurable impact for grocery retailers:

• Reduced stockouts through accurate identification of missing or misaligned products.
• Improved planogram compliance via automated shelf verification from various image capture methods.
• Enhanced operational flexibility, enabling rapid enrollment of new products and seamless expansion across retailers.
• Multi-platform capability, with compatibility across mobile phone cameras, panoramic imaging, and robotic platforms.

By integrating computer vision, scalable infrastructure, and intuitive tools, OneSix built a flexible product identification platform that continues to power real-time shelf analytics and support evolving retail needs.