Solutions: AI Workflows
Retail Loss Prevention
Build scalable solutions for product detection and recognition to prevent shrinkage at the point of sale.
The Loss Prevention Challenge
Retail product shrinkage is a $100 billion per year challenge, with over 65% of shrinkage—and growing—due to theft. With NVIDIA’s retail loss prevention AI workflow, you can quickly build and deploy applications designed to prevent theft. Leveraging the NVIDIA Metropolis microservices and models pretrained on hundreds of the most frequently stolen goods, this AI workflow provides few-shot, active learning to quickly scale to recognize hundreds of thousands of products and provide intelligent alerts with actionable information.
Explore the Retail Loss Prevention AI Workflow
Built on cloud-native NVIDIA Metropolis microservices—a low- or no-code way to build AI applications—this reference workflow for retail loss prevention and frictionless shopping delivers pretrained AI models along with the applications needed to jump-start development and rapidly index hundreds of thousands of store products for cross-camera and barcode-scan identification.
The workflow contains:
- Pretrained models for retail product recognition.
- Similarity search for accurate and fast retrieval.
- Few-shot learning for ease of expansion to hundreds of thousands of products and their variants.
- A reference solution for deploying in production, including barcode scanner comparison.
- Guidance on how to train and customize the solution for a specific use case.
The AI models in this workflow are pretrained to recognize the hundreds of products most frequently lost to theft—including meat, alcohol, and laundry detergent—and to identify them in varying sizes and shapes. With synthetic data generation from NVIDIA Omniverse™, retailers and independent software vendors can customize and further train the models to cover hundreds of thousands of store products. The workflow is based on a state-of-the-art, few-shot learning technique developed by NVIDIA Research that, combined with active learning, identifies and captures any new products scanned by customers and sales associates during checkout to ultimately improve model accuracy.
Decrease Shrinkage With Vision AI-Based Loss Prevention
Key Benefits of the Retail Loss Prevention AI Workflow
Pretrained Models
The AI models in this workflow are pretrained to recognize hundreds of products most frequently lost to theft—including meat, alcohol, and laundry detergent—and to identify them in varying sizes and shapes.
Few-Shot Active Learning
The workflow features a state-of-the-art variation of few-shot learning designed to adapt continuously with limited new product data using object characterization and self-supervised learning algorithms. This unique method of active learning identifies and captures new products and packaging changes scanned during checkout for future recognition through similarity search.
Flexible Reference Architecture
Delivered through cloud native microservices, this AI workflow allows users to jump-start development, easily customize for their solution, rapidly index hundreds of thousands of store products for cross-camera and barcode-scan identification, and enable scalable production deployment.
Accelerate the Development of AI Solutions
AI workflows accelerate the path to AI outcomes. This AI workflow provides a reference for developers to rapidly start creating a flexible and scalable loss prevention solution.
