Nvidia Introduces New AI Workflows to Help the Retail Industry Loss Prevention • TechCrunch

Nvidia Introduces New AI Workflows to Help the Retail Industry Loss Prevention • TechCrunch

Nvidia Introduces New AI Workflows to Help the Retail Industry Loss Prevention • TechCrunch

In a recent episode of “Customer Wars”, a woman put a chainsaw in her pants in an attempt to steal it. And you may have caught a video of thieves robbing Ulta.

Videos like this are all over the internet and the retail industry is reporting that thefts are on the rise. Target attributed losses of hundreds of thousands of dollars in 2022 to organized retail thefts, while Walmart recently said increased thefts could result in higher prices and store closures.

“Shrunk” is a term used by retailers to describe loss due to theft, damage, or misplacement of a product. The National Retail Federation said shrinkage has increasingly become an issue for $100 billion retailers over the last five years. Digging into the shrinkage, the NRF says about 65% of it is due to theft itself.

Go to Nvidia. Known for developing and manufacturing GPUs, the company announced three new retail AI workflows as part of its NVIDIA AI Enterprise software suite. These workflows are designed to help developers build and deploy apps that are designed to prevent theft faster.

The workflows are built on Nvidia’s Metropolis Microservices, a low-code way to build AI applications with or without code, and then integrate them with the company’s legacy systems such as point-of-sale, said Azita Martin, Nvidia’s vice president of retail AI, CPG and QSR, in an initial media briefing this week.

These three tools include:

  • Retail Loss Prevention AI Workflow: This is what you might call the main tool and a bit of fun. Martin said this workflow was pre-trained with hundreds of images of the most commonly stolen products — we’re talking about detergents, alcohol, over-the-counter drugs, and even steaks — so that the AI ​​could recognize the variety of shapes and sizes of bottles and packaging. New products scanned at the checkout are also part of the training.

“We focused primarily on brands produced by consumer goods companies, but these can be customized to suit the needs of a particular retailer,” added Martin. “They can use synthetic data generation and additional data for further training for all the different types of products that a given retailer sells in their stores.”

  • AI workflow with multi-camera tracking: Multi-target and multi-camera capabilities allow app developers to create systems that track objects with multiple cameras throughout the store. Objects are tracked using visual embedding or appearance, instead of personal biometric information, allowing buyers to maintain privacy.

“As customers move around the store, if products are moved, it tracks those products and even tracks baskets or carts from camera to camera,” explained Martin. “It also gives retailers an overview of the customer journey across the store. So this is another area where we are seeing huge interest from retailers.”

  • Retail store analytics workflow: This tool uses computer vision to provide insight into store analytics, including in-store traffic trends, number of customers with shopping carts, and aisle occupancy via custom dashboards.

“Using a heatmap, you know where your customers go most often and what the most popular customer paths are,” said Martin. “This allows the in-store inventory and merchandising to be optimized to increase revenue.”

Additional details on these new tools will be unveiled at the National Retail Federation’s conference in New York beginning on January 15.

Leave a Reply

Your email address will not be published. Required fields are marked *