IoT and the Future of the Retail Industry | Pensa's Richard Schwartz

Engineering & simulationdeep learning / neural networks

The team trained deep learning models to recognize products on retail shelves down to the individual SKU level, enabling the system to automatically infer shelf organization and detect stockouts or misplaced items.

“what we're doing is we've trained the computer to essentially stare at the shelves. And visually, visually, recognize the products that are on the shelf, down to the individual skew or product item level, and also look repeatedly and automatically infer or conclude how the shelf is organized…”

Engineering & simulationdeep learning / neural networks

The AI was used to learn product recognition in a way that generalizes across different retail locations and chains, allowing the system to scale by incrementally learning new products it encounters.

“What we're doing is we're using deep learning or neural networks to essentially learn the way people learn to recognize products. And by doing that, anything that we've trained the system to recognize even once, it should recognize everywhere, regardless of which location, and which retail chain.”