A new AI system developed by Deep Vision Data uses machine learning to monitor retail store stock and merchandising. When used with 100 percent synthetic data, the system instantaneously audits for display planogram compliance, determines the quantity and SKU of out-of-stock items, and even alerts when products are improperly merchandised.
Cincinnati, Ohio-based Kinetic Vision’s Deep Vision Data division recently developed a deep learning AI system that classifies product types, available stock, missing stock and planogram compliance for a retail store merchandising audit system.
“Although created to demonstrate the efficacy of synthetic training data, we feel the model architecture would easily scale to entire stores,” said Kinetic Vision President and CEO Rick Schweet. We’ve shown that synthetic data can be as effective as physical information, or even more so, in training deep learning models. The result is much faster AI system development, often with very significant cost savings.”
The model, based on the ResNet101 architecture, was trained with 20,000 synthetic product images using a 50-50 split of structured and unstructured domain randomized subsets and an 80-20 training/validation data split. Model validation was also done with synthetic training data, and the resulting system correctly classified product brands and out-of-stock items on actual photos with nearly 100 percent accuracy.
Synthetic training data can be used for almost any machine learning application, either to augment a physical dataset or completely replace it. By effectively using domain randomization, synthetic data can be made to be indistinguishable from physical training information.
Equally important, because the user controls the dataset distribution, features or events that rarely occur in physical datasets can be made to be more prevalent, thus enabling model training for those rare occurrences. Synthetic training data is inherently less costly, faster to create, perfectly annotated, and isn’t constrained by availability, time or even the physics of the natural world.
Deep Vision Data synthetic training data are created using a variety of proprietary methods, can be multi-class, and developed for both regression and classification problems. Labeling and semantic segmentation are automatic and 100 percent accurate.