STEMMA Recognition for product recognition

A computer vision module for automatic product recognition at self-service checkout counters and weigh-in stations.

The system uses neural network algorithms to accurately identify products based on their visual characteristics: shape, color, texture, packaging, and distinctive features. The module operates without requiring barcode scanning, which is particularly valuable for unlabeled items—such as fruits, vegetables, baked goods, prepared foods, and fresh produce. The model is trained on a specific retail chain’s product assortment and supports thousands of SKU, ensuring high recognition speed and accuracy even under challenging lighting conditions and high foot traffic.

Problems that product recognition software solves

Problems:

• Customer and staff errors when selecting items from the checkout menu (intentional or accidental).
• Fraud - replacing expensive goods with cheaper ones when weighing or scanning.
• Slowdown of service due to manual search of the item in the list.
• Difficulty working with unlabeled items.
• Loss of sales data due to incorrect product identification.

Business Values

Product recognition software provides:

• Loss reduction - automatic verification of product and selected item matches.
• Faster service - the customer does not have to manually search for a product in the checkout menu.
• Increased accounting accuracy - the system records each sale correctly.
• Increased customer loyalty - reduced waiting time at checkouts and weighing areas.
• Anti-fraud protection - preventing deliberate product substitution.
• Transparent analytics - accurate data on real turnover.