Eye-Shelf is the solution that provides valuable sell-out information by leveraging the power of AI and edge computing.
Thanks to high-performance cameras and computational boards, it can recognize products taken from shelves or refrigerated displays and associate them with general characteristics of the person who made the selection.
It uses advanced Deep Learning (AI) models that are properly trained to recognize specific products and classify the gender and age range of the person who made the selection.
Automating restocking when a "stock-out occurs"
Collecting sell-out data by cross-referencing products with gender
Determining the age of the person who took the item.
Eye-Shelf incorporates computer vision techniques that leverage the power of neural networks to recognize all objects and faces captured by the cameras. Another innovative aspect is the hardware used: it is well-known that deep learning models are resource-intensive.
This led us to select highly advanced components capable of optimally supporting neural networks with the goal of developing a solution based on edge computing—meaning the ability of a system to perform analysis directly on the display, resulting in a decentralized and highly flexible solution.
Benefits
Eye-Shelf can be integrated into existing structures through dedicated kits, and the speed of the product recognition models enables real-time analysis with an accuracy level of 90%. Furthermore, all generated reports are available via a web interface, always accessible and easy to consult.
Our primary target includes goods manufacturers who can drive the adoption of the solution, manufacturers of display stands or shelves, and supermarkets where Eye-Shelf can be integrated into pre-existing environments.
The development of the solution required the contribution of various highly specialized professionals: the Data Engineer handled the data cleaning phase, the Data Scientist worked on data augmentation and deep learning model training, while the AI Engineers focused on developing the software and integrating it with the AI models and highly specialized hardware.
