Computer vision system for a cashier-less store

Our partner approached us to develop an AI-driven customer identification system to improve their store’s security and management, and give insights into customer behavior. The system would need to determine if a customer is registered with the business and if they have a payment method set up, as well as track customer movements and time spent in-store.  We built them a robust video identification system with four key components:

1. A human detection module to identify individuals in the store.
2. A re-identification model to recognize returning customers.
3. An enhanced machine learning tracking algorithm to document movement and time spent within the store.
4. A Flask-based web app to document videos, process them, and display an analysis with a machine learning algorithm.

Additionally, we ensured our partner could adapt the system to track and analyze other variables and designed it to be scalable to easily include other objects.

Project outcomes

1. A highly adaptable system with an intelligent tracking algorithm that can scale to detect new variables and objects.
2. Better in-store security and easier surveillance with real-time customer identification.
3. Improved employee process management, ultimately helping our partner work more efficiently and reorganize their store’s layout.