hello@itrexgroup.com +1 559 550 4575

←←All cases

Edge AI security platform for theft detection in small/mid-size retail stores

Client

A UK-based security solutions provider

Industry

Retail security and surveillance

Services

AI solution architecture, machine learning development

Tech stack

AWS Panorama, Lenovo ThinkEdge SE70, Amazon SageMaker, Python, React, Java Spring Boot, AWS IoT Core, AWS S3 bucket

Challenge

Our client is a UK-based security company on a mission to help small retail shops combat the ongoing issue of petty theft. For many small retailers, theft is a daily concern—and yet one that is often ignored by law enforcement due to the low monetary value of stolen goods. Unlike large chain stores, small retailers cannot afford high-end surveillance systems or pay for ongoing cloud-based analytics, leaving them vulnerable and underserved. The client envisioned a theft detection solution powered by artificial intelligence (AI) that would be both effective and affordable for small retail environments. Their goal was to give independent retailers the same level of security intelligence that enterprise retailers enjoy—without the heavy price tag. To realize this vision, the client partnered with ITRex, a reliable AI development company, to create a scalable, cost-effective, AI-driven system that could be installed with minimal technical effort and no ongoing cloud expenses.

Our team’s contribution

The client asked our team to design and prototype a low-cost solution that retailers can purchase and install. Our responsibility was to:

●

Minimize recurring cloud usage fees by performing video analysis and data storage locally on edge devices

●

Train and implement machine learning models to detect common theft patterns

●

Enable a human-in-the-loop approach that keeps store employees in control of how incidents are handled

●

Establish a roadmap of theft scenarios and work with the client to prioritize detection features for initial deployment

Solution

We delivered an affordable edge AI platform for theft detection. It integrates seamlessly with existing CCTV infrastructure and supports real-time decision-making for store staff. Here’s how we made it happen: Edge-based architecture for cost efficiency To eliminate the need for cloud processing and minimize cost, our team chose to deploy the solution on the Lenovo ThinkEdge SE70, a rugged edge device optimized for AWS Panorama. This device is easy to install in-store, and it supports up to ten CCTV cameras, making it suitable for most small retail layouts. Theft detection and real-time alerts We developed machine learning theft detection models using Amazon SageMaker and trained them on the client’s proprietary dataset of shoplifting patterns. The initial system supports two key theft scenarios:

●

When a person conceals an item in their clothing

●

When a person grabs an item and sprints toward the exit

Once suspicious behavior is detected, the system sends a WhatsApp message to store personnel. To reduce friction and cost, we deliberately avoided building a dedicated mobile app. The alert includes a link to the recorded footage highlighting the potential theft. Evidence management and offender recognition The system also provides advanced features to aid in identifying and processing incidents:

●

Facial recognition to spot repeat offenders

●

Live alerts when someone in the known offenders database enters the store

●

Evidence packaging process that captures and saves critical footage, offender snapshots, and metadata in formats admissible in court

●

Information-sharing capability that enables stores that use our platform to share offender profiles and photos with one another securely

Built-in protocol support We integrated the platform with the ASCONE protocol, a UK-recognized step-by-step decision-making framework that store staff can use to minimize wrongful accusations and take action safely and appropriately. What’s next We’ve successfully rolled out a prototype in two small stores in the UK. Looking ahead, we’re working closely with the client on:

●

Developing an API layer to integrate with external compliance and security tools

●

Retraining ML algorithms to improve detection accuracy based on user feedback

●

Expanding theft detection capabilities to include mobility devices like strollers and wheelchairs, and count the number of items each visitor took from shelves and presented at the counter

●

Designing a mobile app for a richer real-time employee experience

●

Scaling to enterprise retailers with additional features and analytics

●

Adding AI-driven theft pattern analytics to enable smarter store security policies

Impact

Our edge AI platform delivers powerful theft detection and prevention capabilities at a fraction of the cost of traditional solutions. Here is the impact our solution made:

●

50% cost savings in the first year compared to cloud-based alternatives

●

Up to 8x cheaper in subsequent years, depending on the number of cameras and data input volume

●

No hardware upgrades needed. The system works with standard, existing CCTV infrastructure.

●

Reducing police dependence. Store owners now have sufficient evidence to go directly to court, bypassing under-resourced law enforcement for petty theft.

●

Protecting small businesses. Empowers independent retailers to proactively manage theft and enhance safety.