Corax provides behaviour‑based theft detection for Canadian retail stores, turning existing IP cameras into a real‑time monitoring system that alerts staff before merchandise leaves the store. The solution uses pose‑detection AI to read body language, operates at 30 FPS, integrates with any existing camera network without hardware replacement, and is PIPEDA‑compliant, avoiding facial recognition or shopper profiling.
Funding
Funding not disclosed
Founders
Product
Problem
Retail stores often experience small, unnoticed thefts such as items being slipped into bags or tags being removed, which go undetected because existing security cameras only record footage without providing real‑time alerts. This lack of immediate response forces retailers to absorb losses that are difficult to quantify until inventory checks.
Solution
Corax converts a retailer’s existing IP camera network into an active loss‑prevention system by applying a 30 FPS pose‑detection engine that monitors body language to identify theft‑related behavior. The solution operates without facial recognition, ensuring compliance with Canadian privacy regulations (PIPEDA). When suspicious behavior crosses a predefined threshold, the platform sends discreet, real‑time alerts to the appropriate staff members, enabling intervention before merchandise leaves the store. By leveraging current camera infrastructure, Corax eliminates the need for additional hardware investments while providing actionable intelligence to reduce shrinkage.
Target Audience
Corax serves independent and chain retailers in Canada that operate physical stores and seek to improve loss prevention without compromising customer privacy.
Features
- Integration with existing IP cameras, avoiding costly hardware replacements
- Real‑time pose detection at 30 FPS to analyze body movements and identify theft behavior
- PIPEDA‑compliant operation with no facial recognition or storage of shopper identities
- Automated alerts routed to designated staff members the moment suspicious behavior is detected
- Continuous monitoring that distinguishes normal activity from shoplifting with confidence scores
- Edge‑processing option to keep video streams local and minimize bandwidth usage