Store-wide Loss Prevention

A suite of Intel® edge AI applications for retail loss prevention, powered by Scenescape spatial computing and Intel® OpenVINO™ inference.

Use Cases

#	Use Case	Status	Directory
1	Person of Interest (POI)	Available	person-of-interest/
2	Suspicious Activity Detection	Available	suspicious-activity-detection/

---

Person of Interest

Real-time detection of enrolled suspects across multiple cameras using face re-identification and FAISS vector search, with historical investigation capabilities.

For full details see the POI User Guide.

---

Suspicious Activity Detection

MQTT-driven loss prevention service for Scenescape retail deployments. The service monitors person behavior across store zones using real-time tracking, manages session state and detection rules, and stores cropped person frames in SeaweedFS. Behavioral analysis (pose detection + VLM confirmation) runs as a separate service invoked conditionally when a person enters a high-value zone.

Detected Activities:

#	Activity	Trigger	Alert Level
1	Merchandise Concealment	Behavioral Analysis confirms suspicious pose + VLM	WARNING
2	Checkout Bypass	Visited HIGH_VALUE zone, exited without CHECKOUT	WARNING / CRITICAL
3	Loitering	Dwell > threshold in HIGH_VALUE zone	WARNING
4	Repeated Visits	Re-entries ≥ threshold to same HIGH_VALUE zone	WARNING
5	Restricted Zone Violation	Entered RESTRICTED zone	CRITICAL

For full details see the SAD User Guide.