Success story.
AI-powered Food Defense: Security cameras detect risks before they escalate
Two monitors, 32 cameras, and three shifts. A food plant was supposed to strengthen its food safety measures (FSSC 22000, SQF, FSMA), but it was reviewing the recordings after the fact.
We take on any challenge
At Zenta there are no obstacles we cannot overcome.
Challenge
To detect intentional product tampering early: people outside their designated area, blocked cameras, foreign objects. And to do so using steam, reflections, and image obstructions, without overloading an already overloaded team with false alarms.
Solution
We designed a platform that uses AI only where it provides the most benefit (uniformity by zone and unusual objects) and leverages the native analytics of Axis cameras for everything else. We added a vision-language model, filters against false positives, a single alert inbox with video footage of the incident, and an auditable log.
Results from our management
Calibration with real video of the plant.
False positives on normal operation: from 20% to 0%.
Intentional contamination and camera obstruction detected in 19 seconds, at no AI cost.
Cost per camera and latency measured; demo deployed and ready to scale.