AI Vision Worker Efficiency Tracker

A manufacturing facility had no reliable or objective way to measure worker activity on the production floor. Relying on supervisors to manually observe each worker was both strenuous and impractical, often resulting in infrequent and inconsistent assessments. This made it difficult to accurately track productivity, identify idle time, or detect unsafe zone entries. The facility needed a non-intrusive, automated solution that could eliminate the dependency on manual observation entirely.

The solution was built around IP cameras with wide-angle lenses installed across the production floor, providing full video coverage of worker activity across all zones. A Raspberry Pi 5 served as the edge AI processing unit, running an OpenCV-based inference pipeline to analyze video frames in real time without sending footage to an external server.

An activity classification model processed each frame to detect and categorize worker states (such as active, idle, or stationary) using pose estimation techniques. Polygon-based zone definitions were applied across the floor layout to monitor occupancy in safety and restricted areas, triggering alerts whenever a violation was detected.

All activity data was displayed through a web-based dashboard, providing shift-level productivity heatmaps, activity timelines, and zone compliance logs for supervisors to review. An integrated alert engine sent real-time email and browser notifications whenever a safety zone was entered or extended idle time was detected, ensuring immediate awareness without the need for constant manual observation.