12 / Manufacturing
Privacy-Preserving Camera Analytics & Workforce Tracking for Factories
We made 100+ old factory cameras smart without sending a single video frame to the cloud, and tracked who is working where using QR tags instead of face recognition.
- Client
- One of India's Largest Toy Manufacturers (Undisclosed Under NDA)
- Stack & approach
- Computer VisionEdge AIZero Cloud VideoPrivacy-PreservingOn-Prem
The problem
Legacy analog cameras streamed signals incompatible with modern CV, and a full replacement was not viable. Streaming 100+ high-res feeds to the cloud would have meant severe lag and bandwidth cost, and facial recognition for worker tracking was rejected outright on privacy grounds.
Solution & architecture
An edge-first spatial-intelligence layer with zero cloud video. Hardware encoders digitised the analog estate into RTSP streams without replacing a camera. Edge nodes process every feed on-site, and QR codes on machines act as fiducial markers: when a worker occludes a marker, the desk is confirmed in use. Workers stay anonymous and no raw video ever leaves the facility.
Why it is built this way
Zero cloud video
Every frame processed on-site
Privacy by design
QR fiducials, never facial recognition
No camera rip-&-replace
Encoders reuse the analog estate
