Challenge
A smart building operator required occupancy detection across 200+ rooms to optimize HVAC systems.
The solution needed to ensure complete privacy (no camera-based face data storage) while delivering
accurate, real-time occupancy insights.
Solution
- Sensor Fusion: Combined thermal sensors with motion detectors to eliminate the need for cameras.
- Lightweight AI Model: Small LSTM model (8KB, INT8 quantized) for efficient processing.
- Edge Processing: Local inference on Raspberry Pi gateway with occupancy detection in under 100ms.
- Privacy-Focused Cloud: Only aggregated occupancy data per floor/zone is transmitted, with no individual tracking.
- HVAC Integration: Automated temperature adjustments based on real-time occupancy levels.
Outcome
- Energy Efficiency: 18% reduction in energy usage through dynamic HVAC optimization.
- Privacy Assurance: No camera footage or personal data stored.
- Scalable Architecture: Single gateway supports up to 50 sensors efficiently.
- Large-Scale Deployment: Implemented across 200+ buildings covering 10,000+ rooms.
