LabOS is an enterprise-grade laboratory automation and safety platform. This repository contains the LabVision-Hailo module, a production-grade edge AI occupancy system optimized for high-performance hardware accelerators.
- Hardware: Raspberry Pi 5 + Hailo AI HAT+ (26 TOPS).
- Real-time Occupancy: Multi-threaded async inference using HailoRT.
- Sensor Fusion: Combines AI Vision, mmWave presence, and PIR sensors for 100% occupancy accuracy.
- Industrial Logic: 5s occupancy entry buffer / 20s empty exit buffer to ensure stable automation.
- Zero-Lag Pipeline: GStreamer-based RTSP handling with hardware H.264 decoding.
- Training Suite: Scripts for fine-tuning YOLOv8 on custom lab datasets.
- Suppression Engine: Specialized training to ignore furniture and lab equipment.
- Standard Monitoring: CUDA-accelerated inference for NVIDIA GPUs (RTX 3050).
- Industrial Design: Dark-mode interface with glassmorphism and cyan accents.
- Pure Vanilla Stack: Dependency-free HTML/CSS/JS optimized for edge server hosting.
- Inference Engine: Hailo-8 AI Accelerator (26 TOPS)
- Host Platform: Raspberry Pi 5 (16GB RAM)
- Communication: Mosquitto MQTT (Industrial Messaging)
- Development: Python 3.11, C++ (HailoRT), GStreamer
LabVision-Hailo/: Production deployment scripts for Raspberry Pi.src/hailo_app.py: Main async inference & logic application.src/fusion_engine.py: AI + PIR + mmWave sensor fusion logic.
LabVision-AI/: Training and general-purpose monitoring scripts.yolov8s_final.pt: Fine-tuned production weights.index.html: Dashboard entry point.
The system publishes to the following topics:
lab/lab101/zoneA/occupancy/state: Fused occupancy state (JSON).lab/lab101/system/heartbeat: Health and diagnostic data.
LabOS - Intelligent Laboratory Automation & Safety Platform (26 TOPS Edition)