structwatch

A robust structural health monitoring system that fuses high-speed IMU vibration modal analysis with DeepLabV3+ vision-based crack detection over a LoRa mesh network.

Features

• Firmware Pipeline (C99): Radix-2 Cooley-Tukey FFT, Welch PSD analysis, and modal frequency tracking in a 512-word zero-malloc footprint.
• Visual AI: PyTorch DeepLabV3+ semantic segmentation for crack and damage pixel classification.
• Risk Fusion: Kalman filter-based fusion of modal vibration shifts and visual structural damage.
• LoRa Telemetry: Encoded 32-byte wire format optimized for long-range, low-bandwidth data gathering.
• Real-time Dashboard: FastAPI backend serving a vanilla JS WebSocket real-time visualizer.

Quickstart

Firmware Simulation

make compile-firmware
make test-firmware
make simulate

Python Package

python -m venv .venv
source .venv/bin/activate
make install
make test-python

# Run dashboard in simulation mode
structwatch run --simulate

Navigate to http://localhost:8000.

Architecture

See ANTIGRAVITY_PROMPT_structwatch.md for full implementation details.