Smart E-Nose AI — Atmospheric Safety
Five-sensor gas array with a Random Forest classifier and a live glassmorphism dashboard for real-time hazard detection.
The Smart E-Nose AI fuses a five-sensor MQ gas array with a Random Forest machine learning model to classify air quality in real time. Instead of threshold alarms, it reads the combined “signature” of each gas across the sensors and distinguishes clean air, smoke and fire, LPG and methane leaks, alcohol vapour, and polluted or stale air. The hardware layer is an Arduino Uno carrying MQ-2, MQ-3, MQ-5, MQ-7 and MQ-135 sensors; the backend is a Python Flask REST API with SQLite persistence, and a multi-threaded bridge syncs hardware readings to the web within milliseconds. The dashboard uses a glassmorphism interface with particle simulations that turn red during hazard detection and Chart.js time-series analytics. A built-in software sensor simulator lets the whole system run with no hardware attached.
Key Features
- Five-sensor MQ array with multi-sensor fusion across MQ-2, MQ-3, MQ-5, MQ-7 and MQ-135
- Random Forest classifier separating clean air, smoke, gas leaks, alcohol and pollution
- Flask REST API with SQLite persistence and a multi-threaded hardware bridge
- Glassmorphism dashboard with reactive particle background and live time-series charts
- Built-in sensor simulator so the dashboard runs with no hardware attached
- One-command launcher that starts the backend, frontend and simulator together
Repository Topics
Challenges
Sensor noise and overlapping gas signatures make reliable classification far harder than a simple threshold alarm — the model had to learn patterns rather than cut-offs.
Results
A working IoT safety prototype that turns invisible atmospheric data into actionable, real-time insight for schools, offices and industrial sites.