🌍 EnviroSense AI

A Hybrid Deep Learning Ecosystem for Proactive Air Quality Forecasting & Anomaly Detection

EnviroSense AI is an end-to-end environmental intelligence platform. By bridging the gap between raw IoT sensor data and advanced Deep Learning, the system transforms air quality monitoring from a reactive "status check" into a proactive "early warning system."

---

🏗️ System Architecture

The platform follows a modular 4-layer architecture designed for high availability and analytical depth:

• Edge Layer: ESP32 nodes capturing high-fidelity particulate matter (PM1.0, PM2.5, PM10) and particle bin counts (0.3μm–10μm) via UART/Modbus protocols.
• Persistence Layer: TimescaleDB (PostgreSQL) optimized for high-velocity time-series storage and hypertable partitioning.
• Inference Layer: FastAPI serving pre-trained Deep Learning models (LSTM/Autoencoders) for real-time forecasting and anomaly detection.
• Presentation Layer: React PWA utilizing WebSockets for instant data updates and cross-platform accessibility.

---

📊 Data Analytics & AI/ML Methodology

This project implements rigorous data science workflows to ensure accuracy and reliability:

Data Analytics Focus

• Time-Series Decomposition: Statistical separation of raw data into Trend, Seasonality, and Residuals to validate environmental cycles.
• Size-Distribution Profiling: Analyzing shifts in particle bin counts (0.3μm to 10μm) to identify pollutant sources (e.g., smoke vs. dust).
• Environmental Correlation: Quantifying the impact of Temperature and Humidity on Particulate Matter concentrations through multivariate regression.

AI/ML Focus

• Sequence Modeling (LSTM): Implementing Long Short-Term Memory neural networks to capture temporal dependencies and predict PM levels for a 6-hour horizon.
• Unsupervised Anomaly Detection: Using Reconstruction Error thresholds from trained Autoencoders to flag sensor hardware failures or localized extreme events.

---

🛠️ Tech Stack

• Hardware: ESP32, Next-PM Optical Sensor, DHT22.
• Backend: Python 3.10+, FastAPI, SQLAlchemy, Paho-MQTT.
• Frontend: React.js, Vite, Tailwind CSS, Recharts.
• Database: PostgreSQL with TimescaleDB extension.
• ML/DS Libraries: TensorFlow/Keras, Scikit-learn, Pandas, NumPy, Plotly.

---

🖥️ Remote Development & VPS Workflow

To maintain a production-ready environment, development is conducted via Remote-SSH, ensuring that high-compute tasks remain on the server while code is authored in a local IDE.

1. Remote Connection

Standardize connections via a non-root user to maintain security and system integrity.

SSH Configuration (~/.ssh/config):

Host envirosense-vps
    HostName <YOUR_VPS_IP>
    User <YOUR_USERNAME>
    IdentityFile ~/.ssh/id_ed25519

2. Isolated Execution

All analytics and subscriber logic should run within a dedicated Python Virtual Environment to prevent dependency drift.

# Activation and Verification
source .venv/bin/activate
python -c "import sys; print(sys.prefix)" # Confirms isolated environment

3. Port Forwarding for Visualization

When running analytical dashboards (e.g., Plotly Dash or Streamlit) on the VPS, use VS Code Port Forwarding to tunnel the web interface to your local browser:

• Remote Port: 8050 (or your app's port)
• Local Address: localhost:8050

---

📦 Getting Started

1. Clone the Repository:

git clone https://github.com/YourUsername/EnviroSense-AI.git
cd EnviroSense-AI

2. Environment Setup: Install dependencies from requirements.txt into your virtual environment.
3. Database Configuration: Ensure the PostgreSQL pg_hba.conf allows connections for your specific database user.
4. Service Deployment: Configure the subscriber.py as a systemd service for 24/7 data persistence.