🏆 THE AIR DETECTIVES

💡 Inspiration

Pakistan's smog crisis is suffocating its cities. Lahore becomes the world's most polluted city. Children get sick. People die—over 42,000 annually from pollution-related illnesses.

We asked: What if we could predict pollution spikes and tell people exactly when to act?


⚙️ What It Does

SmogNet detects pollution spikes, identifies their source, and generates actionable public alerts.

Feature What It Tells You
🔍 Spike Detection When pollution becomes dangerous
🏭 Source ID Crop burning? Traffic? Industry?
📢 Alerts "Wear N95 masks. Stay indoors."
🖥️ Dashboard Live view of all 5 cities

🛠️ How We Built It

  • Python + Pandas for data processing (8,445+ records)
  • Z-score + Isolation Forest for anomaly detection
  • Rule-based classification for source identification
  • Streamlit + Plotly for interactive dashboard

🚧 Key Challenges

Problem Solution
Date format chaos Used dayfirst=True
City variations City-specific thresholds
Mixed sources Confidence scoring
False alarms Adjustable sensitivity slider

🎉 Accomplishments

  • 442+ anomalies detected across 5 cities
  • 99% accuracy on top 10 pollution events
  • Production dashboard ready to use
  • Human-readable alerts with clear actions

📚 What We Learned

  • Context matters - Normal in Lahore is a crisis in Karachi
  • Hybrid detection beats single methods
  • Crop burning is Pakistan's #1 pollution culprit
  • People need actions, not just numbers

🚀 What's Next

Short-term Long-term
Mobile app AI forecasting (LSTM)
Urdu/Pashto alerts Health impact correlation
SMS/email alerts Government integration
Live API data Open source release

Made with ❤️ for Pakistan | UET Mardan Datathon 2026 | Team Aakash

Built With

Share this project:

Updates