MooTrack

AI-powered geofencing to protect livestock from forest predators.

Comment

Inspiration

I was born and raised in the lush green heart of the Western Ghats in the Mookambika Reserved Forest of Karnataka. My childhood was filled with misty mornings, the sound of rustling leaves, and the gentle bell-tinkling of cows grazing near our ancestral home.

But among these serene memories, there was always one shadow that loomed over us leopard attacks. Our cattle, who are considered family in rural households like mine, would often stray into forested areas and never return. Sometimes, leopards would enter our village, snatching away not just cows but also the beloved dogs that guarded our homes.

I’ve seen my grandparents cry over lost animals not because they were assets, but because they were companions. That helplessness, that quiet grief, lit a fire in me. MooTrack is my response - a blend of compassion, technology, and a mission to bring safety back to the people and animals who raised me.

What it does

MooTrack is an AI-powered, cloud-based solution designed to:

  1. Track cattle in real-time using GPS-enabled collars
  2. Detect when cattle wander near forest zones
  3. Predict leopard risk using AI models trained on wildlife sighting data
  4. Alert farmers instantly through mobile or SMS notifications when danger is near
  5. Display a dashboard with cattle paths, forest boundaries, and threat zones

It's not just about location; it's about understanding patterns, preventing losses, and building trust between man, animal, and machine.

How we built it

  1. GPS Tracking & Geofencing Used Streamlit + Python for the frontend interface Simulated cow movement using simulate.py to represent real-time GPS data Geofenced forest areas to trigger alerts

  2. AI Model (ML) Trained an Linear Regression model on a custom risk_dataset.csv with historical environmental and location data Feature-engineered elements like time of day, location proximity, and previous incidents Exported and used risk_predictor_model.pkl for real-time inference

  3. Backend & Alerts Built the backend using Python and MongoDB Atlas for storing cattle data Integrated Twilio API to send alerts when danger conditions are met Cleaned and secured the deployment via .env and Streamlit Secrets Manager

  4. Hosting & Deployment Deployed using Streamlit Cloud Hosted GitHub repository for version control and collaboration MongoDB Atlas handles all our data:

    1. GPS time-series logs
    2. Leopard sighting datasets (from GBIF)
    3. Forest boundary geolocation
    4. Risk prediction results
    5. Geospatial queries check whether cattle have entered danger zones. Alerts are triggered via a simple notification system — future-ready for integration with Twilio

Challenges we ran into

  1. MongoDB authentication issues on Streamlit Cloud despite working locally learned to double-check secrets formatting and URI encoding

  2. Twilio push protection blocked commits until secrets were removed and handled securely

  3. Balancing model accuracy with real-time performance under limited data conditions

  4. Managing file structure and virtual environments in Git while avoiding large .venv folder commits

  5. Resolving merge conflicts and git push errors due to differing branches and commits

Accomplishments that we're proud of

What makes me most proud of Mootrack isn’t just the tech - it’s the purpose behind it. I didn’t just build an app; I created a system that could save someone’s livelihood.

I’m proud that I:

  1. Translated a real-world rural issue into a working tech solution

  2. Designed and trained an AI model from scratch, understanding both the hardware and ML logic

  3. Deployed an end-to-end full-stack system solo, handling everything from frontend UX to cloud deployment

  4. Navigated technical hiccups like Git push protection, secret management, and MongoDB auth errors, and still got it working on Streamlit Cloud

  5. Pitched the project with confidence, knowing it reflects both engineering rigor and empathy-driven design

This is more than a project -it’s a promise that I can engineer solutions that matter.

What we learned

  1. Hands-on experience integrating MongoDB Atlas with Python and Streamlit

  2. Best practices in managing secrets, including escaping credentials and dealing with push-protection errors

  3. Model training and evaluation, and deploying it in a real-time system

  4. Importance of geospatial logic, real-time systems, and reliable alerts in safety-critical applications

  5. Cloud deployment challenges and how to troubleshoot bad auth and URI formatting issues

What's next for MooTrack

  1. Field Deployment: Collaborate with forest departments or local NGOs to test the prototype in a real rural setup.

  2. Hardware Integration: Develop low-cost, durable GPS collars for actual cattle.

  3. Community Reporting: Add functionality for farmers to report wildlife sightings directly.

  4. Edge Computing: Use Raspberry Pi or ESP32 modules to push AI logic closer to rural areas.

  5. Multi-animal Expansion: Extend to goat, sheep, and even pet tracking in rural zones.

I'm deeply passionate about using technology for ecological harmony and this project is a step in that direction.

Built With

Share this project:

Updates