PyPatrol

• 

  Home Page

• 

• 

• 

• 

• 

• 

• 

• 

Inspiration

- The Problem: Miami-Dade faces a serious invasive species crisis with Burmese pythons, tegu lizards, green iguanas, African land snails, cane toads, and feral hogs disrupting ecosystems and costing millions.
- My Spark: As a solo founder passionate about tech and the environment, I saw an opportunity to use AI and big data to make sense of scattered data – turning chaos into actionable insights that help protect Miami’s unique ecosystem.

What It Does

- Data Aggregation: Scrapes data from wildlife databases, social media, and news outlets for real-time sightings.
- AI Verification: Uses NLP to process text reports and an image recognition model (which you can plug in later) to confirm sightings.
- Geospatial Mapping: Clusters verified sightings to generate interactive maps and heatmaps of invasive species hotspots.
- Actionable Insights: Predicts future spread trends, allowing agencies to focus their mitigation efforts where it matters most.

How I Built It

- End-to-End Pipeline: Developed Python scripts for data scraping from sources like EDDMapS, USGS NAS, and iNaturalist.
- AI Integration: Leveraged spaCy for text analysis and set up a framework for image recognition (currently returning dummy data so I can later drop in my custom .pt model).
- Clustering & Visualization: Implemented DBSCAN clustering for geospatial analysis and built an interactive dashboard using Folium.
- Tech Stack: Entire solution built as a solo project using open-source libraries in Python, with a mobile frontend for visualization.

Challenges I Ran Into

- Data Heterogeneity: Merging data from multiple sources required extensive cleaning and standardization.
- Tech Integration: Combining web scraping, NLP, computer vision, and geospatial clustering in one streamlined pipeline was challenging.
- Performance: Ensuring real-time updates without overwhelming the system demanded careful algorithm optimization.
- Mapping Precision: Accurately overlaying heatmaps on Miami’s actual landmass (especially zooming in on the Homestead–Hollywood area) took some fine-tuning.

Accomplishments I’m Proud Of

- Seamless Data Pipeline: Created a robust, end-to-end solution that aggregates, processes, and visualizes real-time data.
- Interactive Dashboard: Developed an engaging, easy-to-use interface that shows actionable insights at a glance.
- Scalable Architecture: Designed the platform to be extended to other invasive species and regions, proving the concept’s scalability.
- Solo Innovation: As a one-person team, I managed to integrate cutting-edge AI techniques with environmental data – bridging tech and conservation.

What I Learned

- The Power of Data: Clean, well-integrated data is key to turning a complex ecological problem into actionable insights.
- Interdisciplinary Problem Solving: Merging ecology with AI and geospatial analysis opens up new ways to address real-world challenges.
- Resilience as a Solo Founder: Managing all aspects of a project—from coding to design to research—has taught me invaluable lessons in perseverance and resourcefulness.
- User-Centric Design: The importance of a simple, intuitive interface that allows users (from government officials to conservationists) to quickly grasp the data.

What’s Next for PyPatrol

- Refine AI Models: Integrate a fully trained object detection model and fine-tune NLP pipelines with more domain-specific data.
- Expand Data Sources: Incorporate additional public and private datasets to improve accuracy and coverage.
- Real-Time Alerts: Develop a real-time notification system (using web sockets or push notifications) to alert agencies as new hotspots emerge.
- Platform Expansion: Extend PyPatrol to monitor other invasive species and even explore partnerships with local governments and startups to scale the solution statewide.

Built With

• colab
• opencv
• python
• react-native