BeeGuardian

BeeGuard - AI-Powered Bee Health Monitoring System

Live Demo: https://elcago.github.io/BeeGuardian/

Overview

BeeGuardian is an integrated beekeeping management platform that combines AI-powered health analysis with localized resource matching, weather-based recommendations, and comprehensive education - making advanced beekeeping technology accessible to beekeepers of all experience levels.

The system analyzes visual and audio signals using a multimodal neural network to detect hive health issues with 92.6% accuracy on 1,639+ samples, then provides actionable guidance including local suppliers, treatment timing based on weather conditions, and educational resources.

Project Structure

BeeGuardian/
├── index.html              # Frontend web application
├── model_server.py         # Backend Flask API server
├── requirements.txt        # Python dependencies
├── README.md              # This file
└── bee_health_final_model.h5  # Trained ML model

Features

Fully Functional (Web Demo)

• U.S. colony loss data visualization (2006-2024)
• Smart local resource finder with problem-specific search
• Weather-based beekeeping recommendations
• Comprehensive education hub with 9 learning modules

Backend Required (ML Analysis)

• AI-powered hive health analysis (4 health categories)
• Image, audio, and video file analysis
• Bee counting and behavior detection

Setup Instructions

Frontend Only (Quick Start)

1. Visit the live site: https://elcago.github.io/BeeGuardian/
2. All features work except ML analysis (requires backend)

Full System with Backend

Prerequisites:

• Python 3.8 or higher
• Trained model file: bee_health_final_model.h5 (contact for access)

Step 1: Clone the repository git clone https://github.com/elcago/BeeGuardian.git cd BeeGuardian

Step 2: Install Python dependencies pip install -r requirements.txt

Step 3: Add the model file Place bee_health_final_model.h5 in the project root directory.

Step 4: Run the backend server python model_server.py

You should see: Loading models… Model loaded successfully! Starting BeeGuardian server on http://localhost:5001

Step 5: Serve the frontend (in a new terminal) python3 -m http.server 8000

Step 6: Open in browser Navigate to http://localhost:8000

API Endpoints

The backend provides several REST API endpoints:

• GET / - Server status page
• GET /api/health - Health check
• POST /api/analyze - Upload and analyze files
• POST /api/stream - Analyze live stream frames
• GET /api/model/info - Model information

Model Details

Architecture: Cross-Attention Multimodal Neural Network (CAMNN)

• Visual pathway: CNN feature extraction (224x224 images to 4096 features)
• Audio pathway: MFCC + Mel Spectrogram + Chromagram + STFT to 4096 features
• Fusion: Cross-attention mechanism for multimodal integration

Performance:

• Training samples: 1,639+ images and audio recordings
• Accuracy: 93% on test set
• Classes: Healthy, Varroa Mites, Missing Queen, Hive Robbing

Input formats:

• Images: PNG, JPG, JPEG, GIF
• Audio: WAV, MP3
• Video: MP4, AVI

Technologies Used

Frontend

• HTML5, CSS3, JavaScript
• Chart.js for data visualization
• Google Maps Places API
• OpenWeatherMap API
• Responsive design with mobile support

Backend

• Python 3.8+
• Flask web framework
• TensorFlow/Keras for ML inference
• OpenCV for image processing
• Librosa for audio feature extraction

Data Sources

• Colony loss data: Bee Informed Partnership & USDA NASS (2006-2024)
• Educational content: University extension services, peer-reviewed research
• Weather data: OpenWeatherMap API
• Local resources: Google Maps Places API

Project Background

This application addresses the critical issue of declining bee populations (40% annual colony losses) by providing beekeepers with AI-powered diagnostic tools, practical local resources, and evidence-based educational content.

For Users

Quick Demo

1. Visit the live site to see working features: https://elcago.github.io/BeeGuardian/
2. Watch the demonstration video showing ML analysis in action
3. View source code directly on GitHub

Running Backend Locally

While the backend can be run locally following the setup instructions above, the frontend demo showcases the complete user interface and integration design. The ML model file is not included in the repository due to size constraints (150MB+).

Technical Highlights

• Intelligent problem-specific resource search with relevance categorization
• Asynchronous API coordination (geocoding to search to details)
• Real-time weather integration with beekeeping recommendations
• Multimodal ML model with cross-attention fusion
• Comprehensive error handling and graceful degradation

License

MIT License - See LICENSE file for details

Acknowledgments

• Bee Informed Partnership for colony loss data
• University extension services for educational resources
• Research institutions studying multimodal neural networks for agricultural applications

Contact

For questions about the model or demo access, please open an issue on GitHub.

This site is open source. Improve this page.