Skip to main content

👭 Bee interaction detection

🎯 Purpose​

Detects and analyzes social behaviors between bees at the hive entrance, including feeding interactions and communication behaviors.

🎭 User Story​

  • As a beekeeper interested in colony social dynamics
  • I want to automatically detect when bees are interacting with each other at the entrance
  • So that I can understand social behaviors, feeding patterns, and potential stress indicators in my colony

🚀 Key Benefits​

  • Behavioral insights: Understanding of colony social dynamics and health indicators
  • Feeding pattern analysis: Detection of trophallaxis (food sharing) between bees
  • Stress detection: Abnormal interaction patterns may indicate colony stress or disease
  • Research value: Valuable data for understanding bee social behavior

🔧 Technical Overview​

Uses computer vision to detect when multiple bees are in close proximity and exhibit specific behavioral patterns such as antenna contact, feeding positions, or grooming behaviors. Analyzes bee pose and proximity relationships over time.

📋 Acceptance Criteria​

  • Detects bees within interaction distance (<2 bee body lengths)
  • Identifies feeding/trophallaxis behaviors with >70% accuracy
  • Tracks interaction duration and frequency
  • Distinguishes between different interaction types (feeding, grooming, fighting)
  • Records interaction data with timestamps and bee IDs

🚫 Out of Scope​

  • Individual bee identification across sessions
  • Detailed pose estimation (handled by separate bee pose features)
  • Audio-based interaction detection
  • Inside-hive social behavior analysis

🏗️ Implementation Approach​

  • Detection: Multi-object tracking to identify bee pairs/groups
  • Proximity analysis: Distance calculation between bee centroids
  • Behavior classification: Machine learning model trained on interaction patterns
  • Data output: Structured interaction events sent via telemetry API

📊 Success Metrics​

  • Interaction detection accuracy >70% compared to manual annotation
  • False positive rate <20%
  • Processing capability for 5+ simultaneous bee interactions
  • Real-time analysis at 15+ FPS
  • Behavioral pattern recognition for 3+ interaction types

🔗 Related Features​

📚 Resources & References​

  • LabelBee platform interaction analysis features
  • Research on bee trophallaxis and social behaviors
  • Computer vision approaches for animal behavior analysis

💬 Notes​

This feature provides valuable research data and may serve as an early indicator of colony health issues through social behavior changes.