AMY ZINGER

CREATIVE DIRECTOR

Bee Communication

Bee Communication
Layered Paper

Overview

This installation visualizes a single honeybee’s waggle dance—a sophisticated communication method that conveys the distance, direction, and quality of a food source to follower bees. Using data captured from a fixed-position camera, the piece translates these complex motion patterns into physical form through precision laser cutting. The two translucent layers represent distinct but interdependent modes of bee movement:

    • Rear Layer: Precise thorax positions during a single waggle dance
    • Front Layer: The same bee’s surrounding exploratory behavior—before and after communication

Nothing in this installation is symbolic or interpretive. It is the raw geography of a message that only another bee could have understood. The translucent materials allow viewers to observe both the focused communication pattern and the broader movement context simultaneously, creating a layered experience that mirrors the complexity of natural bee behavior.

Data Analysis

The data foundation for this project comes from the Berlin 2019 honeybee study, a sophisticated multi-camera research project combining computer vision and behavioral classification. After extensive review of the dataset, a single exemplary dance sequence was selected to represent the fascinating patterns of honeybee communication.

Selected Dance Record

    • Dance ID: 14589364613185451070
    • Camera: 1 (right-side hive)
    • Dancer ID: 34
    • Date: 2019-08-21

This specific dance was chosen through systematic visual inspection using a custom JavaScript-based alignment tool that enabled dynamic review of movement and waggle phase overlays across the hive. The selection criteria prioritized dances with clear, well-defined patterns and comprehensive tracking data.

Data Sources

The raw data was drawn from several interrelated datasets:

    • Berlin2019_dances.csv – Metadata about all detected dances
    • Berlin2019_waggle_phases.csv – Each classified waggle phase, including position and angle
    • 2019-08-21.csv – Per-frame positional tracks of all bees recorded on August 21, 2019
    • Supplemental files – Data for followers, classifier labels, and manual annotations

Data Processing Methodology

Due to the fragmented nature of the source datasets, preparing this installation required significant cross-referencing and filtered processing:

    1. Initial Filtering
        • The full 2019-08-21.csv file was filtered for cam_id == 1
        • Results were joined with Berlin2019_dances.csv to isolate all dances from Camera 1
    2. Activity Analysis
        • To identify compelling sequences, the following metrics were compared:
          • Number of waggle phases (joined from waggle_phases)
          • Number of track points (joined from frame-level tracking)
        • This comparison generated dances_most_active_0821_cam1.csv with derived columns for waggle_count and track_count
    3. Candidate Selection
        • Dances were systematically reviewed using a custom visualization tool
        • Selection prioritized visual coherence and narrative clarity, not just statistical metrics
    4. Detailed Extraction
        • Once Dance ID 14589364613185451070 was selected, two filtered datasets were produced:
          • waggle_phases_most_active_0821_cam1_id.csv – 54 rows capturing the bee’s thorax location during each waggle phase
          • tracks_most_active_0821_cam1_id.csv – 3,003 track points reflecting the bee’s movement before and after the waggle sequence
    5. Visual Refinement
        • Using the custom JavaScript tool, the data was:
          • Verified for alignment between waggle phases and track points
          • Adjusted for scale, rotation, and translation to achieve visual balance
          • Exported as clean SVG paths for each layer, ready for laser cutting

Data Representation System

Layer Source Dataset Data
Points		 Physical Representation
Rear Layer waggle_phases_most_active_0821_cam1_id.csv 54 Thorax locations during waggle phases
Front Layer tracks_most_active_0821_cam1_id.csv 3,003 Continuous motion of the same bee
Metadata dances_most_active_0821_cam1.csv 1 Used for validation and context

All positions are represented in millimeters. The coordinate system from the source datasets already incorporated camera homography correction, requiring only relative alignment adjustments (scale, rotation, offset) for final presentation.

Construction Specs and Materials

  • Dimensions: 24″ × 24″ × 8″
  • Viewing: Front-facing with illumination from behind

Materials:

    • Translucent Yupo paper (two data layers plus an additional diffusion layer)
    • ½” pine plywood frame
    • Adjustable LED keylight panel (rear-mounted)

Fabrication Process:

    1. Data export to precision SVG files
    2. SVG processing in Adobe Illustrator:
        • Created outlines and merged paths for laser cutter compatibility
        • Added small paper anchors for structural support (after initial prototype revealed stability issues)
        • Exported processed file to SVG format compatible with Lightburn software
    3. Laser cutting of Yupo paper layers
    4. Manual assembly with diffusion layer
    5. Integration with lighting system
    6. Frame construction and mounting

Lighting System:

    • Adjustable LED panel positioned behind the layers
    • Diffusion layer to create even illumination
    • Brightness calibrated to optimize visibility of both data layers simultaneously

Code Architecture

The data visualization and extraction process relied on a custom-built tool developed specifically for this project:

Development Framework:

    • JavaScript + HTML5 Canvas
    • PapaParse for CSV loading and processing
    • Custom SVG export logic with millimeter precision

Key Features:

    • Dynamically load and parse raw bee dance data
    • Filter by camera position and dance ID
    • Display and overlay waggle/track data with adjustable visibility
    • Perform alignment adjustments (scale, rotation, X/Y offset)
    • Export each layer to SVG for laser cutting fabrication

The tool enabled iterative refinement of the visual presentation while maintaining absolute fidelity to the original data points. This approach ensured that the final installation accurately represents the actual movement patterns recorded in the original research.

References

Dormagen, D. M., Wild, B., Wario, F., & Landgraf, T. (2023). Data used in Machine learning reveals the waggle drift’s role in the honey bee dance communication system (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7928121

von Frisch, K. (1967). The Dance Language and Orientation of Bees. Harvard University Press.

 

Return to Home >