Skip to content
Freie Universität Berlin

BioRobotics Lab

Research Assistant · Tim Landgraf · Oct 2025 – Present · Berlin

Working on the "Honeybees under pesticides" experiment, researching 8 weeks of 24-hour recordings across 4 beehives.

I retrained a YOLO vision model in PyTorch to adapt to varying light conditions at the feeding camera. I also optimized the computer vision tracking system for untagged bees, increasing average tracking duration from 1.4s to 5.6s.

The paper replicated is "Automatic methods for long-term tracking and the detection and decoding of communication dances in honeybees."

Tracking Improvement

Optimized the computer vision tracking system for untagged bees, dramatically increasing average tracking duration.

4x IMPROVEMENT
BEFORE1.4s

AVG TRACKING DURATION

AFTER5.6s

AVG TRACKING DURATION

Research Setup

8 weeks

RECORDING DURATION

24hr

CONTINUOUS MONITORING

4 hives

BEEHIVE COUNT

YOLO v8
VISION MODEL