Keywords: fruit flies behavior, behavioral clustering, fruit flies sleep patterns
Project partner: Hugo Gillet, Dr. Alexandra M. Medeiros Vieira da Silva, Prof. Dr. Anissa Kempf, Biozentrum, University of Basel
CeDA collaborator: Konstantinos Ntemos
Repository: behavioral-fingerprinting-of-fruit-flies

Project objectives

  • Develop machine learning models to accurately capture the behavior patterns of fruit flies.
  • Discover distinct sleep patterns on fruit flies.

Approaches

  • Given a video of a fruit fly as input, fit a skeleton (with each point representing a distinct body part of the fly) on the depicted fruit fly.
  • Given the spatial positions of the body parts of the fly over time, perform clustering to capture distinct behavioral patterns (unspervised).
  • Use of annotated video clips of distinct fly behavioral patterns, build classification models to enhance the unsupervised learning performance (semi-supervised).



Figure 1. Example of a video frame depicting a fruit fly performing Proboscis Extension. As we observe on the upper left corner, the classification algorithm accurately captures this behavior ΒΆ