Virtual Reality and Tracking the Mating Behavior of Fruit Flies: a Machine Learning Approach

Abstract

Study of social behaviors of Drosophila melanogaster, i.e., fruit flies, is used to understand certain behaviors of human. Automation of capturing and real-time analysis of fruit fly social movements helps quantitative study of those behavior which in turns can be applied for studying human behavior. To achieve this, we present Virtual Reality environment setting to stimulate fruit flies behavior and tracking of their motions automatically in real-time. As an experiment in a real application, we selected study of mating behavior of fruit flies. Male fruit flies tend to extend their mating duration when exposed to rivals as published in previous biology studies. We designed a Virtual Reality environment where synthetic male fruit flies are virtually simulated to stimulate a male fruit fly to study the effect of rivals. Bezier curve fitting and Gaussian random distribution are utilized for movement simulation. A machine learning approaches (logistic regression) employed to track, detect, and classify fruit flies mating behavior. We performed connected component labeling as an operator for tracking and classification of mating and non-mating status for comparison purposes. The machine learning based approach shows superior result in terms of speed and accuracy.