Efficient Pose Estimation in Social Insects Residing in Colonies Using DeepPoseKit

Abstract

Pose estimation is vital for figuring out how insects that live in colonies interact and conduct themselves, providing significant insights into their social dynamics and group decision-making. For the purpose of examining numerous aspects of insect behavior, such as feeding patterns, communication, and movement, accurate and effective pose estimation techniques are essential. Deep learning algorithms have become effective tools for pose estimation tasks over a wide range of domains in recent years. This research describes the use of DeepPoseKit, an advanced deep learning framework, for posture estimation of insects such as ants and termites which reside within colonies. A data set of high-definition footage of the insects striking various stances are used to train and evaluate the system. Our work includes the creation of an annotation collection, initialization of the annotations, and training of the model. The method offers a potent tool for behavior analysis and identifying fresh targets for intervention, which has the potential to revolutionize the area of insect study © 2023 IEEE.