Establish a simple and quantitative deep learning-based method to analyse complicated intra- and inter-species social interaction behaviour for four stag beetle species.

Abstract

Stag beetles (Lucanidae) exhibit diverse social behaviours, yet quantifying these interactions remains challenging. Understanding social interactions within and between species is crucial for comprehending their behaviour, ecology and evolution. Stag beetles exhibit diverse social behaviours, including intraspecific competition, courtship and interspecific interactions, often involving complex physical displays and subtle cues. Traditional ethological methods for analysing these behaviours are time-consuming, subjective and limited in their ability to capture the nuances of dynamic interactions. This project aims to develop a simple and quantitative deep learning-based method to analyse complicated intra- and inter-species social interaction behaviour in four stag beetle species. This study utilizes DeepLabCut™ (DLC), a state-of-the-art deep learning-based pose estimation tool, to analyse and compare intra- and inter-species social interactions in four stag beetle species: Phalacrognathus muelleri, Prosopocoilus astacoides, Dorcus titanus and Prosopocoilus inclinatus. High-resolution videos of staged encounters were collected, and DLC was trained to accurately track key body parts of individual beetles. Behavioural parameters such as distance between individuals, orientation angles and movement trajectories were extracted from the pose data. Statistical analyses were conducted to identify species-specific differences in social behaviour, including aggression levels, courtship displays and dominance hierarchies. This study demonstrates the effectiveness of DLC in objectively quantifying complex social interactions in insects, providing valuable insights into the social ecology and evolutionary divergence of stag beetles.