Studying termite collective behavior and movement patterns have been giving us insight into how complex phenomena may emerge from simple, individual actions. Although interesting and relevant, studies of this kind are often difficulted by the hardship of obtaining data from real and living samples. Even when confined and observed in controlled ambient and with aid of video recording, the raw footage material requires the attention and time from a human observer to extract the relevant data from video. In this paper we present a few computational tools, based on digital image processing and machine learning, devised to facilitate this process of data extraction. Using Python and scientific open source packages we were able to track termites on video and report their position, encounters, distances and social roles (workers or soldiers) in an almost one hundred percent automated method, while providing a data pipeline that reports the experiments results in a easy way for use in other tools of data analysis. The use of modern tracking algorithms and offline trained neural networks enables that this process occurs in almost real time on a regular modern computer. Funding: FAPEMIG (Fundação de Amparo à Pesquisa do Estado de Minas Gerais)

Funding: FAPEMIG (Fundação de Amparo à Pesquisa do Estado de Minas Gerais)