Project: “CollembolAI”: Computer vision applied to soil mesofauna.
We are adapting computer vision approaches to assist the analyses of soil mesofauna communities and to assists with the digitization of vast collection of tiny animals in ethanol, such as the Collembola and Protura specimens that make up the majority of the Apterygota collection in Senckenberg Görlitz.
We develop an automatic macrophotography platform to capture the content of petri dishes in high definition. We also develop a deep-learning application to assist in the creation of object-detectors and classifiers that can automatically count and classify the animals on pictures (https://github.com/stasys-hub/Collembola_AI).
CollembolAI is under active development. We aim at creating machine-learning models that will recognize the major groups of Collembola (being our core expertise) but also the major groups of soil invertebrates. We also aim at bringing computer vision and molecular techniques together in order to create an efficient pipeline to characterize soil communities species diversity, species abundances and biomass, and species morphology.
CollembolAI is developed in collaboration with Stanislav Sys, Stephan Weißbach and Prof. Dr. Susanne Gerber, Institute of Human Genetics, Johannes Gutenberg-University Mainz