Alisa Ficiciyan
Research Mentor: Bing Ye
Mentor Department: Life Sciences Institute, University of Michigan, Medicine
Author(s): Alisa Ficiciyan, Robert Tomlinson, Bing Ye
Session: Session 2 (10:00 AM – 10:50 AM)
Presentation Type: Poster 62
Abstract
Accurate quantification of animal behavior is crucial for studying neural basis of behavior and for developing treatments of medical conditions with behavioral manifestation. LabGym is an open-source software tool that uses artificial intelligence (AI) to analyze animal movements and actions from video recordings. In this project, I focused on improving the accessibility of LabGym to bridge technical gaps between cross-disciplinary researchers in diverse fields such as neuroscience, psychology, evolutionary biology, and more. I evaluated LabGym’s existing codebase and tested key features, including its background subtraction and Detectron2-enabled ‘detector’ tracking methods, to process a completely automated behavior analysis workflow. I also worked through a workflow on another similar deep-learning-based AI software called SLEAP, to make comparisons between it and LabGym. Throughout this process, I looked for ways to make LabGym simpler, more useful, and more user-friendly, aiming to help scientists who may not have advanced computer skills. Building on this evaluation, I integrated resources directly into the LabGym interface, including links to the main website and module-specific tutorial videos at relevant workflow steps. I also began developing a workflow map within the UI to clarify how LabGym’s modules connect and to guide users through the overall workflow process. Planned enhancements include progress bars for processes in key modules, displaying per-behavior example counts in the categorizer training module, and a “model library” to help users review, organize, and share training configurations for their categorizers. Overall, these improvements make LabGym more accessible and powerful, enabling scientists across disciplines to analyze animal behavior more efficiently.


