This project aims to link detailed facial biomechanics with brain-wide neural activity by integrating the Cheese3D multiview facial tracking system with the International Brain Laboratory (IBL) ecosystem. Cheese3D is a high-speed, multi-camera system that reconstructs three-dimensional facial kinematics of head-fixed mice with micron-scale precision across multiple anatomical landmarks, including the eyes, ears, whisker pad, and jaw. By capturing coordinated motion across the face, the system provides a rich behavioral readout that can reflect internal states and neural activity. The project will combine these multiview facial recordings with electrophysiology to investigate how facial muscle dynamics relate to neural population activity across distributed motor and limbic circuits.

The collaboration with the IBL Core will focus on integrating Cheese3D with modern machine-learning approaches and IBL data infrastructure. This includes training multiview pose estimation models using LightningPose, applying self-supervised video representation learning through the BEAST framework, and linking facial movement embeddings with Neuropixels recordings. Together, these efforts aim to create a scalable pipeline for multiview facial tracking and neural decoding, enabling researchers to infer internal state and neural activity from facial movement patterns. The resulting datasets, models, and analysis tools will be released as open-source resources to support broader use within the neuroscience community.