Decoding Animal Behavior Through Pose Tracking

Explore cutting-edge techniques for decoding animal behavior through pose tracking in this comprehensive lecture. Delve into the world of behavioral quantification and its significance in neuroscience as Talmo Pereira from Princeton University presents advanced computational methods for automating animal behavior analysis. Learn how high-speed videography and pose tracking are revolutionizing our understanding of how neural circuits control complex behaviors in various species, from fruit flies to giraffes. Discover the application of deep learning and computer vision techniques in generalizing human pose estimation methods to animals, addressing challenges such as limited labeled data and multi-instance tracking. Gain insights into dataset-tailored neural network architecture design and unsupervised action recognition for interpreting unconstrained behavior. The lecture also includes a hands-on tutorial using the SLEAP framework, demonstrating how to train and evaluate deep learning models for animal pose tracking directly in the browser.

Intro
Behavior through the lens of neuroscience
Background: How do we track animals?
Deep learning for animal pose estimation: LEAP
SLEAP: Dealing with multiple instances (Top-down)
SLEAP: Dealing with multiple instances (Bottom-up)
General-purpose animal pose tracking
SLEAP: How do we adapt to different datasets?
SLEAP: Data-specific network architecture design
Unsupervised behavior recognition
Future directions