Item

Abstract

Quantifying behavior is crucial for many applications in neuroscience. Videography provides easy methods to observe animals, yet extracting particular aspects of a behavior can be highly time consuming. In motor control studies, humans or other animals are often marked with reflective markers
to assist with computer-based tracking, yet markers are intrusive, especially for smaller animals, and the number and location of the markers must be determined a priori. Here we provide a highly efficient method of markerless tracking in mice based on transfer learning with very few training samples (~ 200 frames). We demonstrate the versatility of this framework by tracking various body parts of mice in different tasks: odor trail-tracking (by one or multiple mice simultaneously), and a skilled forelimb reach
and pull task. For example, during the skilled reaching behavior, individual digit joints can be automatically tracked from the hand. Remarkably, even when a small number of frames are labeled, the algorithm achieves excellent tracking performance on test frames that is comparable to human accuracy.