B-soID: unsupervised behavior analysis

Eric Yttri from Carnegie Mellon University has shared the following about B-SOiD, an open source unsupervised algorithm for discovery of spontaneous behaviors:

“Capturing the performance of naturalistic behaviors remains a tantalizing but prohibitively difficult field of study – current methods are difficult, expensive, low temporal resolution, or all of the above. Recent machine learning applications have enabled localization of limb position; however, position alone does not yield behavior. To provide a high temporal resolution bridge from positions to actions and their kinematics, we developed Behavioral Segmentation of Open-field In DeepLabCut, or B-SOiD. B-SOiD is an unsupervised learning algorithm that discovers and classifies actions based on the inherent statistics of the data points of the data points provided (including any marker or markerless system, not just deeplabcut). Our algorithm enables the automated segregation of different, sub-second behaviors with a single bottom-up perspective video camera – and does so without considerable effort or potential bias from the user. This open-source platform opens the door to the efficient study of spontaneous behavior and its neural mechanisms. It also readily provides critical behavioral metrics that historically have been difficult to quantify, such as grooming and stride-length in OCD and stroke research.”

This research tool was created by your colleagues. Please acknowledge the Principal Investigator, cite the article in which the tool was described, and include an RRID in the Materials and Methods of your future publications.  Project portal RRID:SCR_021410; Software RRID:SCR_021385