Vision-Based System for Automated Mouse Behavior Recognition

Automated quantitative analysis of animal behavior is an important tool for assessing the effects of diseases, drugs, gene mutations, and perturbations in neural circuits. Automating behavioral analysis addresses the inherent imitations of human assessment (time, cost, and reproducibility) and allows researchers to study behaviors over longer time scales than typically studied.

We provide a general purpose system for recognizing behavior in singly housed mice. The system has been trained to recognize 8 stereotypical mouse behaviors: eat, drink, groom, hang, micromovement, rear, rest, and walk. Given a video of a mouse in the home-cage, the system labels each frame with one of these 8 behaviors.

With our system, we have also released two databases of mice actions. The full database includes over ten hours of continuous labeled mice behavior. The clipped database includes 4200 short clips of the most unambiguous examples of each behavior. In the near future, we plan on releasing the training framework, so other researchers can use our framework to recognize new behaviors.

The software and database are described in the following article. We request that authors cite this paper in publications describing work carried out with this system and/or the video database.

H. Jhuang, E. Garrote, X. Yu, V. Khilnani, T. Poggio, A. Steele and T. Serre. Automated Home-Cage Behavioral Phenotyping of Mice. Nature communications, Sep., 2010 [pdf]
drink
eat
groom
hang
micro-movement
rear
rest
walk

System Download

The software is free for academic and non-commercial research (license). The author is not responsible for implications for the use of this software.

  • Testing Code - A simple gui for evaluating mice on the 8 stereotypical behaviors: drink, eat, groom, hang, micromovement, rear, rest, walk.
  • Training Code - A simple gui for training and testing new behaviors.

The softwared is available to download, please request a download link here. Report bugs to sbraun [at] mit.edu. Please see System Documentation section for installation instruction, User Guide, and FAQ.

Video Database Download

  • Full Database - Over 10.6 hours of continously labeled video (8 day videos and 4 night videos) for the eight behaviors of interest: drink, eat, groom, hang, micromovemnt, rear, rest, walk.
  • Clipped Database - 4200 short clips containing a single behavior. The clips were curated from 9000 to 4,200 clips so as to only include the most unambiguous examples of each behavior.

Demo

Documentation

Contact

For technical questions, comments and questions about the system and/or datasets, please contact Hueihan Jhuang (hueihan.jhuang [at] tuebingen.mpg.de)(hueihan@mit.edu is no longer in use) or Stav Braun (sbraun [at] mit.edu).

For potential collaborations, please contact Prof. Thomas Serre (Thomas_Serre [at] brown.edu) and Prof. Tomaso Poggio (tp [at] csail.mit.edu).

Change Log

  • 02/21/2012 - upload the link of 8 high resolution clips and the demo videos (host is changed from MITTechTV to YouTube)
  • 07/25/2011 - Testing + training system released