Behavioral Clustering#
Motion Sequencing (MoSeq): a method to discover the syllables and grammar that comprise mouse 🐁 behavior. Recently, Caleb Weinreb and colleagues, bioRxiv 2023 have been expanding this approach, which originally relied on depth sensing technology, to work with pose estimation data. Thus, it is broadly applicable, independent of what cameras were used in experiments. That makes it ideally suited for this course.
Figure showing example behavioral syllables, Source: Datta Lab, Harvard Medical School.
A detailed overview of keypoint-MoSeq can be found on the project’s website.
For this course, we recommend to go over this example Colab notebook (with example data, it will run for about 40 min): https://colab.research.google.com/github/dattalab/keypoint-moseq/blob/main/docs/keypoint_moseq_colab.ipynb
Do you wonder what you’ll get without running it? Check this doc’s entry out for an already run notebook! You can also find a more minimal notebook, in the docs that runs in less time!
If you ran Keypoint-Moseq, you can also analyze the syllable segmentations.
Code Repository: dattalab/keypoint-moseq Data: https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fdrive.google.com%2Fdrive%2Ffolders%2F1UNHQ_XCQEKLPPSjGspRopWBj6-YNDV6G%3Fusp%3Dshare_link FAQ: https://keypoint-moseq.readthedocs.io/en/latest/FAQs.html