Implementation of a keypoint-adapted-SLDS aproach for unsupervised behavior discovery. Check out the example keypoint notebook for how to use the code.
The standard MoSeq HDP-ARHMM can also be run using code from this repo. The model is identical to that in pyhsmm-autoregressive. The key differences are that this new code is written with jax; lacks C-dependencies; and is functional rather than object oriented. Example AR-HMM code is provided in the depth-PCA notebook.