Data Pipeline

Each node in the following diagram represents the analysis code in the pipeline and the corresponding table in the database. Within the pipeline, Element MoSeq connects to upstream Elements including Lab, Animal, Session, and Event. For more detailed documentation on each table, see the API docs for the respective schemas.

The Element is composed of two main schemas, moseq_train and moseq_infer. The moseq_train schema is designed to handle the analysis and ingestion of PCA model for formatted keypoint tracking and train the Kepoint-MoSeq model. The moseq_infer schema is designed to handle the analysis and ingestion of Keypoint-MoSeq's motion sequencing on video recordings by using one registered model.

Diagrams

moseq_train module

• The moseq_train schema is designed to handle the analysis and ingestion of PCA model for formatted keypoint tracking and train the Kepoint-MoSeq model.

  

moseq_infer module

• The moseq_infer schema is designed to handle the analysis and ingestion of Keypoint-MoSeq's motion sequencing on video recordings by using one registered model.

  

Table Descriptions

lab schema

• For further details see the lab schema API docs

Table	Description
Device	Scanner metadata

subject schema

• Although not required, most choose to connect the Session table to a Subject table.

• For further details see the subject schema API docs

Table	Description
Subject	Basic information of the research subject

session schema

• For further details see the session schema API docs

Table	Description
Session	Unique experimental session identifier

moseq_train schema

• For further details see the moseq_train schema API docs

Table	Description
KeypointSet	Store keypoint data and video set directory for model training.
KeypointSet.VideoFile	IDs and file paths of each video file that will be used for model training.
Bodyparts	Store the body parts to use in the analysis.
PCATask	Staging table to define the PCA task and its output directory.
PCAPrep	Setup the Keypoint-MoSeq project output directory (kpms_project_output_dir) creating the default config.yml and updating it in a new dj_config.yml.
PCAFit	Fit PCA model.
LatentDimension	Calculate the latent dimension as one of the autoregressive hyperparameters (ar_hypparams) necessary for the model fitting.
PreFitTask	Specify parameters for model (AR-HMM) pre-fitting.
PreFit	Fit AR-HMM model.
FullFitTask	Specify parameters for the model full-fitting.
FullFit	Fit the full (Keypoint-SLDS) model.

moseq_infer schema

• For further details see the moseq_infer schema API docs

Table	Description
Model	Register a model.
VideoRecording	Set of video recordings for the Keypoint-MoSeq inference.
VideoRecording.File	File IDs and paths associated with a given recording_id.
PoseEstimationMethod	Pose estimation methods supported by the keypoint loader of keypoint-moseq package.
InferenceTask	Staging table to define the Inference task and its output directory.
Inference	Infer the model from the checkpoint file and save the results as results.h5 file.
Inference.MotionSequence	Results of the model inference.
Inference.GridMoviesSampledInstances	Store the sampled instances of the grid movies.