Model Schema

Schema: model

This schema manages video recordings, model definitions, and pose-estimation outputs. It is the main route from camera streams to body-part trajectories.

What this schema answers

  • Which recording/model pair produced a pose estimate?

  • Which task is staged (load/trigger) and where is output stored?

  • Which body-part trajectories are available for alignment with events or neural data?

Core tables used most often

  • model.VideoRecording, model.VideoRecording.File

  • model.VideoRecordingNew, model.VideoRecordingNew.File

  • model.Model, model.Model.BodyPart

  • model.PoseEstimationTask, model.PoseEstimation

  • model.PoseEstimationTaskNew

  • model.RecordingInfo, model.RecordingInfoNew

Query anchors

  • recording_id

  • model_name

  • task_mode

(model.PoseEstimationTaskNew & {'scan_id': 'scanXXXXXXXX'})
(model.PoseEstimationNew & {'recording_id': 'LE_eye_left_tracking_YYYY-MM-DD'})

Diagram

Model schema

Notes for users

  • Keep recording_id and model_name explicit in all notebook outputs.

  • Validate task_mode before interpreting empty computed tables.

  • Use model body-part vocabularies as controlled contracts across analyses.

Element lineage

ADAMACS model tables build on Element DeepLabCut patterns. Reference: https://docs.datajoint.com/elements/element-deeplabcut/latest/

External references