DISK Schema
Schema: disk
This schema integrates DISK-based imputation for pose and mocap streams. It is an optional enhancement layer used when missing-value reconstruction is needed.
What this schema answers
Which DISK model was used for an imputation run?
Which imputation task was staged and executed?
How many frames/values were imputed and with what uncertainty?
Core ADAMACS tables
disk.DISKModeldisk.DLCImputationTask,disk.DLCImputation,disk.DLCImputation.BodyPartPositionDISKdisk.MocapImputationTask,disk.MocapImputationdisk.MocapImputation.TrackingPositionDISKdisk.MocapImputation.RigidBodyPositionDISK
Query anchors
disk_model_nametask_modesession_idscan_id
scan_key = {'session_id': 'sessXXXXXXXX', 'scan_id': 'scanXXXXXXXX'}
(disk.DLCImputationTask & scan_key)
(disk.MocapImputationTask & scan_key)
Diagram
Notes for users
Treat imputed outputs as a distinct data product, not a silent replacement of raw values.
Preserve uncertainty and mask fields in downstream analyses.
Report model/version metadata whenever imputed values are used.
External references
DISK source and README: https://github.com/bozeklab/DISK
DeepLabCut docs (for upstream pose streams): https://deeplabcut.github.io/DeepLabCut/
Element DeepLabCut integration: https://docs.datajoint.com/elements/element-deeplabcut/latest/
DataJoint query operators: https://docs.datajoint.com/core/datajoint-python/latest/query/operators/