Annotation pipeline
Data quality
The workflow that turns source data into labels ready for their intended use.
A pipeline can include selection, preparation, annotation, review, correction and release. Quality checks belong at explicit points in that workflow. Creating labels and checking labels are different jobs, and they can be performed by different teams or systems against the same acceptance criteria.
In practice
A team mines useful clips, has its supplier annotate them, then reviews the labels before assembling a training release.