Know where your dataset came from

Captured media carries a source snapshot and capture reason so you can trace an image after collection, filter the dataset and understand why a frame was selected. Source configuration and the history attached to a media item have different lifetimes.

What is recorded

  • Origin and source name at capture time.
  • Producer and session identifiers when available, to distinguish a selected camera, screen grant or video from a reusable source configuration.
  • Capture time and reason: manual, fixed interval, motion, low confidence or novelty.
  • For a still collected from a video-file source, the presentation timestamp of the captured pixels.

Use source history in the media grid

Use the source and capture-reason filters to inspect a particular collection or the images chosen by Smart capture. Combine filters with annotation status and the other media filters, then review or batch-annotate the relevant selection. Media information shows the available provenance; older items may have no capture metadata.

Preserve history without copying source credentials

The name is stored with the media so deleting or renaming a source does not erase its history. AnnotateIt project archives and its dataset metadata sidecar preserve supported provenance; third-party tools are not required to read that sidecar. Source credentials and stream URLs are not part of capture provenance.

Source metadata does not automatically prevent train/test leakage between similar frames from one recording. Plan the split deliberately and inspect related frames before evaluating a model.

See also

Video tutorial

AnnotateIt tutorial