In This Article

  • How to run automatic head detection

  • How to review head detection

  • How to remove an automatic head detection

  • How to resize head groups


Veritone Redact’s Head Detection engine works by scanning through your video frame by frame looking for any objects that appear to be a head. The engine will look for a pair of shoulders with something round on top. As long as a head is clearly visible on-screen, the Head Detection engine will track this head as it moves across the screen.

If you’d like to learn more about how Veritone Redact’s head detection engine works, please view this help article.

PRO TIP: The Head Detection engine is not facial recognition. For example, the engine is unable to detect the difference between a civilian and a police officer.

How to Run Automatic Head Detection

After selecting the media that you would like to upload into Veritone Redact, simply select “Run Head and Object Detection” and click “Upload”. This will scan your media for heads during the upload process.

If you did not run head and object detection while uploading your media, you can do so from the File Manager Page. Click on the file name on the Redact Homepage to open this page.

Click Detect Heads and Objects on the Video tab. The head detection cognitive engine will commence processing. You can perform other work on the file while waiting for head detection to finish, i.e. add user-defined regions, run audio transcription, etc. or return to the homepage to work on a different file.

Note: Processing times vary based on the size of the file. Cognitive engine processing occurs in the cloud and does not require the browser to be open.

The processing status of the file's head detection can be viewed in the Notification Center.

How to Review Head Detection

When head detection processing is complete, enable all automatic head detections and populate the Head timeline. To do this, select the checkbox below Detect Heads and Objects, under the Video panel on the right side.

Auto-detected heads are outlined or blackfilled depending on the Preview selection made in the lower left corner of the page. The boxed area represents the area that will be obfuscated once the video is processed for redaction.

Under the Video tab located on the right-hand side of the File Manager Page, each thumbnail image represents a single frame of video that contains a detected head. To advance the video to the specific frame at which that head appears, click on the corresponding time segment. You can drag the red marker handle on the scrub bar or utilize keyboard shortcuts to advance at different speeds as desired.

How to Delete an Automatic Head Detection

If there is a false positive or a head that does not need to be redacted, you can simply uncheck the box associated with that specific head.

A detected head must be selected for the redaction mask to display in the video.

How to Resize Head Groups

All detected heads have a default padding of 20% added to account for movement. You can increase or decrease this padding by selecting the bounding box to the right of the timelines and adjusting accordingly.

If an auto-detection mask is not correctly sized or placed, you can re-size or adjust its positioning by clicking into the green box and using your mouse to adjust as needed.

You can also re-size or adjust a single headgroup if an auto-detection mask is not correctly sized or placed. To do so, click into the green head detection box and use your mouse to drag the corner and adjust as needed.

Note: If you noticed a head, or in some instances, a partial head due to an obstruction in the media, was not captured by our head detection engine, you will need to draw a User Defined Region (UDR) to redact this. Learn more about UDRs here.

If you have any questions about head detection, please contact the Veritone Support team at [email protected].

Did this answer your question?