Object Tracking

Object tracking in videos

Our solution features automatic object tracking in videos to ensure that moving objects are reliably detected and anonymised.

Understanding Object Tracking

Object tracking is a computer vision technique that follows objects as they move across frames in a video. We use object tracking in conjunction with our object detection algorithm. The object detection model identifies objects of interest (such as faces or license plates) in each frame. Then, the object tracking algorithm tracks these objects across frames, ensuring they are consistently anonymized even if they temporarily disappear from view.

Benefits of Object Tracking

• Reduced Flickering: Object tracking significantly reduces the occurrence of flickering detections, where an object is anonymized in one frame and then reappears unanonymized in the next. This improves the overall quality and consistency of the anonymized video.

• Improved Accuracy: By tracking objects across frames, our object tracking feature can help to correct errors in the object detection process. This leads to more accurate and reliable anonymization.

How to Use Object Tracking in Celantur

Object tracking is enabled by default for video anonymisation. You can disable it using --no-object-tracking as a command-line argument in Batch and Stream mode or no-object-trackingas query parameter in REST API (v1) mode.