Player not loading? Watch on YouTube
This tutorial explains region-based object tracking with Ultralytics TrackZone and YOLO26. The speaker uses Google Colab to define a zone and process a video, with examples of how selected regions could support traffic analysis or queue monitoring.
The walkthrough starts with installing the Ultralytics Python package and setting up video capture and an output writer. The speaker explains that Colab is a headless environment, so the notebook saves results to a video file for inspection after processing. The runtime can use a CPU or a selected GPU.
Region coordinates can describe a rectangle or an arbitrary polygon. The TrackZone configuration includes those points, a model, and a display setting. The speaker also explains how to supply a custom model path, select object classes, and adjust line width. Each video frame passes through the tracker, and the returned detections and counts can feed further analytics.
The custom-video example tracks people inside a polygon. The speaker reports roughly 25 to 30 frames per second for that run; this is a demonstration result, not a general performance guarantee. The tutorial also warns that uploaded Colab files disappear when the environment terminates, so users need to save their work.
