Determine which video-level computer vision task you need to perform based on the insights you want to gain.