Automatically generated using scene segmentation, object detection, action recognition, and semantic analysis.