Abstract: Some implementations include methods generating visualization emphasis object for players and may include receiving a video clip associated with a sporting event participated by a plurality of players using a playing field with a play object. The players in a frame of the video clip may be detected. Players who are on the playing field may be identified from the detected players. Each of the players identified to be on the playing field may be associated with a rectangular bounding box that provides an outline of each of the players. A player who has possession of the play object may be identified. A visualization emphasis object may be generated and placed on the player who has possession of the play object. The visual emphasis object may have a size proportional to height of a bounding box associated with the player having the possession of the play object.

Abstract: Techniques for reducing the latency of annotating a replay video segment may include receiving a video segment with content involving multiple individuals. An annotation task is performed concurrently with a tracking task. The annotation task receives annotation data to indicate which of the individuals is an individual of interest in a subset of frames of the video segment, and the tracking task tracks the individuals in the video segment by generating bounding objects corresponding to the individuals. The annotation data can be associated with the bounding objects to detect a bounding object for the individual of interest, and a visualization emphasis object is generated based on the detected bounding object in a replay video segment to identify the individual of interest.

Abstract: Systems and techniques for ranking and selection of headshots from a collection of images. The systems and techniques for ranking use heuristics to rank extracted headshots from a set of images based on features of the faces within the images. The ranking may be used to generate a training dataset for a machine learning model to determine quality scores for faces within images. The ranked images may then be stored for later access in reference to a video or set of images containing the individual.