US 12,328,479 B2
Systems and methods for real-time processing of live media streams
Stephan Leroux, East Gwillimbury (CA)
Assigned to Shopify Inc., Ottawa (CA)
Filed by Shopify Inc., Ottawa (CA)
Filed on Sep. 2, 2022, as Appl. No. 17/901,898.
Prior Publication US 2024/0080529 A1, Mar. 7, 2024
Int. Cl. H04N 21/47 (2011.01); G06Q 30/0601 (2023.01); G06V 20/40 (2022.01); H04N 5/268 (2006.01); H04N 21/431 (2011.01); H04N 21/472 (2011.01); H04N 23/69 (2023.01)
CPC H04N 21/47217 (2013.01) [G06Q 30/0603 (2013.01); G06V 20/41 (2022.01); H04N 5/268 (2013.01); H04N 21/431 (2013.01); H04N 23/69 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
obtaining video data for a video stream;
setting a detected first subject of the video stream as a default object of interest;
detecting, in at least one video frame of the video stream, a first object from a first set of defined objects associated with the video stream; and
in response to detecting the first object in the at least one video frame:
identifying the first object as a current target object associated with the video stream;
detecting a transition cue associated with the first object based on:
determining that the first object is associated with a smaller subject distance to a camera capturing the video stream than a previous target object that is different from the current target object; or
identifying keywords comprising descriptors of the first object in speech detected in audio data associated with the video stream; and
responsive to detecting the transition cue associated with the first object, causing the current target object to be displayed in video focus in at least one subsequent video frame of the video stream by causing transition from displaying a previous target object in video focus for the video stream, to displaying the current target object in video focus for the video stream;
determining that none of the first set of defined objects is detected in a further subsequent video frame of the video stream; and
in response to determining that none of the first set of defined objects is detected in the further subsequent video frame, returning video focus to the default object of interest based on setting the first subject as the current target object associated with the video stream.