US 11,776,262 B1
Identifying compelling scenes at scale by mapping characteristics
Kewen Chen, King, WA (US); Tu Anh Ho, Mercer Island, WA (US); Christina Botkins, Seattle, WA (US); and Najmeh Sadoughi Nourabadi, Bellevue, WA (US)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Jun. 28, 2021, as Appl. No. 17/304,932.
Int. Cl. G06V 20/40 (2022.01); H04N 21/234 (2011.01); H04N 21/25 (2011.01); H04N 21/2668 (2011.01); G06F 18/21 (2023.01)
CPC G06V 20/41 (2022.01) [G06F 18/21 (2023.01); G06V 20/46 (2022.01); H04N 21/23418 (2013.01); H04N 21/251 (2013.01); H04N 21/2668 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
identifying a plurality of indices, each index corresponding to a continuous portion of a media presentation that is associated with one or more timestamps or one or more frames;
determining a set of attribute scores for each index, wherein each attribute score represent a confidence level that a corresponding attribute applies to a corresponding portion of the media presentation, wherein at least some of the attribute scores are determined by applying a classifier trained on clips that correspond with one of a plurality of action attributes or one of a plurality of emotion attributes;
applying a set of weights to the set of attribute scores, each weight representing a compelling-ness of an attribute;
determining a compelling-ness score for each index based on the set of attribute scores and the set of weights;
identifying a plurality of clips based on the compelling-ness scores, wherein each clip includes a contiguous range of content of the media presentation beginning and ending at corresponding shot boundaries, each clip corresponding to a contiguous range of the plurality of indices;
determining a clip compelling-ness score for each clip of the plurality of clips based on the compelling-ness score for each index included in that clip;
ranking the plurality of clips based on the clip compelling-ness scores;
identifying a subset of the plurality of clips based on the ranking of the plurality of clips, and
providing one or more clips of the subset of the plurality of clips to a first client device.