US 11,973,993 B2
Machine learning based media content annotation
Jonathan Bennett-James, Merthyr Tydfil (GB); and Craig Holbrook, Merthyr Tydfil (GB)
Assigned to NAGRAVISION S.A., Cheseaux-sur-Lausanne (CH)
Filed by NAGRAVISION S.A., Cheseaux-sur-Lausanne (CH)
Filed on Oct. 26, 2021, as Appl. No. 17/510,722.
Claims priority of provisional application 63/106,784, filed on Oct. 28, 2020.
Prior Publication US 2022/0132179 A1, Apr. 28, 2022
Int. Cl. H04N 21/235 (2011.01); G06N 20/00 (2019.01); G06V 20/40 (2022.01); H04N 21/266 (2011.01)
CPC H04N 21/2353 (2013.01) [G06N 20/00 (2019.01); G06V 20/47 (2022.01); H04N 21/26603 (2013.01)] 16 Claims
OG exemplary drawing
 
12. A system for annotating media content, including:
a memory; and
one or more processors coupled to the memory and configured to:
obtain media content;
generate, use one or more machine learning models, a metadata file for at least a portion of the media content, the metadata file including one or more metadata descriptions;
obtain a plurality of template sentences, each template sentence of the plurality of template sentences including one or more placeholder metadata tags;
determine a subset of metadata descriptions from the one or more metadata descriptions having confidence scores greater than a confidence threshold;
generate a plurality of sentences at least in part by replacing placeholder metadata tags of the plurality of template sentences with one or more metadata descriptions of the subset of metadata descriptions;
determine, using a machine learning model, a subset of sentences from the plurality of sentences to generate a scene description; and
annotate the media content use the scene description.