US 12,022,138 B2
Model serving for advanced frequency management
Khaldun Matter Ahmad AlDarabsah, Santa Clara, CA (US); Hailong Geng, Beijing (CN); Yu Tao Zhao, Olympia, WA (US); Yoshihiro Tanaka, Redmond, WA (US); Haofei Wang, Redwood City, CA (US); Mark Alden Rotblat, Lafayette, CA (US); Jaya Kawale, San Jose, CA (US); Chang She, San Francisco, CA (US); Marios Assiotis, Park City, UT (US); Joseph Gallagher, San Francisco, CA (US); Chiyu Zhong, Bloomington, IN (US); and Amir Mazaheri, Mountain View, CA (US)
Assigned to Tubi, Inc., San Francisco, CA (US)
Filed by Tubi, Inc., San Francisco, CA (US)
Filed on Feb. 21, 2022, as Appl. No. 17/676,760.
Claims priority of provisional application 63/213,177, filed on Jun. 21, 2021.
Prior Publication US 2022/0405809 A1, Dec. 22, 2022
Int. Cl. G06Q 30/00 (2023.01); G06Q 30/0241 (2023.01); G06Q 30/0242 (2023.01); G06Q 30/0251 (2023.01); G06V 10/70 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 20/40 (2022.01); H04N 21/234 (2011.01); H04N 21/25 (2011.01); H04N 21/262 (2011.01); G05B 19/418 (2006.01)
CPC H04N 21/23424 (2013.01) [G06Q 30/0245 (2013.01); G06Q 30/0251 (2013.01); G06Q 30/0277 (2013.01); G06V 10/70 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 20/41 (2022.01); G06V 20/46 (2022.01); H04N 21/251 (2013.01); H04N 21/26208 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A system for entity detection using artificial intelligence, comprising:
a computer processor;
a deep learning model service executing on the computer processor and configured to enable the computer processor to:
select a set of frames from a media item according to a predefined selection procedure;
analyze the set of frames to determine a set of candidate brand-probability pairs, each associated with one frame of the set of frames;
a voting engine configured to execute a voting selection procedure, comprising:
obtaining, as an input to the voting selection procedure, at least one hyperparameter value representing a threshold for determining whether candidate brand-probability pairs are to be included in a result set;
determining that a first brand-probability pair of the set of candidate brand-probability pairs does not meet the threshold;
excluding the first brand-probability pair from the result set based on determining that the first brand-probability pair does not meet the threshold;
sorting the result set by a count of occurrences of each brand of the candidate brand-probability pairs in the set of frames; and
selecting at least one final brand-probability pair from the result set; and
an offline transcoding service configured to:
store the final brand-probability pair in a repository with a relation to an identifier of the media item, wherein the repository is configured to query for the final brand-probability pair by the identifier of the media item; and
a frequency management service configured to:
receive a request for a digital advertisement, the request comprising a recipient identifier;
identify a candidate digital advertisement obtained from a realtime bidding service;
identify a brand identifier associated with the candidate digital advertisement, wherein the brand identifier corresponds to the final brand-probability pair stored in the repository;
perform a query against the repository, wherein the query comprises the brand identifier and the recipient identifier;
receive a response from the repository comprising a quantity of impressions associated with the brand identifier and the recipient identifier over a duration of time;
determine that the quantity of impressions exceeds a predefined frequency management limit; and
exclude the candidate digital advertisement from a result set provided for display by a client device associated with the recipient identifier.