Granted Patent B2

US 12,519,990 · App. 18/676,456 · Granted Jan 6, 2026

Inventors: Khaldun Matter Ahmad AlDarabsah (Santa Clara, CA); Hailong Geng (Beijing, CN); Yu Tao Zhao (Olympia, WA); Yoshihiro Tanaka (Redmond, WA); Haofei Wang (Redwood City, CA); Mark Alden Rotblat (Lafayette, CA); Jaya Kawale (San Jose, CA); Chang She (San Francisco, CA); Marios Assiotis (Park City, UT); Joseph Gallagher (San Francisco, CA); Chiyu Zhong (Bloomington, IN); Amir Mazaheri (Mountain View, CA)

Assignee: Tubi, Inc.

H04N21/23424G06Q30/0245G06Q30/0251G06Q30/0277G06V10/70G06V10/774G06V10/776G06V20/41G06V20/46H04N21/251H04N21/26208

Abstract

Claims (73)

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, according to a predefined selection procedure, a subset of frames of a media item and analyze the subset to determine a set of candidate entity-probability pairs, each representing a probability of a brand being associated with the media item, wherein the probability exceeds a predefined probability threshold, the predefined probability threshold optionally comprising a tuned hyperparameter value;

a voting engine configured to select at least one final brand for association with the media item in a transcoding repository based on a count of occurrences of the final brand within the set of candidate entity-probability pairs;

an online transcoding service configured to interface with the transcoding repository to determine occurrences of the final brand within a set of media items served to a set of recipients, wherein the set of media items comprises the media item;

a frequency management service configured to:

store a quantity of impressions associated with the final brand for each of the set of recipients in a lookup repository;

manage frequencies of serving media content associated with the final brand to the set of recipients based on the quantity of impressions, wherein the quantity of impressions for each recipient is incremented for every impression by the recipient of media content associated with the final brand; and

calculate a fingerprint of the media item using a deterministic algorithm, wherein the online transcoding service is further configured to query the transcoding repository with the fingerprint of the media item to identify whether the media item has been previously processed.

2 . The system of claim 1 , wherein the frequency management service is operatively connected to a realtime bidding service, and wherein the frequency management service is further configured to:

increment and track the quantity of impressions associated with the final brand for each of a set of digital advertisements provided for the set of recipients by the realtime bidding service;

receive a request for a digital advertisement for a recipient of the set of recipients; and

exclude a subset of the set of digital advertisements from fulfilling the request based on determining that the quantity of impressions associated with the final brand and the recipient over a duration of time exceeds a predefined frequency management limit.

3 . The system of claim 1 , wherein the frequency management service is further configured to:

receive a request for media content for a recipient of the set of recipients;

identify a candidate media item for fulfilling the request; and

wherein the online transcoding service is further configured to:

obtain a fingerprint of the candidate media item;

query a transcoding repository with the fingerprint of the candidate media item;

obtain a result for the query indicating that the candidate media item is not tracked in the transcoding repository; and

queue the candidate media item for transcoding, wherein the candidate media item is not throttled for the request by the frequency management service based on the result for the query indicating that the candidate media item is not tracked in the transcoding repository.

4 . The system of claim 1 , wherein selecting the at least one final brand for association with the media item comprises:

calculating a ranking score for each of the subset of frames of the media item based on at least one general characteristic;

sorting each of the subset of frames of the media item by their ranking score; and

selecting a portion of the subset of frames based on the sorting.

5 . The system of claim 1 , wherein the deterministic algorithm matches multiple similar media items to the same fingerprint.

6 . The system of claim 1 , wherein an offline transcoding service is further configured to store multiple brand probability pairs in the transcoding repository with association to the media item.

7 . The system of claim 1 , wherein probabilities of the candidate entity-probability pairs each represent at least one selected from a group consisting of (i) a likelihood that the media item contains the associated entity and (ii) a strength of association between the entity and the media item.

8 . A method for entity detection using artificial intelligence, comprising:

selecting, according to a predefined selection procedure, a subset of frames of a media item and analyzing, using at least one computer processor, the subset to determine a set of candidate entity-probability pairs, each representing a probability of a brand being associated with the media item, wherein the probability exceeds a predefined probability threshold, the predefined probability threshold optionally comprising a tuned hyperparameter value;

selecting at least one final brand for association with the media item in a transcoding repository based on a count of occurrences of the final brand within the set of candidate entity-probability pairs;

interfacing with the transcoding repository to determine occurrences of the final brand within a set of media items served to a set of recipients, wherein the set of media items comprises the media item;

storing a quantity of impressions associated with the final brand for each of the set of recipients in a lookup repository;

managing frequencies of serving media content associated with the final brand to the set of recipients based on the quantity of impressions, wherein the quantity of impressions for each recipient is incremented for every impression by the recipient of media content associated with the final brand; and

calculating a fingerprint of the media item using a deterministic algorithm, wherein the online transcoding service is further configured to query the transcoding repository with the fingerprint of the media item to identify whether the media item has been previously processed.

9 . The method of claim 8 , further comprising:

incrementing and tracking the quantity of impressions associated with the final brand for each of a set of digital advertisements provided for the set of recipients by a realtime bidding service;

receiving a request for a digital advertisement for a recipient of the set of recipients; and

excluding a subset of the set of digital advertisements from fulfilling the request based on determining that the quantity of impressions associated with the final brand and the recipient over a duration of time exceeds a predefined frequency management limit.

10 . The method of claim 8 , further comprising:

receiving a request for media content for a recipient of the set of recipients;

identifying a candidate media item for fulfilling the request;

obtaining a fingerprint of the candidate media item;

querying a transcoding repository with the fingerprint of the candidate media item;

obtaining a result for the query indicating that the candidate media item is not tracked in the transcoding repository; and

queueing the candidate media item for transcoding, wherein the candidate media item is not throttled for the request by the frequency management service based on the result for the query indicating that the candidate media item is not tracked in the transcoding repository.

11 . The method of claim 8 , wherein selecting the at least one final brand for association with the media item comprises:

calculating a ranking score for each of the subset of frames of the media item based on at least one general characteristic;

sorting each of the subset of frames of the media item by their ranking score; and

selecting a portion of the subset of frames based on the sorting.

12 . The method of claim 8 , wherein the deterministic algorithm matches multiple similar media items to the same fingerprint.

13 . A non-transitory computer-readable storage medium comprising a plurality of instructions for entity detection using artificial intelligence, the plurality of instructions configured to execute on at least one computer processor to enable the at least one computer processor to:

select, according to a predefined selection procedure, a subset of frames of a media item and analyze the subset to determine a set of candidate entity-probability pairs, each representing a probability of a brand being associated with the media item, wherein the probability exceeds a predefined probability threshold, the predefined probability threshold optionally comprising a tuned hyperparameter value;

select at least one final brand for association with the media item in a transcoding repository based on a count of occurrences of the final brand within the set of candidate entity-probability pairs;

interface with the transcoding repository to determine occurrences of the final brand within a set of media items served to a set of recipients, wherein the set of media items comprises the media item;

store a quantity of impressions associated with the final brand for each of the set of recipients in a lookup repository;

manage frequencies of serving media content associated with the final brand to the set of recipients based on the quantity of impressions, wherein the quantity of impressions for each recipient is incremented for every impression by the recipient of media content associated with the final brand; and

calculate a fingerprint of the media item using a deterministic algorithm, wherein the online transcoding service is further configured to query the transcoding repository with the fingerprint of the media item to identify whether the media item has been previously processed.

14 . The non-transitory computer-readable storage medium of claim 13 , the plurality of instructions further configured to execute on the at least one computer processor to enable the at least one computer processor to:

increment and track the quantity of impressions associated with the final brand for each of a set of digital advertisements provided for the set of recipients by a realtime bidding service;

receive a request for a digital advertisement for a recipient of the set of recipients; and

exclude a subset of the set of digital advertisements from fulfilling the request based on determining that the quantity of impressions associated with the final brand and the recipient over a duration of time exceeds a predefined frequency management limit.

15 . The non-transitory computer-readable storage medium of claim 13 , the plurality of instructions further configured to execute on the at least one computer processor to enable the at least one computer processor to:

receive a request for media content for a recipient of the set of recipients;

identify a candidate media item for fulfilling the request;

obtain a fingerprint of the candidate media item;

query a transcoding repository with the fingerprint of the candidate media item;

obtain a result for the query indicating that the candidate media item is not tracked in the transcoding repository; and

queue the candidate media item for transcoding, wherein the candidate media item is not throttled for the request based on the result for the query indicating that the candidate media item is not tracked in the transcoding repository.

16 . The non-transitory computer-readable storage medium of claim 13 , wherein selecting the at least one final brand for association with the media item comprises:

calculating a ranking score for each of the subset of frames of the media item based on at least one general characteristic;

sorting each of the subset of frames of the media item by their ranking score; and

selecting a portion of the subset of frames based on the sorting.

Assignments (1)

Continuity (3)

References Cited (100)