Video object tagging based on machine learning

Abstract

Aspects of the subject disclosure may include, for example, a method in which a processing system obtains a sample of a content stream directed to a user device, identifies a type of the content stream, and selects a model for recognizing objects appearing in the content stream. The system analyzes the content stream in accordance with the model to recognize the object, generates a label for the object, and associates the label with the object in the content stream. The system also delivers the content stream for presentation at the user device; the label is delivered in-line with respect to the content stream and is generated in real time with respect to the presentation. The method further includes training the model in accordance with a machine learning procedure; the model is refined based on the analyzing of the content stream. Other embodiments are disclosed.

What is claimed is: 
     
         1 . A device comprising:
 a processing system including a processor; and   a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations, the operations comprising:   identifying a type of content in a content stream directed to a user device;   selecting, in accordance with the type of content, a model for use in recognizing an object appearing in the content stream, by mapping the type of content to a database storing a plurality of models, the database accessible to the processing system and indexed to the type of content, wherein the selected model is determined to be most likely of the plurality of models to yield accurate recognition of the object;   analyzing the content stream in accordance with the model to recognize the object;   generating a label for the object based on the analyzing;   associating the label with the object in the content stream;   determining, subsequent to the associating, that an attribute of the label matches a portion of a user profile;   delivering the content stream for presentation at the user device, the label being delivered in-line with respect to the content stream, the labeled object being presented in an enhanced format, wherein the labeled object represents an advertised product; and   training the model in accordance with a machine learning procedure, whereby the model is refined based on the analyzing of the content stream.   
     
     
         2 . The device of  claim 1 , wherein the operations further comprise obtaining a sample of the content stream, and wherein the type of content is identified based on the sample. 
     
     
         3 . The device of  claim 1 , wherein the user device is associated with a subscriber to a communication network. 
     
     
         4 . The device of  claim 3 , wherein the analyzing is performed without reference to interests of the subscriber. 
     
     
         5 . The device of  claim 3 , wherein the determining that an attribute of the label matches a portion of a user profile indicates that the label correlates with an interest of the subscriber. 
     
     
         6 . The device of  claim 1 , wherein the label is generated in real time with respect to the presentation. 
     
     
         7 . The device of  claim 1 , wherein the label has an accompanying a link to a site offering information regarding the advertised product. 
     
     
         8 . The device of  claim 1 , wherein the enhanced format comprises a display of the label, a highlighted display of the object, a display of the object at a higher resolution than that of another object in the content stream, a display of the object in a predefined portion of a display area of the user device, or a combination thereof. 
     
     
         9 . The device of  claim 1 , wherein a selectable item is displayed in association with the labeled object. 
     
     
         10 . The device of  claim 1 , wherein the operations further comprise extracting metadata relating to the labeled object, wherein the metadata is delivered in-line with respect to the content stream. 
     
     
         11 . A method comprising:
 identifying, by a processing system including a processor, a type of content in a content stream directed to a user device;   selecting, by the processing system in accordance with the type of content, a model for use in recognizing an object appearing in the content stream, by mapping the type of content to a database storing a plurality of models, the database accessible to the processing system and indexed to the type of content, wherein the selected model is determined to be most likely of the plurality of models to yield accurate recognition of the object;   analyzing, by the processing system, the content stream in accordance with the model to recognize the object;   generating, by the processing system, a label for the object based on the analyzing;   associating, by the processing system, the label with the object in the content stream;   determining, by the processing system, subsequent to the associating, that an attribute of the label matches a portion of a user profile;   delivering, by the processing system, the content stream for presentation at the user device, the label being delivered in-line with respect to the content stream, the labeled object being presented in an enhanced format, wherein the labeled object represents a selectable advertised product; and   training, by the processing system, the model in accordance with a machine learning procedure, whereby the model is refined based on the analyzing of the content stream.   
     
     
         12 . The method of  claim 11 , further comprising obtaining, by the processing system, a sample of the content stream, and wherein the type of content is identified based on the sample. 
     
     
         13 . The method of  claim 11 , further comprising extracting, by the processing system, metadata relating to the labeled object, wherein the metadata is delivered in-line with respect to the content stream. 
     
     
         14 . The method of  claim 11 , wherein the user device is associated with a subscriber to a communication network. 
     
     
         15 . The method of  claim 14 , wherein the content stream is produced by the communication network, and wherein the method is performed as a service provided at an edge of the communication network. 
     
     
         16 . The method of  claim 14 , wherein the content stream is delivered via a programmable data pipeline of the communication network, the data pipeline being programmed to generate the label. 
     
     
         17 . A non-transitory machine-readable medium comprising executable instructions that, when executed by a processing system including a processor, facilitate performance of operations, the operations comprising:
 identifying a type of content in a content stream directed to a user device, wherein the user device is associated with a subscriber to a communication network;   selecting, in accordance with the type of content, a model for use in recognizing an object appearing in the content stream, by mapping the type of content to a database storing a plurality of models, the database accessible to the processing system and indexed to the type of content, wherein the selected model is determined to be most likely of the plurality of models to yield accurate recognition of the object;   analyzing the content stream in accordance with the model to recognize the object;   generating a label for the object based on the analyzing;   associating the label with the object in the content stream;   determining, subsequent to the associating, that an attribute of the label matches a portion of a user profile;   delivering the content stream for presentation at the user device, the label being delivered in-line with respect to the content stream, the labeled object being presented in an enhanced format, wherein the labeled object represents an advertised product; and   training the model in accordance with a machine learning procedure, whereby the model is refined based on the analyzing of the content stream.   
     
     
         18 . The non-transitory machine-readable medium of  claim 17 , wherein the operations further comprise obtaining a sample of the content stream, and wherein the type of content is identified based on the sample. 
     
     
         19 . The non-transitory machine-readable medium of  claim 17 , wherein the operations further comprise extracting metadata relating to the labeled object, wherein the metadata is delivered in-line with respect to the content stream. 
     
     
         20 . The non-transitory machine-readable medium of  claim 17 , wherein the content stream is produced by the communication network, and wherein the operations at least in part are performed as a service provided at an edge of the communication network.