Video classification and preview selection

Abstract

Systems and methods select and provide video snippets in a matrix interface. Example methods include obtaining a portion of a video stream, assigning the portion to a class, determining that the assigned class is a promoted class, and generating a snippet for the video stream using the portion. Other example methods include determining, using a trained video classifier, a set of video streams that have at least one portion that is classified as a promoted class, calculating, for each of the video streams in the set, an aggregate score for the video stream, selecting video streams with highest aggregate scores, generating a snippet for each of the selected video streams, and providing the snippets in a user interface. Methods may also include selecting portions of video streams that are responsive to a user-provided parameter and generating a snippet for the video stream a portion classified as a promoted class.

What is claimed is: 
     
         1 . A method comprising:
 obtaining portions of a video stream;   determining, from the portions, a subset of portions that depict content determined to be in a promoted class, wherein a promoted class has an associated percentage that represents subject matter of rare occurrence within a statistically relevant sample of video streams;   generating a snippet for the video stream using at least one portion in the subset; and   providing the snippet for display in a user interface, the snippet being selectable and the user interface being configured to, responsive to selection of the snippet, provide the video stream for viewing.   
     
     
         2 . The method of  claim 1 , wherein generating the snippet includes:
 obtaining a preceding portion and a succeeding portion of the at least one portion; and   providing the preceding portion, the at least one portion, and the succeeding portion as the snippet, so the at least one portion represents a middle of the snippet.   
     
     
         3 . The method of  claim 1 , wherein the subset includes a second portion, the at least one portion being a first portion, and generating the snippet includes stitching together the first portion and the second portion as the snippet. 
     
     
         4 . The method of  claim 3 , wherein the first portion and the second portion are not consecutive portions. 
     
     
         5 . The method of  claim 4 , wherein generating the snippet includes:
 obtaining a first preceding portion and a first succeeding portion of the first portion;   obtaining a second preceding portion and a second succeeding portion of the second portion; and   stitching the first portion and the second portion by stitching the first preceding portion, the first portion, the first succeeding portion, the second preceding portion, the second portion, and the second succeeding portion as the snippet.   
     
     
         6 . The method of  claim 1 , wherein determining the subset includes:
 determining, using a trained video classifier, a respective class for each of at least some portions of the video stream; and   selecting portions for the subset that have a respective class classified as a promoted class, wherein the promoted class has an associated percentage that meets a threshold.   
     
     
         7 . The method of  claim 1 , wherein the video stream is one video stream in a plurality of video streams and the method further includes:
 generating a respective snippet for each of the plurality of video streams from respective subsets of preview-eligible portions,   wherein the user interface displays the respective snippets in a matrix and selection of a respective snippet in the matrix initiates display of the video stream associated with the selected respective snippet.   
     
     
         8 . The method of  claim 1 , wherein the video stream is one video stream in a plurality of video streams and the method further includes:
 determining a respective subset of portions, for each video stream in the plurality of video streams, that are preview-eligible based on the class of subject matter depicted in the portion, wherein each class depicted in a portion has a respective confidence score for the portion; and   generating a respective snippet for each of the plurality of video streams from respective subsets of preview-eligible portions, the respective snippet for a video stream of the plurality of video streams being based on the confidence scores of the subset of portions for the video stream.   
     
     
         9 . The method of  claim 8 , further comprising:
 ranking the plurality of video streams based on an aggregation of the respective confidence scores selecting the at least one portion from among the subset of portions responsive to determining that the at least one portion has a highest confidence score from among the subset of portions.   
     
     
         10 . The method of  claim 1 , wherein the at least one portion has at least two class assignments, each class assignment having a respective confidence score, and the method further includes:
 calculating an aggregate confidence score for each portion in the subset of portions; and   selecting the at least one portion based on the at least one portion having a highest aggregate confidence score.   
     
     
         11 . A computing device comprising:
 at least one processor; and   a memory storing instructions that, when executed by the at least one processor, are configured to:   obtain portions of a video stream,   determine, from the portions, a subset of portions that depict content determined to be in a promoted class, wherein a promoted class has an associated percentage that represents subject matter of rare occurrence within a statistically relevant sample of video streams,   generate a snippet for the video stream using at least one portion in the subset, and   provide the snippet for display in a user interface, the snippet being selectable and the user interface being configured to, responsive to selection of the snippet, provide the video stream for viewing.   
     
     
         12 . The computing device of  claim 11 , wherein generating the snippet includes:
 obtaining a preceding portion and a succeeding portion of the at least one portion; and   providing the preceding portion, the at least one portion, and the succeeding portion as the snippet, so the at least one portion represents a middle of the snippet.   
     
     
         13 . The computing device of  claim 11 , wherein the subset includes a second portion, the at least one portion being a first portion, and generating the snippet includes stitching together the first portion and the second portion as the snippet. 
     
     
         14 . The computing device of  claim 13 , wherein the first portion and the second portion are not consecutive portions. 
     
     
         15 . The computing device of  claim 14 , wherein generating the snippet includes:
 obtain a first preceding portion and a first succeeding portion of the first portion;   obtain a second preceding portion and a second succeeding portion of the second portion; and   stitch the first portion and the second portion by stitching the first preceding portion, the first portion, the first succeeding portion, the second preceding portion, the second portion, and the second succeeding portion as the snippet.   
     
     
         16 . The computing device of  claim 11 , wherein determining the subset includes:
 determine, using a trained video classifier, a respective class for each of at least some portions of the video stream; and   select portions for the subset that have a respective class classified as a promoted class.   
     
     
         17 . The computing device of  claim 11 , wherein the video stream is one video stream in a plurality of video streams and the memory stores instructions that are further configured to:
 generate a respective snippet for each of the plurality of video streams from respective subsets of preview-eligible portions,   wherein the user interface displays the respective snippets in a matrix and selection of a respective snippet in the matrix initiates display of the video stream associated with the selected respective snippet.   
     
     
         18 . The computing device of  claim 11 , wherein the video stream is one video stream in a plurality of video streams and the memory stores instructions that are further configured to:
 determine a respective subset of portions, for each video stream in the plurality of video streams, that are preview-eligible based on the class of subject matter depicted in the portion, wherein each class depicted in a portion has a respective confidence score for the portion; and   generate a respective snippet for each of the plurality of video streams from respective subsets of preview-eligible portions, the respective snippet for a video stream of the plurality of video streams being based on the confidence scores of the subset of portions for the video stream.   
     
     
         19 . The computing device of  claim 18 , further comprising:
 rank the plurality of video streams based on an aggregation of the respective confidence scores selecting the at least one portion from among the subset of portions responsive to determining that the at least one portion has a highest confidence score from among the subset of portions.   
     
     
         20 . The computing device of  claim 11 , wherein the at least one portion has at least two class assignments, each class assignment having a respective confidence score, and the memory stores instructions that are further configured to:
 calculate an aggregate confidence score for each portion in the subset of portions; and   select the at least one portion based on the at least one portion having a highest aggregate confidence score.