Visual salience of online video as a predictor of success

Number: 9911202
Published: 2018-03-06
Filed: 2015-08-24
Assignee: Disney Enterprises, Inc.
Inventors: Frey; Seth et al.
CPC: G06V10/454; G06T7/60; G06V40/20; H04N19/154; G06V20/41; H04N19/17
Verdict: Set aside online-video visual-salience success prediction, content analytics
Source: Google Patents · FreePatentsOnline

Abstract

Systems, methods, and computer program products to perform an operation comprising computing a saliency value for a video based on saliency values a set of pixels in each frame of the video, computing, for the video, an expected value for a metric by a predictive algorithm based on the saliency value for the video, and outputting the expected value for the metric as an indication of an expected outcome for the metric achieved by the video.

Background

BACKGROUND(1) Field of the Invention(2) Embodiments disclosed herein relate to computer software. More specifically, embodiments disclosed herein relate to computer software that uses visual salience of a video as a predictor of success of the video.(3) Description of the Related Art(4) Producers of video content need analytics that can predict the success of a video before it has been released. Traditionally, predictions for feature films have been based on aspects surrounding the video, such as the video's genre, budget, the popularity of starring actors, critical reviews, and the like. Recently, social media content and other Internet sources have been leveraged to predict success. Many of these factors are subjective measures that potentially bias the results. For example, when relying on box office sales as a success measure, the results strongly depend on the chosen time window (such as first weeks, cumulative period in a theater, theater and video sales, etc.). Further still, available prediction models focus on full-length feature films, and are therefore of limited value when predicting the success of shorter videos such as commercials, trailers, and other content that is becoming prevalent on various streaming websites. However, computational measures of video assets themselves can serve as useful predictors. Specifically, computational models of human visual attention have not been applied to predict video success.SUMMARY(5) Embodiments disclosed herein provide sys

Claims

1. A method, comprising: computing a saliency value for a video based on saliency values for a set of pixels in each frame of the video; computing, for the video, an expected value for a metric by a predictive algorithm based on the saliency value for the video, wherein the metric comprises at least one of: (i) a number of views the video will receive, and (ii) a duration of the video a viewer will watch; and outputting the expected value for the metric as an indication of an expected outcome for the metric achieved by the video. 8. A computer program product, comprising: a non-transitory computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by a processor to perform an operation comprising: computing a saliency value for a video based on saliency values for a set of pixels in each frame of the video; computing, for the video, an expected value for a metric by a predictive algorithm based on the saliency value for the video, wherein the metric comprises at least one of: (i) a number of viewers that will watch a second video promoted by the video, and (ii) a likelihood that a viewer will click a link to an advertisement in the video; and outputting the expected value for the metric as an indication of an expected outcome for the metric achieved by the video. 15. A system, comprising: a processor; and a memory containing a program which when executed by the processor performs an operation comprising: computing a saliency value for a video based on saliency values for a set of pixels in each frame of the video; computing, for the video, an expected value for a metric by a predictive algorithm based on the saliency value for the video, wherein the metric comprises at least one of: (i) an expected amount of advertising revenue the video will generate, and (ii) a likelihood that a viewer will stop watching the video at each second of the video; and outputting the expected value for the metric as an indication of an expected outcome for the metric achieved by the video.