Disney Enterprises, Inc.

Pablo Beltran Sanchidrian

Anthony M Accardo

Miquel Angel Farre Guiu

Marc Junyent Martin

Avner Swerdlow

Katharine S Ettinger

Matthew C Petrillo

Monica Alfaro Vendrell

Evan A Binder

Aljoscha Smolic

Aljosa Aleksej Andrej Smolic

Miquel Angel Farré Guiu

Michael Clements

Seth Frey

Monica Alfaro

Miquel Á Farré Guiu

A media content tagging system includes a computing platform having a hardware processor, and a system memory storing a tag selector software code configured to receive media content having segments, each segment including multiple content elements each associated with metadata tags having respective pre-computed confidence scores. For each content element, the tag selector software code assigns each of the metadata tags to at least one tag group, determines a confidence score for each tag group based on the pre-computed confidence scores of its assigned metadata tags, discards tag groups having less than a minimum number of assigned metadata tags, and filters the reduced number of tag groups based on the second confidence score to identify a further reduced number of tag groups. The tag selector software code then selects at least one representative tag group for a segment from among the further reduced number of tag groups.

Automation of media content tag selection

Systems, methods, and articles of manufacture to perform an operation comprising processing, by a machine learning (ML) algorithm and a ML model, a plurality of images in a first dataset, wherein the ML model was generated based on a plurality of images in a training dataset, receiving user input reviewing a respective set of tags applied to each image in the first data set as a result of the processing, identifying, based on a first confusion matrix generated based on the user input and the sets of tags applied to the images in the first data set, a first labeling error in the training dataset, determining a type of the first labeling error based on a second confusion matrix, and modifying the training dataset based on the determined type of the first labeling error.

Semiautomatic machine learning model improvement and benchmarking

According to one implementation, a content classification system includes a computing platform having a hardware processor and a system memory storing a video asset classification software code. The hardware processor executes the video asset classification software code to receive video clips depicting video assets and each including images and annotation metadata, and to preliminarily classify the images with one or more of the video assets to produce image clusters. The hardware processor further executes the video asset classification software code to identify key features data corresponding respectively to each image cluster, to segregate the image clusters into image super-clusters based on the key feature data, and to uniquely identify each of at least some of the image super-clusters with one of the video assets.

Video asset classification

There is provided a method that includes receiving a video having video shots, and creating video shot groups based on similarities between the video shots, where each video shot group of the video shot groups includes one or more of the video shots and has different ones of the video shots than other video shot groups. The method further includes creating at least one video supergroup including at least one video shot group of the video shot groups based on interactions among the one or more of the video shots in each of the video shot groups, and divide the at least one video supergroup into connected video supergroups, each connected video supergroup of the connected video supergroups including one or more of the video shot groups based on the interactions among the one or more of video shots in each of the video shot groups.

Systems and methods for contextual video shot aggregation

Systems, methods, and computer program products to perform an operation comprising receiving a plurality of superclusters that includes at least one of a plurality of shot clusters, wherein each of the shot clusters includes at least one of a plurality of video shots, and wherein each video shot includes one or more video frames, and computing an expected value for a metric based on the plurality of superclusters.

Shot structure of online video as a predictor of success

Growing community of inventors

Barcelona, Spain

Pablo Beltran Sanchidrian