Agency for Science, Technology and Research

Nbcuniversal Media, LLC

Sap Se

Comcast Cable Communications, LLC

Sujoy Roy

Sheau Bao Ng

Wei Yun Yau

Kong Wah Wan

Oliver Jojic

Amit Bagga

Sean Saito

Hassan Sayyadi-Harikandehei

Shannon McKenzie

Jeffrey J Reale

Yau Wei Yun

Mohammad Khalid Bin Ahmad

Jianshu Li

Xianjun Wang

Qing'An Li

Implementations include receiving an input image including a face, processing the input image through a global transformation network to provide a set of global transformation parameters, applying the set of global transformation parameters to the input image to provide a globally transformed image, processing the globally transformed image through a global representation learning network to provide a set of global features, processing the set of global features through a part localization network to provide a set of part localization parameters, applying the set of part localization parameters to the globally transformed image to provide a locally transformed image, processing the locally transformed image through a part representation learning network to provide a set of local features, and outputting a label representing at least one attribute depicted in the input image based on fusing global feature(s) from the set of global features, and local feature(s) from the set of local features.

Landmark-free face attribute prediction

Methods, systems, and computer-readable storage media for receiving a set of training images and a set of classification labels, generating a set of target codebooks based on the set of classification labels, the set of target codebooks being provided as a first set of vectors of random value and dimension, generating a set of output codebooks based on the set of training images, the set of output codebooks being provided as a second set of vectors of random value and dimension, training a ML model by minimizing a loss function provided as a mean-squared-error (MSE) loss function, the loss function being measured by the Euclidean distance between an output codebook of the set of output codebooks and a target codebook of the set of target codebooks.

Platform for preventing adversarial attacks on image-based machine learning models

Systems and methods for transforming a similarity matrix used to generate content item recommendation lists are described. The system may determine a weight (used to transform the similarity matrix) sufficient to place one or more content items on the content item recommendation list or at a particular position on the recommendation list. The transformed similarity matrix may further be reduced in size, and content recommendations may be generated using the transformed or reduced similarity matrix.

Recommendation system using a transformed similarity matrix

A content adaptation method includes: obtaining a primary metadata profile associated with a particular time point of primary content; obtaining secondary metadata profiles each associated with corresponding secondary content of a plurality of secondary content; identifying one of the plurality of secondary content associated with a secondary metadata profile having a desired similarity value with the primary metadata profile associated with the primary content; and matching the identified secondary content with the particular time point of the primary content.

Selecting adaptive secondary content based on a profile of primary content

A method and system for modeling the interests of an individual television viewer based on logging the user's television usage activities for the purpose of determining personalized recommendations is presented. According to one aspect of the claimed subject matter, user input in response to a questionnaire, the user's recorded television usage, the user's previous interactions with the personalized recommendations and contextual data such as external events and social group influences are used to model the probability of a user's interest in available media content. Data from the model is subsequently arranged to display units of available media content with the highest probabilities of user interest to the user. According to further embodiments, specific user behavior in previously consumed media content such as a user's selection, a user's continuous consumption, and a user's discontinuation of consumption is tracked and used as input.

Method and system for personalized recommendation modeling

Growing community of inventors

Singapore, Singapore

Sujoy Roy