Microsoft Technology Licensing, LLC

Beijing University of Technology

Microsft Technolgy Licensing, LLC

Yanfeng Sun

Arun Sacheti

Karim Hasham

Wei Wang

Rui Cai

Haifeng Li

Xiao Kong

Aaron Chun Win Yuen

Lei Zhang

Hong-Jiang Zhang

Mingjing Li

Parthasarathy Govindarajen

Fujiao Ju

Ming Gang Liu

Qifa Ke

Kun Wu

Yi Li

Rui Hu

David R Parlin

Ryuichi Hirano

Malik Mehdi Pradhan

Avinash Vemuluru

Yandong Guo

Hua Ding

Vincent Kin-Wah Leung

Gautam Vishwas Vaidya

Aaron Chun-Win Yuen

Soohoon Cho

Alexandre Michelis

Jeffrey Lee-Yu Su

Alexandre B R Michelis

Yogesh Ajit Vaidya

Alexandre Bernard Raymond Michelis

Nan Wu

Emily Lin

Deepak Santhanam

A processing method for image tensor data, which can greatly reduce the number of free parameters in the model, limit the weight layer flexibly, and can be applied to any order of image tensor data. In this processing method for image tensor data, TTRBM model of Restricted Boltzmann machine with tensor train format is introduced. The input and output data of this method are both represented by tensors, and the weight of the middle layer is also represented by tensors, and the restricted weight has the structure of tensor train. The number of free parameters in the middle layer is controlled by adjusting the rank of tensor train decomposition. The rank of TT decomposition is adjusted, and different feature representations with the same size are expressed.

Processing method for image tensor data

A processing method for high-order tensor data, which can avoid that the vectorization process of the image observation sample set damage the internal structure of the data, simplify the large amount of redundant information in the high-order tensor data in the image observation sample set, and improve the image processing speed. In this processing method for high-order tensor data, the high-order tensor data are divided into three parts: the shared subspace component, the personality subspace component and the noise part; the shared subspace component and the personality subspace component respectively represent the high-order tensor data as a group of linear combination of the tensor base and the vector coefficient; the variational EM method is used to solve the base tensor and the vector coefficient; design a classifier to classify the test samples by comparing the edge distribution of samples.

Processing method for high order tensor data

Methods and systems for providing targeted recommendations are provided. A user model is generated from user data. Feed candidates are generated based on the user model. The generated feed candidates are ranked on a predetermined scale. At least one targeted recommendation from the ranked feed candidates. The at least one targeted recommendation feed is provided to a user associated with the user data.

Personalized user-categorized recommendations

The technology described herein provides an efficient mechanism for quickly analyzing huge amounts of media content to find media content (hereafter 'content' or 'media content') that is relevant to a user. The technology analyzes features of a curator to classify curators by interest and/or find curators with similar content recommendations. The curator data can be used to make curator recommendations to users based on the user's interests. The technology described herein collects curator data from multiple content sites and analyzes the data to identify curators that recommend similar content on different content sites.

Recommending a content curator

Various embodiments described herein are directed to providing cohesive related searches with dynamically generated titles. One or more related queries are retrieved for a search query and are grouped into one or more groupings based on a predefined relationship between the search query and a subset of the related queries. A dynamically generated first title is retrieved for a first grouping from the predefined relationship between the search query and the related queries in first grouping. The first title and the related queries in the first grouping are presented. A selection pivot is provided to select a second grouping, where the selection pivot contains an indicator of the predefined relationship between the search query and the related queries in the second grouping. Upon selecting the second grouping, a dynamically generated second title is retrieved, and the second title and the related queries in the second grouping are presented.

Cohesive related searches with dynamically generated titles

The technology described herein provides an efficient mechanism for generating image tags. Image data from a plurality of sources may be analyzed to identify relevant text items from the aggregated data. The relevant text items may be keywords describing a subject of an image, an entity of an image, a location of an image, or the like. From the aggregated image data, one or more image tags may be generated and stored as an offline dataset with an image identifier. Upon detecting a prompt such as a user issuing a search query for an image, the image identifier is used to perform a look up of the image and associated image tags to be provided.

Generating image tags

A user query may be evaluated to provide a result set. In case the results do not reflect the user's intent, the device may provide recourse options for adjusting the query in a manner that yields more desirable results, e.g., a suggestion at the top of the result set for a different spelling, or recommendations at the end of the results set for additional query techniques that may yield more accurate results. However, such presentation of recourse options may clutter the user interface and/or go unnoticed by the user. Instead, an adjusted query may be identified with an interpreted probability of reflecting the intent of the query. An adjustment option describing the adjusted query may be inserted into the result set, between a higher-probability first result and a lower-probability second result. Selection of the adjustment option may cause the adjusted query to be evaluated on behalf of the user.

Providing query recourse with embedded query adjustment options

Anchor images and information associated therewith are accumulated during a Web crawling operation. One or more rules are applied to the accumulated candidate anchor images to filter out candidate anchor images that are not appropriate for use as the anchor image for a particular target video. The remaining candidate anchor image is then selected as the anchor image for the particular video.

Anchor image identification for vertical video search

Video entity templates defining common features that relate to various metadata types shared among a group of video Web pages are generated for target Web sites. Metadata associated with videos contained within Web pages belonging to a particular target Web site can then be automatically and accurately extracted using a video entity template generated for the particular target Web site. This metadata can then be indexed for use by video search applications in providing video search results.

Scalable metadata extraction for video search

Methods and computer-readable media for propagating content category information to images stored in a database are described. A seed image that is associated with a known content category is received. A content-based image retrieval is conducted using the seed image as a search query image. A number of search result images are identified. The content category is propagated to the search result images. Metadata associated with the search result images is aggregated and analyzed to identify domains that should also be associated with the content category. Additional images that are associated with the domain are identified and the content category propagated thereto. The process is iterated using the additional images as search query images for the content-based image retrieval.

Image metadata propagation

A process for comparing two digital images is described. The process includes comparing texture moment data for the two images to provide a similarity index, combining the similarity index with other data to provide a similarity value and determining that the two images match when the similarity value exceeds a first threshold value.

Integrated solution to digital image similarity searching

In one aspect, the present disclosure describes a process for automatic artifact compensation in a digital representation of an image. The process includes detecting, by a processor, regions corresponding to facial images within the digital representation; locating, by the processor, red-eye regions within the detected regions; and automatically modifying, by the processor, the located red-eye regions to provide a modified image.

Modification of red-eye-effect in digital image

Growing community of inventors

Redmond, WA, United States of America

Yanfeng Sun