Microsoft Technology Licensing, LLC

Micro Focus Lcc

Entit Software LLC

Longsand Limited

Pashmina Cameron

Sean Blanchflower

Richard John Black

Antony Ian Taylor Rowstron

Thomas Joseph Cashman

George Saklatvala

Timothy John Deegan

Ioan Alexandru Stefanovici

David Lara Saucedo

Ariel Gomez Diaz

James Hilton Clegg

Peter Gyula Scholtz

Govert Michael Verkes

David Collier

Andrew William Fitzgibbon

Erroll William Wood

Federica Bogo

Cheng Zhang

Sebastian Tschiatschek

Yingzhen Li

Camilla Alice Longden

Richard Eric Turner

Timothy Woods

Sasa Galic

Mohammad Sadegh Ali Akbarian

Angus Lamb

Evgeny Sergeevich Saveliev

Jose Miguel Hernández Lobato

A method for predicting the pose of an articulated object includes receiving spatial information for n joints of the articulated object. The spatial information for the n joints is passed to a machine learning model previously trained to receive spatial information for n+m joints as input, wherein m>=1. From the machine learning model, a pose prediction for the articulated object is received as output based at least on the spatial information for the n joints, and without spatial information for the m joints.

Pose prediction for articulated object

A computer-implemented method of training an auxiliary machine learning model to predict a set of new parameters of a primary machine learning model, wherein the primary model is configured to transform from an observed subset of a set of real-world features to a predicted version of the set of real-world features.

Auxiliary model for predicting new model parameters

A method for virtually representing human body poses includes receiving positioning data detailing parameters of one or more body parts of a human user based at least in part on input from one or more sensors. One or more mapping constraints are maintained that relate a model articulated representation to a target articulated representation. A model pose of the model articulated representation and a target pose of the target articulated representation are concurrently estimated based at least in part on the positioning data and the one or more mapping constraints. The previously-trained pose optimization machine is trained with training positioning data having ground truth labels for the model articulated representation. The target articulated representation is output for display with the target pose as a virtual representation of the human user.

Concurrent human pose estimates for virtual representation

One example provides a system for reading birefringent data. The system comprises one or more light sources, a first polarization state generator positioned to generate first polarized light from light of a first wavelength band output by the one or more light sources, a second polarization state generator positioned to generate second polarized light from light of a second wavelength band output by the one or light sources, an image sensor configured to acquire an image of the sample region via the first polarized light and the second polarized light, a polarization state analyzer disposed between the sample region and the image sensor, a first bandpass filter configured to pass light of the first wavelength band onto the image sensor, and a second bandpass filter configured to pass light of the second wavelength band onto the image sensor.

Efficient reading of birefringent data

According to examples, an apparatus may identify a first viewpoint that an entity expressed in a first media file, identify a second viewpoint expressed in a second media file that is attributed to the entity, determine that the second viewpoint is dissimilar to the first viewpoint and in response to the determination that the second viewpoint is dissimilar to the first viewpoint, may output a message.

Entity viewpoint determinations

In some examples, a camera tracking failure in a simultaneous localization and mapping (SLAM) process with respect to a first SLAM map may be identified. Responsive to identification of the camera tracking failure, a second SLAM map for the SLAM process may be initialized. A video frame tracked through the second SLAM map may be accessed. Matched features between the video frame and individual keyframes of the first SLAM map may be identified to determine a keyframe subset. The keyframe subset may be analyzed to determine a candidate camera position from among the keyframe subset. The candidate camera position may be tracked with respect to the first SLAM map for subsequent video frames. The first SLAM map may be stitched to the second SLAM map responsive to tracking the candidate camera position with respect to the first SLAM map for the subsequent.

Stitching maps generated using simultaneous localization and mapping

A plurality of binary support vector machines (SVM) are trained to vote between two of a plurality of classes. Each of the binary SVMs vote between a different pair of the plurality of classes. A set of validated samples is input to each of the binary SVMs. The set of validated samples includes samples of each of the classes. Likelihood estimates are determined based on votes of the binary SVMs. A prior probability distribution (prior) is assigned for each of the classes. A posterior probability is calculated for each of the classes based on the determined likelihood estimates and the assigned priors.

Calculating posterior probability of classes

According to examples, an index of entries may be retrieved, in which each of the entries may correspond to phrases that are analogous to other phrases in content on a network. A plurality of phrases that are similar to the entries in the index may be identified in a content accessed on the apparatus and a determination may be made, based on a user interaction with the accessed content via a user interface, that the user has read a first phrase in the plurality of phrases. The index may be updated to indicate that the user has read the first phrase and, based on the updated index, a second phrase in a subsequently accessed content that is similar to the first phrase may be determined. In addition, the second phrase in the subsequently accessed content may be displayed differently from other displayed phrases in the subsequently accessed content.

Distinguish phrases in displayed content

In some examples, a system includes a color cluster learning engine and a color recognition engine. The color cluster learning engine may be configured to obtain a set of training images, process the training images to obtain clusters of pixel colors for the training images, identify learned color clusters from the clusters of pixel colors obtained from the training images, and label the learned color clusters with color indicators. The color recognition engine may be configured to receive an input image for color identification, process the input image to obtain a particular cluster of pixel colors that covers the highest number of pixels in the input image, match the particular cluster to a particular learned color cluster labeled with a particular color indicator, and identify a color of the input image as specified by the particular color indicator.

Color recognition through learned color clusters

A pair of video frames may be used for simultaneous localization and mapping (SLAM) initialization. The pair of frames may be determined according to a translation threshold. Whether the translation threshold is met may be determined by obtaining the pair of video frames and estimating the translation between the frames.

Simultaneous localization and mapping initialization

Examples disclosed herein relate to detecting change in a 3-dimensional (3D) view. The examples enable determining a first set of sparse local features from a first set of frames of an initial 3D scene; determining whether the initial 3D scene is different from a current 3D scene based on the first set of sparse local features; and providing information about any determined differences between the initial 3D scene and the current 3D scene.

Growing community of inventors

Cambridge, United Kingdom

Pashmina Cameron