Abstract: According to an aspect, a method may include receiving two-dimensional (2D) positions of at least one of a first reflective marker or a second reflective marker of a physical component, estimating a three-dimensional (3D) position of the first reflective marker and a 3D position of the second reflective marker based on the 2D positions, and computing an orientation of the physical component in 3D space based on the 3D position of the first reflective marker, the 3D position of the second reflective marker, and positioning information of the first and second reflective markers in the physical component.

Abstract: An augmented reality (AR) device, such as AR glasses, may include a microphone array. The sensitivity of the microphone array can be directed to a target by beamforming, which includes combining the audio of each microphone of the array in a particular way based on a location of the target. The present disclosure describes systems and methods to determine the location of the target based on a gaze of a user and beamform the audio accordingly. This eye-tracked beamforming (i.e., foveated beamforming) can be used by AR applications to enhance sounds from a gaze direction and to suppress sounds from other directions. Additionally, the gaze information can be used to help visualize the results of an AR application, such as speech-to-text.

Abstract: A method can include selecting, from at least a first avatar and a second avatar based on at least one attribute of a calendar event associated with a user, a session avatar, the first avatar being based on a first set of images of a user wearing a first outfit and the second avatar being based on a second set of images of the user wearing a second outfit, and presenting the session avatar during a videoconference, the presentation of the session avatar changing based on audio input received from the user during the videoconference.

Abstract: Neural network-based categorization can be improved by incorporating graph neural networks that operate on a graph representing the taxonomy of the categories into which a given input is to be categorized by the neural network based-categorization. The output of a graph neural network, operating on a graph representing the taxonomy of categories, can be combined with the output of a neural network operating upon the input to be categorized, such as through an interaction of multidimensional output data, such as a dot product of output vectors. In such a manner, information conveying the explicit relationships between categories, as defined by the taxonomy, can be incorporated into the categorization. To recapture information, incorporate new information, or reemphasize information a second neural network can also operate upon the input to be categorized, with the output of such a second neural network being merged with the output of the interaction.

Abstract: Systems and methods for training models to predict dense correspondences across images such as human images. A model may be trained using synthetic training data created from one or more 3D computer models of a subject. In addition, one or more geodesic distances derived from the surfaces of one or more of the 3D models may be used to generate one or more loss values, which may in turn be used in modifying the model's parameters during training.

Abstract: A system including a computer vision model configured to perform a machine learning task is described. The computer vision model includes multiple wrapped convolutional layers, in which each wrapped convolutional layer includes a respective convolutional layer configured to receive, for each time step of multiple time steps, a layer input and to process the layer input to generate an initial output for the current time step, and a respective note-taking module configured to receive the initial output and to process the initial output to generate a feature vector for the current time step, the feature vector representing local information of the wrapped convolutional layer. The model includes a summarization module configured to receive the feature vectors and to process the feature vectors to generate a revision vector for the current time step, the revision vector representing global information of the plurality of wrapped convolutional layers.

Abstract: A method includes providing an interactive graphical user interface comprising a first menu providing one or more input options, a second menu providing one or more machine learning models, and a third menu providing one or more output formats. The method also includes generating a graph in a portion of the interactive graphical user interface by detecting one or more user selections of an input option, a machine learning model, and an output format, displaying nodes corresponding to the input option, the machine learning model, the output format, and displaying edges connecting the first node to the second node, and the second node to the third node. The method additionally includes applying the machine learning model to an input associated with the input option to generate an output in the output format. The method further includes providing, by the interactive graphical user interface, the output in the output format.

Abstract: Systems and methods for training models to predict dense correspondences across images such as human images. A model may be trained using synthetic training data created from one or more 3D computer models of a subject. In addition, one or more geodesic distances derived from the surfaces of one or more of the 3D models may be used to generate one or more loss values, which may in turn be used in modifying the model's parameters during training.

Abstract: Systems and methods are provided in which physical objects in the ambient environment can function as user interface implements in an augmented reality environment. A physical object detected within a field of view of a camera of a computing device may be designated as a user interface implement in response to a user command. User interfaces may be attached to the designated physical object, to provide a tangible user interface implement for user interaction with the augmented reality environment.

Abstract: A method including determining a gaze direction of a user of a wearable device, capturing an image using a forward-looking camera of the wearable device, detecting a surroundings of the user based on the image, determining whether or not the user is distracted based on the gaze direction and the surroundings, and in response to determining the user is distracted, causing an operation to be performed on the wearable device, the operation configured to cause the user to change the user's attention.

Abstract: A method performed by a computing system comprises generating text from audio data and determining an end portion of the text to include in a summarization of the text based on a length of a portion of the audio data from which the text was generated and which ends with a proposed end portion and a time value associated with the proposed end portion, the proposed end portion including a word from the text.

Abstract: A head-mounted device, such as augmented reality glasses, can be configured to track screen time. The head-mounted device can advantageously track screen time across a variety of devices whether or not they are configured to such measurements. Various modes for tracking the screen time may be used based on accuracy, power consumption, and privacy. The head-mounted device may track in a mode based on a battery level, a device setting or a user input. The head-mounted device may accumulate screen times in a database and in some implementations, can link screen times with devices in the database based on device information gathered by the head-mounted device during the tracking process. The head-mounted device may be configured to generate and display an alert based on one or more rules about screen time.

Abstract: Systems and methods are provided in which physical objects in the ambient environment can function as user interface implements in an augmented reality environment. A physical object detected within a field of view of a camera of a computing device may be designated as a user interface implement in response to a user command. User interfaces may be attached to the designated physical object, to provide a tangible user interface implement for user interaction with the augmented reality environment.

Abstract: According to a general aspect, a method can include receiving a photo of a virtual conference participant, and a depth map based on the photo, and generating a plurality of synthesized images based on the photo. The plurality of synthesized images can have respective simulated gaze directions of the virtual conference participant. The method can also include receiving, during a virtual conference, an indication of a current gaze direction of the virtual conference participant. The method can further include animating, in a display of the virtual conference, an avatar corresponding with the virtual conference participant. The avatar can be based on the photo. Animating the avatar can be based on the photo, the depth map and at least one synthesized image of the plurality of synthesized images, the at least one synthesized image corresponding with the current gaze direction.

Abstract: Techniques of introducing virtual objects into a physical environment of AR system include displacing vertices of a mesh representing the physical environment based on a live depth map. For example, an AR system generates a mesh template, i.e., an initial mesh with vertices that represents a physical environment and a depth map that indicates a geometry of real objects within the physical environment. The AR system is configured to represent the real objects in the physical environment by displacing the vertices of the mesh based on depth values of the depth map and parameter values of a pinhole camera model. The depth values may be taken from the perspective of an illumination source in the physical environment.

Abstract: A system and method for relighting of an augmented reality scene are provided. Image data and depth data corresponding to a physical environment may be captured by an image sensor of a computing device, for display in an AR scene generated by the computing device. The system may use the image data and depth data to generate a live depth map of the physical environment. A virtual light source may be placed in the AR scene of the physical environment in response to a user input. The system may relight the AR scene in response to the placement of the virtual light source based on the image data and the depth map.

Abstract: Smart devices can be configured to collect and share various forms of context data about where a user is located (e.g., location), what a user will be doing (e.g., schedule), and what a user is currently doing (e.g., activity). This context data may be combined with fingerprint data (e.g., biometrics) to help identify the fingerprint data. For example, a location of a user may help associated speech detected at that location with the user. These associations may be stored in a mapping database that can be updated over time to reduce ambiguities in identification. The mappings in the database may be used to train a machine learning model to recognize fingerprints as identities, which may be useful in applications, such as speaker identification.

Abstract: The disclosed systems and method correlates user behaviors with audio processing to achieve more accurate conclusions about sounds in a user's environment. These conclusions may, in turn, be used to adjust the way a device, such as AR glasses, operate or respond to the sounds. For example, audio events determined from processing speech can be correlated with behavior events determined by sensing a user to improve a speech-to-text transcript of the speech by separating, or otherwise altering, the text in the transcript by speaker.

Abstract: Augmented reality devices can be configured to display messages in response to sounds from an environment. A variety of techniques can be combined to localize and track the sources of the sounds in the environment. Messages created in response to the sounds can then be anchored to their corresponding sources in order to provide a user with a clear understanding of the location of sources of the messages. Additionally, these anchored messages can be enhanced with additional information, such as identification, to further the user’s understanding of the sources of the messages. The anchored messages can track relative movement to integrate with the AR environment.

Abstract: Systems and methods are related to tracking an attention of a user with respect to content presented on a virtual screen, detecting a defocus event associated with a first region of the content, and determining a next focus event associated with a second region of the content. The determination can be based at least in part on the defocus event and on the tracked attention of the user. The systems and methods can include generating, based on the determined next focus event, a marker for differentiating the second region of the content from a remainder of the content, and in response to detecting a refocus event associated with the virtual screen, triggering execution of the marker associated with the second region of the content.