Abstract: The present disclosure is directed to encoding images. In particular, one or more computing devices can receive data representing one or more machine learning (ML) models configured, at least in part, to encode images comprising objects of a particular type. The computing device(s) can receive data representing an image comprising one or more objects of the particular type. The computing device(s) can generate, based at least in part on the data representing the image and the data representing the ML model(s), data representing an encoded version of the image that alters at least a portion of the image comprising the object(s) such that when the encoded version of the image is decoded, the object(s) are unrecognizable as being of the particular type by one or more object-recognition ML models based at least in part upon which the ML model(s) configured to encode the images were trained.

Abstract: Implementations described herein relate to generating candidate suggestion(s) that are personalized to a given user locally at an in-vehicle computing device of a vehicle of the given user. For example, implementations can identify an occurrence of a given suggestion state, generate the candidate suggestion(s) that are personalized to the user for the given suggestion state, store the candidate suggestion(s) in on-device storage of the in-vehicle computing device in association with the given suggestion state, and cause a given one of the candidate suggestion(s) to be provided for presentation to the given user. Accordingly, upon a subsequent occurrence of the given suggestion state, implementations can obtain the candidate suggestion(s) stored in association with the given suggestion state, and cause the given one of the candidate suggestion(s), or an additional one of the candidate suggestion(s), to be provided for presentation to the given user.

Abstract: Methods, apparatus, and computer readable media are described herein for allowing a first user to interface with an automated assistant to assign tasks to additional user(s), and/or for causing notification(s) of the assigned task to be rendered to the additional user(s) via corresponding automated assistant interface(s). In various implementations, one or more criteria can be utilized in selecting a group of client device(s), linked to the additional user, via which to provide the notification(s) for the task assigned to the additional user. Also, in various implementations condition(s) for providing the notification(s) for the task can be determined, and the notification(s) provided based on determining satisfaction of the condition(s).

Abstract: A method for providing availability of geographic data to enterprise clients. The method is implemented by processing hardware and includes generating a storage storing geographic information available to an enterprise client via an API call, where the enterprise client configured to (i) receive service requests from user devices and (ii) invoke the API to provide, in response to the service requests, information related to geography. When the enterprise client invokes the API to submit a query (304), the method includes: in a first instance, transmitting the query to a geographic service via a communication network (306) and generating a network-based response to the query using geographic information received from the geographic service in response to the query (312), and in a second instance, generating a storage-based response to the query using the geographic information stored in the storage (322).

Abstract: A method for evaluating synthesized speech, including receiving a speech sample in a first language; and determining a rating of the speech sample based on an encoding of the speech sample by an artificial intelligence encoding model, the rating of the speech sample corresponding to a naturalness of the speech sample, wherein the encoding of the speech sample is based on a first training stage of the encoding model using a first set of training data and a second training stage of the encoding model using a second set of training data, the first set of training data includes unlabeled speech audio, unlabeled text, and paired speech audio and text data in the first language and at least one additional language, and the second set of training data includes rated speech audio.

Abstract: A method including receiving, from a device, interaction data associated with an event including a first identifier, identifying a second identifier associated with the first identifier, identifying one or more previous interactions associated with the first identifier or the second identifier, annotating the interaction data based on the identified one or more previous interactions, wherein the annotation indicates a preference for previous interactions associated with the first identifier or previous interactions associated with the second identifier, and transmitting an indication that the event is associated with one or more of the previous interactions, wherein the one or more previous interactions are determined based on the preference indicated by the annotation.

Abstract: Systems and methods for identifying a personalized prior within a generative model's latent vector space based on a set of images of a given subject. In some examples, the present technology may further include using the personalized prior to confine the inputs of a generative model to a latent vector space associated with the given subject, such that when the model is tasked with editing an image of the subject (e.g., to perform inpainting to fill in masked areas, improve resolution, or deblur the image), the subject's identifying features will be reflected in the images the model produces.

Abstract: Using the techniques discussed herein, a set of images is captured by one or more array imagers (106). Each array imager includes multiple imagers configured in various manners. Each array imager captures multiple images of substantially a same scene at substantially a same time. The images captured by each array image are encoded by multiple processors (112, 114). Each processor can encode sets of images captured by a different array imager, or each processor can encode different sets of images captured by the same array imager. The encoding of the images is performed using various image-compression techniques so that the information that results from the encoding is smaller, in terms of storage size, than the uncompressed images.

Abstract: Spatial audio is rendered at a companion device or server connected to a wearable device, where the spatial audio is rendered based on a first pose estimate of the wearable device that is estimated at the companion device or server. The rendered spatial audio is then transmitted to the wearable device. The rendered spatial audio is refined at the wearable device based on a second pose estimate of the wearable device that is estimated at the wearable device. The refined spatial audio is then provided for playback via speakers of the wearable device.

Abstract: To provide ride services within a mapping application in a client computing device without directing the user to a separate ride service application, the mapping application invokes one or several ride service APIs to access ride service data from various ride service providers. For example, the mapping application receives a request for travel directions to a destination and generates multi-modal travel directions which include a route segment where the mode of transportation is a ride service. The mapping application invokes one or several ride service APIs to retrieve a price estimate, estimated wait time, or any other suitable information regarding the ride service route segment. Accordingly, the mapping application provides the multi-modal travel directions to a user including information regarding the ride service route segment.

Abstract: An HMD employs a multifocal module having one or more voltage controlled optical elements. By changing the voltage applied to the one or more optical elements, the HMD can change the focal length of an image generated by the micro-display. The multifocal module can include a cholesteric liquid crystal (CLC) element and a polarizer switch. By controlling the voltage applied to each of the CLC element and the polarizer switch, the HMD can change the focal length of the image passed through the multifocal module. The HMD can thereby change the focal length of any image generated by the micro-display.

Abstract: The present disclosure provides systems and methods that enable parameter-efficient transfer learning, multi-task learning, and/or other forms of model re-purposing such as model personalization or domain adaptation. In particular, as one example, a computing system can obtain a machine-learned model that has been previously trained on a first training dataset to perform a first task. The machine-learned model can include a first set of learnable parameters. The computing system can modify the machine-learned model to include a model patch, where the model patch includes a second set of learnable parameters. The computing system can train the machine-learned model on a second training dataset to perform a second task that is different from the first task, which may include learning new values for the second set of learnable parameters included in the model patch while keeping at least some (e.g., all) of the first set of parameters fixed.

Abstract: Implementations are provided for generating a plurality of simulated training instances based on a recorded user-directed robot control episode, and training one or more robot control policies based on such training instances. In various implementations, a three-dimensional environment may be simulated and may include a robot controlled by an external robot controller. A user may operate the robot controller to control the robot in the simulated 3D environment to perform one or more robotic tasks. The user-directed robot control episode, including responses of the external robot controller and the simulated robot to user commands and/or the virtual environment, can be captured. Features of the captured user-directed robot control episode can be altered in order to generate a plurality of training instances. One or more robot control policies can then be trained based on the plurality of training instances.

Abstract: Implementations relate to an automated assistant that provides and manages output from one or more elements of output hardware of a computing device. The automated assistant manages dynamic adjustment of access permissions to the computing device according to, for example, a detected presence of one or more users. An active-user queue can be established each time a unique user enters a viewing window of a camera of the computing device when, up to that point, no user was considered active. Multiple image frames can be captured via the camera and processed to determine whether an initial user remains in the viewing window and/or whether another user has entered the viewing window. The initial user can be considered active as long as they are exclusively detected in the viewing window. Restricted content associated with the user may be rendered by the computing device whilst the user is active.

Abstract: Implementations set forth herein relate to an automated assistant that can adapt to circumstances in which a user may invoke an automated assistant with an intention of interacting with the automated assistant via a non-default interface. For example, in some instances, a user may invoke an automated assistant by selecting a selectable GUI element. In response, the automated assistant can determine that, in the current context, spoken utterances may not be suitable for providing to the automated assistant. Based on this determination, the automated assistant can cause a keyboard interface to be rendered and/or initialized for receiving typed inputs from the user. Should the user subsequently change contexts, the automated assistant can determine that voice input is now suitable for user input and initialize an audio interface in response to the user providing an invocation input in the subsequent context.

Abstract: Display systems, such as near eye display systems or wearable heads up displays, may include a laser projection system having an optical engine and an optical scanner. Light output by the optical engine may be directed into the optical scanner as two angularly separated laser light beams. The angularly separated laser light beams typically have different angles of incidence on a second scan mirror of the optical scanner. Respectively different levels of magnification are applied to the beam diameter of each of the angularly separated laser light beams in a first dimension, such that the angularly separated laser light beams have respectively different beam diameters upon incidence at the second scan mirror. In some embodiments, the different beam diameters of the angularly separated laser light beams result in regions of incidence of each of the angularly separated laser light beams on the second scan mirror being equal or substantially similar.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting hotwords using a server. One of the methods includes receiving an audio signal encoding one or more utterances including a first utterance; determining whether at least a portion of the first utterance satisfies a first threshold of being at least a portion of a key phrase; in response to determining that at least the portion of the first utterance satisfies the first threshold of being at least a portion of a key phrase, sending the audio signal to a server system that determines whether the first utterance satisfies a second threshold of being the key phrase, the second threshold being more restrictive than the first threshold; and receiving tagged text data representing the one or more utterances encoded in the audio signal when the server system determines that the first utterance satisfies the second threshold.

Abstract: A computer-implemented method for k-anonymizing a dataset to provide privacy guarantees for all columns in the dataset can include obtaining, by a computing system including one or more computing devices, a dataset comprising data indicative of a plurality of entities and at least one data item respective to at least one of the plurality of entities. The computer-implemented method can include clustering, by the computing system, the plurality of entities into at least one entity cluster. The computer-implemented method can include determining, by the computing system, a majority condition for the at least one entity cluster, the majority condition indicating that the at least one data item is respective to at least a majority of the plurality of entities. The computer-implemented method can include assigning, by the computing system, the at least one data item to the plurality of entities in an anonymized dataset based at least in part on the majority condition.

Abstract: Generating edge-depth values for an object, utilizing the edge-depth values in generating a 3D point cloud for the object, and utilizing the generated 3D point cloud for generating a 3D bounding shape (e.g., 3D bounding box) for the object. Edge-depth values for an object are depth values that are determined from frame(s) of vision data (e.g., left/right images) that captures the object, and that are determined to correspond to an edge of the object (an edge from the perspective of frame(s) of vision data). Techniques that utilize edge-depth values for an object (exclusively, or in combination with other depth values for the object) in generating 3D bounding shapes can enable accurate 3D bounding shapes to be generated for partially or fully transparent objects. Such increased accuracy 3D bounding shapes directly improve performance of a robot that utilizes the 3D bounding shapes in performing various tasks.

Abstract: Implementations relate to techniques for providing context-dependent search results. The techniques can include receiving a query and background audio. The techniques can also include identifying the background audio, establishing concepts related to the background audio and obtaining terms related to the concepts related to the background audio. The techniques can also include obtaining search results based on the query and on at least one of the terms. The techniques can also include providing the search results.