Abstract: Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to receive, at a client device, a machine learning model from a server, detect a usage pattern for a content item, store an association between the content item and the detected usage pattern in local data, train the machine learning model using local data for the content item with the detected usage pattern to generate a trained machine learning model, generate an update for the machine learning model, privatize the update for the machine learning model, and transmit the privatized update for the machine learning model to the server.

Abstract: Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to select a set of content items from a content item collection based upon a temporal relevance and a contextual relevance to a period of time, rank the set of content items based on at least one of a content item category or a content item predefined relevance score, partition the period of time into a set of time slots to schedule for rendering content in an application, rank the set of time slots based on device usage analysis for the period of time, and schedule the set of content items into the set of time slots in accordance with the rankings.

Abstract: Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to receive, at a client device, a machine learning model from a server, detect a usage pattern for a content item, store an association between the content item and the detected usage pattern in local data, train the machine learning model using local data for the content item with the detected usage pattern to generate a trained machine learning model, generate an update for the machine learning model, privatize the update for the machine learning model, and transmit the privatized update for the machine learning model to the server.

Abstract: Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to select a set of content items from a content item collection based upon a temporal relevance and a contextual relevance to a period of time, rank the set of content items based on at least one of a content item category or a content item predefined relevance score, partition the period of time into a set of time slots to schedule for rendering content in an application, rank the set of time slots based on device usage analysis for the period of time, and schedule the set of content items into the set of time slots in accordance with the rankings.

Abstract: Exemplary embodiments for performing server-assisted object recognition and tracking are disclosed herein. For example, in certain embodiments of the disclosed technology, one or more objects are efficiently recognized and tracked on a mobile device by using a remote server that can provide high capacity computing and storage resources. With the benefit of high-speed image processing on a remote server and high-bandwidth communication networks connecting the mobile device and the remote server, it is possible to identify an object and to track changes in the object's characteristics or location, so that a user experiences seamless, real-time tracking.

Abstract: Exemplary embodiments for performing server-assisted object recognition and tracking are disclosed herein. For example, in certain embodiments of the disclosed technology, one or more objects are efficiently recognized and tracked on a mobile device by using a remote server that can provide high capacity computing and storage resources. With the benefit of high-speed image processing on a remote server and high-bandwidth communication networks connecting the mobile device and the remote server, it is possible to identify an object and to track changes in the object's characteristics or location, so that a user experiences seamless, real-time tracking.

Abstract: Techniques and tools are described for performing perspective correction using a reflection. Reflective properties of a surface being photographed can be used to determine a rotation of the device taking the photograph relative to the surface. Light sourced or produced by the device can be used to create a reflection spot in the picture. A position of the reflection spot within the picture is calculated and used to determine the rotation. The rotation can be used for performing perspective correction on the picture, or on another picture taken by the device.

Abstract: Techniques and tools are described for performing perspective correction using a reflection. Reflective properties of a surface being photographed can be used to determine a rotation of the device taking the photograph relative to the surface. Light sourced or produced by the device can be used to create a reflection spot in the picture. A position of the reflection spot within the picture is calculated and used to determine the rotation. The rotation can be used for performing perspective correction on the picture, or on another picture taken by the device.

Abstract: Techniques and tools are described for performing perspective correction using a reflection. Reflective properties of a surface being photographed can be used to determine a rotation of the device taking the photograph relative to the surface. Light sourced or produced by the device can be used to create a reflection spot in the picture. A position of the reflection spot within the picture is calculated and used to determine the rotation. The rotation can be used for performing perspective correction on the picture, or on another picture taken by the device.

Abstract: Techniques and tools are described for performing perspective correction using a reflection. Reflective properties of a surface being photographed can be used to determine a rotation of the device taking the photograph relative to the surface. Light sourced or produced by the device can be used to create a reflection spot in the picture. A position of the reflection spot within the picture is calculated and used to determine the rotation. The rotation can be used for performing perspective correction on the picture, or on another picture taken by the device.

Abstract: Exemplary embodiments for performing server-assisted object recognition and tracking are disclosed herein. For example, in certain embodiments of the disclosed technology, one or more objects are efficiently recognized and tracked on a mobile device by using a remote server that can provide high capacity computing and storage resources. With the benefit of high-speed image processing on a remote server and high-bandwidth communication networks connecting the mobile device and the remote server, it is possible to identify an object and to track changes in the object's characteristics or location, so that a user experiences seamless, real-time tracking.

Abstract: Because of the small size and mobility of smart phones, and because they are typically hand-held, it is both natural and feasible to use hand, wrist, or arm gestures to communicate commands to the electronic device as if the device were an extension of the user's hand. Some user gestures are detectable by electro-mechanical motion sensors within the circuitry of the smart phone. The sensors can sense a user gesture by detecting a physical change associated with the device, such as motion of the device or a change in orientation. In response, a voice-based or image-based input mode can be triggered based on the gesture. Methods and devices disclosed provide a way to select from among different input modes to a device feature, such as a search, without reliance on manual selection.