Amazon Technologies, Inc.

Qualcomm Incorporated

Google Inc.

Han Zhang

Jianchun Dong

Xuan Zhong

Gengsheng Zhang

Ning Luo

Bo Zheng

Jung Sik Yang

Yinghua Yang

Zhen Xu

Niranjan Madan Mohan Bhatia

Min Wang

Eliot Kim

Chia-Jean Wang

Mangesh M Chansarkar

Sercan Omer Arik

Bozhao Tan

Chun Sik Jeong

Geng Ye

Tomas Jon Pfister

David Tuck

Honglak Lee

Oliver Huy Doan

Nikhil Gautam

Zizhao Zhang

Yinfei Yang

Jason Michael Baldridge

Yuna Hu

Jeffrey John Easter

Chung-Ling Hsu

Jing Yu Koh

Long Zhao

Yana Potashnik

Alfred Ying-Fai Lui

A method includes receiving image data including a series of image patches of an image. The method includes generating, using a first set of transformers of a vision transformer (V-T) model, a first set of higher order feature representations based on the series of image patches and aggregating the first set of higher order feature representations into a second set of higher order feature representations that is smaller than the first set. The method includes generating, using a second set of transformers of the V-T model, a third set of higher order feature representations based on the second set of higher order feature representations and aggregating the third set of higher order feature representations into a fourth set of higher order feature representations that is smaller than the third set. The method includes generating, using the V-T model, an image classification of the image based on the fourth set.

Aggregating nested vision transformers

A computer-implemented method includes receiving, by a computing device, a particular textual description of a scene. The method also includes applying a neural network for text-to-image generation to generate an output image rendition of the scene, the neural network having been trained to cause two image renditions associated with a same textual description to attract each other and two image renditions associated with different textual descriptions to repel each other based on mutual information between a plurality of corresponding pairs, wherein the plurality of corresponding pairs comprise an image-to-image pair and a text-to-image pair. The method further includes predicting the output image rendition of the scene.

Cross-modal contrastive learning for text-to-image generation based on machine learning models

In an aspect, a user equipment (UE) receives a spoofing alert message from either a server or an internet-of-things (IOT) device that indicates whether a spoofed Global Navigation Satellite System (GNSS) condition is present. Based on determining that the spoofing alert message indicates that a spoofed GNSS condition is present, the UE determines, based on the spoofing alert message, a location of a spoofer broadcasting a spoofed GNSS signal, determines, based on the location of the spoofer and a current location of the UE, that the UE is within a receiving area of the spoofed GNSS signal, and determines a position of the UE without using the spoofed GNSS signal.

Detecting spoofed global navigation satellite system (GNSS) signals

Techniques for providing positioning information to a mobile device are discussed. An example apparatus for determining a location of a mobile device includes at least one server comprising a data structure containing precise positioning subscription options associated with the mobile device, a plurality of client internet of things devices configured to communicate with the at least one server, such that at least one of the plurality of client internet of things devices is a serving internet of things device configured to provide precise positioning information to the mobile device, and such that the serving internet of things device is selected from the plurality of client internet of things devices based on the precise positioning subscription options.

Precise positioning services with an Internet of Things network

An Real-Time Kinematic (RTK) and/or Differential GNSS (DGNSS) system is disclosed in which correction data from a plurality of reference stations is provided to the mobile device. A selection of reference stations (from which correction data is provided to the mobile device) can be made based on factors such as the approximate location of the mobile device, geometry of the reference stations, and/or other factors. The mobile device can combine the correction data from the plurality of reference stations in different ways to determine an accurate position fix for the mobile device, without interpolating correction data from the plurality of reference stations.

Real-time kinematic (RTK) and differential global navigation satellite system (DGNSS) corrections using multiple reference stations

A Real-Time Kinematic (RTK) solution is provided to mobile devices having multi-constellation, multi-frequency (MCMF) functionality, in which a single base station may have a baseline much farther than traditional base station. To enable this, embodiments account for differences in atmospheric effects between the rover station and base station when determining a GNSS position fix for a mobile device (rover station), allowing for a separate tropospheric delay error for a base station to be determined. Embodiments may use additional satellite measurements for which no RTK correction is available, and may further use orbital clock correction for these additional satellite measurements.

Ultra-long baseline RTK

A head-mounted wearable device (HMWD) with a form factor of eyeglasses incorporates hinges between a front frame and temples. The hinges provide a fully enclosed passage for a flexible printed circuit (FPC). The FPC allows electrical connectivity between devices in the front frame and in the temple. The hinge includes a cam device that provides a biasing force that allows the temples to remain in one of four configurations: closed, transit, neutral, and maximum neutral. When in the transit configuration the biasing force provided by the cam device tends to move the temples towards the closed configuration. When in the maximum neutral configuration the biasing force provided by the cam device tends to move the temples towards the neutral configuration. During use, this biasing of the temples provides a gentle clamping force that assists in maintaining the HMWD on the user's head.

Spring hinge with accommodation for flexible printed circuit

A head-mounted wearable device provides many functions to a user. The functions may be provided based in part on whether the device is being worn (donned) or not worn (doffed). Force sensing resistors in the device provide output indicative of forces associated with wearing the device. A time series of values indicative of the magnitude of a force may be determined. An increase or decrease in force values measured at different times may indicate that the device has been donned or doffed, respectively. If a subsequent force value does not differ substantially from a previous force value, this may indicate that the device has remained donned or doffed.

Sensors for determining don/doff status of a wearable device

A head-mounted wearable device (HMWD) incorporates a visual indicator device (VID) that provides a visual cue to a wearer about activity of the HMWD. The VID may be positioned in a frame of the HMWD to be within an upper portion of a wearer's field of view during wear. The VID may provide a diffuse illumination that is visible to the wearer but does not obtrusively illuminate the face of the wearer, resulting in the visual indicator being unobtrusive to bystanders. Activation, intensity, color, blink rate, or other characteristics of the visual indicator may be used to provide the wearer with information. For example, the VID may illuminate with a solid color while a microphone on the HMWD is active, or blink while the microphone is muted. Input from an ambient light sensor may be used to adjust the operation of the VID during operation.

Visual indicator for head-mounted device

Wearable device

A head-mounted wearable device (HMWD) may be fashioned that includes temples that pass near each ear of the user. To accommodate different head shapes and sizes, a head contact piece may be mounted to the temple. Head contact pieces with different profiles may be used to improve user comfort, aid in retaining the HMWD on the user's head, improve performance of output devices such as bone conduction speakers, and so forth.

Ergonomic spacer for head-mounted wearable device

A user interface for a head-mounted wearable device provides many functions to a user. Information about whether the device is being worn (donned) or not worn (doffed) may be acquired in several ways. Force sensing resistors in the device may provide output indicative of the pressure associated with wearing the device. Output from a bone conduction (BC) microphone may be analyzed to determine if the device is in use. Piezoelectric BC speakers typically used to present audio output to a wearer provide changes in voltage corresponding to a change in pressure associated with wearing the device. A BC speaker may emit a signal that is detected by the BC microphone, with changes in the signal strength being used to determine if the device is donned or doffed.

Wearable device don/doff sensor

A head-mounted wearable device (HMWD) may be fashioned that includes temples that pass near each ear of the user. A temple may incorporate a bone conduction speaker (BCS) to provide audio output to the user. During wear, a portion of the BCS is in physical contact with a head of the user to transfer the vibrations from the BCS to the user. To accommodate different head shapes and sizes, a head contact piece may be mounted to the BCS. The user may add, remove, or swap the head contact piece from an assortment of such pieces to find a profile that is comfortable to the wearer and improves performance of the BCS. The head contact piece may be magnetically affixed, allowing for easy changeout.

Field replaceable spacer for head-mounted wearable device

A head-mounted wearable device (HMWD) includes a bone conduction speaker (BCS) to provide audio output to a user wearing the HMWD. The BCS may be mounted within a recess of a temple of the HMWD. During wear a portion of the BCS is in physical contact with a head of the user. In one implementation, a head contact piece may be mounted to the side of the BCS that is proximate to the head. The head contact piece is contoured to maximize the contact area with the head, and may have a cross section that is approximately wedge shaped, with a thicker portion of the wedge proximate to the front of the HMWD. The BCS may include low damping elements to provide protection to the BCS from damage and improve performance, as well as high damping elements to reduce the transfer of vibrations from the BCS to the temple.

Growing community of inventors

Mountain View, CA, United States of America

Han Zhang