公开(公告)号：US20240338572A1

公开(公告)日：2024-10-10

申请号：US18681763

申请日：2021-08-06

Applicant: Google LLC

Inventor： Nicholas Gillian ,  Lawrence Au

IPC: G06N3/096 ,  G06N3/0455

CPC classification number: G06N3/096 ,  G06N3/0455

Abstract: The present disclosure provides computer-implemented methods, systems, and devices for efficient training of models for use in embedded systems. A model training system accesses unlabeled data elements. The model training system trains one or more encoder models for data encoding of using each unlabeled data element as input. The model training system generates an encoded version of each of a plurality of labeled data elements. The model training system trains decoder models for label generation using the encoded version of the second data set as input. The model training system generates provisional labels for the unlabeled data elements in the first data set, such that each unlabeled data element has an associated provisional label. The model training system trains one or more student models using the unlabeled data elements from the first data set and the associated provisional labels.

公开(公告)号：US20230061808A1

公开(公告)日：2023-03-02

申请号：US17790418

申请日：2019-12-30

Applicant: Google LLC

Inventor： Nicholas Gillian

IPC: G06N5/04

Abstract: A set of interactive objects can implement a machine-learned model for monitoring an activity while communicatively coupled over one or more networks. The machine-learned model can be configured to generate data indicative of at least one inference associated with the activity based at least in part on sensor data associated with two or more interactive objects of the set of interactive objects. The computing system can determine for each interactive object a respective portion of the machine-learned model for execution by the interactive object during at least a portion of the activity. The computing system can generate for each interactive object configuration data indicative of the respective portion of the machine-learned model for execution by the interactive object during the portion of the activity. The computing system can communicate the configuration data indicative of the respective portion of the machine-learned model for execution by to each interactive object.

公开(公告)号：US20250139514A1

公开(公告)日：2025-05-01

申请号：US18833523

申请日：2022-01-26

Applicant: Google LLC

Inventor： Lauren Marie Bedal ,  Lawrence Au ,  Mei Lu ,  Nicholas Gillian ,  Leonardo Giusti

IPC: G06N20/00 ,  G06F3/044 ,  G06F3/04883 ,  G06N3/044 ,  G06N3/0464 ,  G06N3/08

Abstract: The present disclosure provides computer-implemented methods, systems, and devices for efficient bilateral training of users and devices with touch input systems. An interactive object generates, based on a first output of the machine-learned model in response to sensor data associated with a first touch input, first inference data indicating a negative inference corresponding to a first gesture. The interactive object generates, based on an output of the machine-learned model in response to sensor data associated with a second touch input, second inference data indicating a positive inference corresponding to the first gesture. The interactive object, in response to generating the positive inference subsequent to the negative inference, generates training data as a positive training example of the first gesture. The interactive object trains the machine-learned model based at least in part on the training data.

公开(公告)号：US20220269350A1

公开(公告)日：2022-08-25

申请号：US17631788

申请日：2020-07-30

Applicant: Google LLC

Inventor： Nicholas Gillian ,  Daniel Lee Giles

IPC: G06F3/01 ,  G06N3/02 ,  G06F1/16 ,  D03D15/533 ,  D03D15/25 ,  D03D1/00

Abstract: Computing systems and related methods are provided for discovery of undefined user movements. Sensor data associated with one or more sensors of a wearable device can be obtained and input into one or more machine-learned models that have been trained to learn a continuous embedding space based at least in part on one or more target criteria. Data indicative of a position of the sensor data within the continuous embedding space can be obtained as an output of the one or more machine-learned models. A functionality associated with the position of the sensor data within the continuous embedding space can be determined. The functionality associated with the position of the sensor data within the continuous embedding space can be initiated.

公开(公告)号：US20220301353A1

公开(公告)日：2022-09-22

申请号：US17701338

申请日：2022-03-22

Applicant: Google LLC

Inventor： Nicholas Gillian ,  Daniel Lee Giles

IPC: G06V40/20 ,  G06T13/40 ,  G06N20/20

Abstract: A system for motion capture of human body movements includes sensor nodes configured for coupling to respective portions of a human subject. Each sensor node generates inertial sensor data as the human subject engages in a physical activity session and processes the inertial sensor data according to a first machine-learned model to generate a set of local motion determinations. One or more computing devices receive the sets of local determinations and process them according to a second machine-learned model to generate a body motion profile. The computing device(s) provide an animated display of an avatar moving according to the body motion profile and generate training data based on input received from a viewer in response to the animated display. The computing device(s) modify at least one of the first machine-learned model or the second machine-learned model based at least in part on the training data.

公开(公告)号：US20200320416A1

公开(公告)日：2020-10-08

申请号：US16839966

申请日：2020-04-03

Applicant: Google LLC

Inventor： Nicholas Gillian ,  Ivan Poupyrev ,  Gerard Pallipuram

IPC: G06N5/04 ,  G06N20/20

Abstract: A computing system includes at least a first computing device and a second computing device that is physically separate from the first computing device. The computing devices comprise a plurality of processors and a plurality of non-transitory computer-readable media that collectively store a multi-headed machine-learned model that is distributed across the computing devices. The multi-headed machine-learned model comprises a first model head provisioned at the first computing device and configured to receive sensor data from one or more sensors. The first model head is configured to generate a first set of feature representations based at least in part on the sensor data. The multi-headed machine-learned model comprises a second model head provisioned at the second computing device and configured to generate a second set of feature representations in response to receiving data associated with the first set of feature representations from the first computing device.

公开(公告)号：US20200320412A1

公开(公告)日：2020-10-08

申请号：US16839962

申请日：2020-04-03

Applicant: Google LLC

Inventor： Nicholas Gillian ,  Ivan Poupyrev ,  Gerard Pallipuram

IPC: G06N5/04 ,  G06N20/00 ,  G06F3/0488 ,  G06F3/044 ,  G06F3/0484

Abstract: An interactive object includes one or more sensors configured to generate sensor data in response to at least one of a movement of the interactive object or a touch input provided to the interactive object. The interactive object includes at least a first computing device communicatively coupled to the one or more sensors. The first computing device includes one or more non-transitory computer-readable media that store a first model head of a multi-headed machine-learned model that is configured for distribution across a plurality of computing devices including the first computing device. The multi-headed machine-learned model is configured for at least one of a gesture detection or a movement recognition associated with the interactive object. The first model head is configured to selectively generate at least one inference based at least in part on the sensor data and one or more inference criteria.