Back to results
US 12,400,636 B2Granted
Methods and apparatus for training a classification device with constrained capabilities
AssigneeArm Limited
InventorsRemy Pottier, Wei Yuan, James Edward Myers, Andrew William Dunn, Noel Francis Hurley
Life of this patent
Estimating the term…
Abstract
A method for training a classification device, the method comprising: receiving classification device constraints at an intermediary device; receiving training data at the intermediary device; matching the training data to the classification device constraints to provide constrained training data; mapping the constrained training data to classification device functionality to provide a command model; and transmitting the command model to the classification device.
Claims19
A method for training a classification device, the method comprising:
- receiving classification device constraints indicative of limited capabilities of the classification device at an intermediary device;
- receiving training data at the intermediary device;
- determining, at the intermediary device, classification device functionality comprising one or more functions which the classification device can perform;
- matching, at the intermediary device, the training data to the classification device constraints to provide constrained training data;
- mapping, at the intermediary device, the constrained training data to classification device functionality to provide a command model, such that the command model is to define a function of the one or more functions which the classification device is to perform responsive to a user command and such that the defined function is to personalise an operation of an object for a user;
- and
- transmitting, from the intermediary device, the command model to the classification device,
- wherein:
- the training data comprises image training data;
- and
- the command model enables a personalised user experience, the classification device receiving operational image inputs from one or more users to control operation of an object to provide the personalised user experience.
8 more elements
An intermediary device comprising:
- communication circuitry to:
- receive classification device constraints;
- and
- receive training data;
- and
- processing circuitry to:
- determine a classification device functionality of a classification device, the classification device functionality comprising one or more functions which the classification device can perform;
- match the training data to the classification device constraints to provide constrained training data;
- and
- map the constrained training data to the classification device functionality to provide a command model, where the command model is to define a function of the one or more functions which the classification device is to perform responsive to a user command and such that the defined function is to personalise an operation of an object for a user, wherein:
- the communication circuitry transmits the command model to the classification device;
- the training data comprises image training data;
- and
- the command model enables a personalised user experience, the classification device to receive operational image inputs from one or more users to control operation of an object to provide the personalised user experience.
11 more elements
A system for training a classification device, comprising:
- an intermediary device to:
- receive classification device constraints;
- receive training data;
- determine classification device functionality comprising one or more functions which the classification device can perform;
- match the training data to the classification device constraints to provide constrained training data;
- map the constrained training data to the classification device functionality to provide a command model;
- and
- transmit the command model to the classification device;
- and
- a classification device to:
- provide the classification device constraints to the intermediary device;
- and
- receive the command model from the intermediary device, where the command model is to define a function of the one or more functions which the classification device is to perform responsive to a user command such that the defined function is to personalise an operation of an object for a user, wherein:
- the training data comprises image training data;
- and
- the command model enables a personalised user experience, the classification device to receive operational image inputs from one or more users to control operation of an object to provide the personalised user experience.
13 more elements
Dependencies are inferred from the claim text.