Back to results
US 11,590,650 B2Granted
Generation method for training dataset, model generation method, training data generation apparatus, inference apparatus, robotic controller, model training method and robot
AssigneePreferred Networks, Inc.
InventorsKenta Yonekura, Kuniyuki Takahashi
Life of this patent
Estimating the term…
Abstract
One aspect of the present disclosure relates to a generation method for a training dataset, comprising: capturing, by one or more processors, a target object to which a marker unit recognizable under a first illumination condition is provided; and acquiring, by the one or more processors, a first image where the marker unit is recognizable and a second image obtained by capturing the target object under a second illumination condition.
Claims20
A generation method for a training dataset, comprising:
- acquiring, by one or more processors, a first image by capturing, under a first illumination condition, a target object to which a marker is provided;
- and
- acquiring, by the one or more processor, a second image by capturing the target object under a second illumination condition,
- wherein the marker is recognizable under invisible light, the second illumination condition is under visible light, and the first illumination condition is under invisible light.
1 more element
A training data generation system, comprising:
- one or more memories;
- and
- one or more processors configured to:
- emit first illumination conditioned light to a target object to which a marker is provided;
- emit second illumination conditioned light to the target object;
- switch between illumination of the first illumination conditioned light to the target object and illumination of the second illumination conditioned light to the target object;
- acquire a first image by capturing the target object in a state where the marker becomes recognizable by emitting the first illumination conditioned light to the target object;
- and
- acquire a second image by capturing the target object in a state where the second illumination conditioned light is emitted to the target object,
- wherein the marker is recognizable under invisible light, the second illumination conditioned light is visible light, and the first illumination conditioned light is invisible light.
7 more elements
A robot, comprising:
- one or more memories that store a model trained in accordance with a model generation method including using a first image obtained by capturing, under a first illumination condition, a target object to which a marker recognizable under the first illumination condition is provided, and a second image obtained by capturing the target object under a second illumination condition to generate the model the model;
- and
- one or more processors configured to control an operation of the robot in a state where second illumination conditioned light is emitted to a to-be-operated object.
A generation method for a training dataset, comprising:
- acquiring, by one or more processors, a first image by capturing, under a first illumination conditioned light, a target object to which a marker recognizable under the first illumination conditioned light is provided;
- and
- acquiring, by the one or more processors, a second image by capturing the target object under a second illumination conditioned light,
- wherein the marker includes a recursive reflective material, the second illumination conditioned light is environmental light, and reflected light from the marker is captured under the first illumination conditioned light.
1 more element
A training data generation system, comprising:
- one or more memories;
- and
- one or more processors configured to:
- emit first illumination conditioned light to a target object to which a marker is provided;
- emit second illumination conditioned light to the target object;
- switch between illumination of the first illumination conditioned light to the target object and illumination of the second illumination conditioned light to the target object;
- acquire a first image by capturing the target object in a state where the marker becomes recognizable by emitting the first illumination conditioned light to the target object;
- and
- acquire a second image by capturing the target object in a state where the second illumination conditioned light is emitted to the target object,
- wherein the marker includes a recursive reflective material, and the second illumination conditioned light is environmental light, and reflected light from the marker is captured under the first illumination conditioned light.
7 more elements
Dependencies are inferred from the claim text.