Back to results
US 2026/0237177 A1Published application
Determining object orientation from an image with machine learning
AssigneeNVIDIA Corporation
InventorsJonathan Tremblay, Ming-Yu Liu, Dieter Fox, Philip Ammirato
Life of this patent
Estimating the term…
Abstract
Apparatuses, systems, and techniques to determine orientation of an objects in an image. In at least one embodiment, images are processed using a neural network trained to determine orientation of an object.
Claims21
(canceled)
A robotic system, comprising:
- one or more processors to:
- use one or more neural networks to identify an orientation of one or more objects at least partially depicted in one or more images, wherein the one or more neural networks identify the orientation of the one or more objects based on comparisons between features extracted from the one or more images at least partially depicting the one or more objects and features extracted from rendered images of the one or more objects generated using corresponding three-dimensional (3D) models of the objects;
- adjust the orientation of the one or more objects to achieve an alignment between features extracted from the one or more images and the features extracted from the rendered images;
- and
- cause the robotic system to use the orientation to manipulate the one or more objects.
2 more elements
A method, comprising:
- using one or more neural networks to identify an orientation of one or more objects at least partially depicted in one or more images, wherein the one or more neural networks identify the orientation of the one or more objects based on comparisons between features extracted from the one or more images at least partially depicting the one or more objects and features extracted from rendered images of the one or more objects generated using corresponding three-dimensional (3D) models of the objects;
- adjusting the orientation of the one or more objects to achieve an alignment between features extracted from the one or more images and the features extracted from the rendered images;
- and
- causing a robotic system to use the orientation to manipulate the one or more objects.
1 more element
An autonomous vehicle, comprising:
- a camera attached to the autonomous vehicle that obtains one or more images;
- and
- a processor comprising one or more circuits to:
- use one or more neural networks to identify an orientation of one or more objects at least partially depicted in the one or more images, wherein the one or more neural networks identify the orientation of the one or more objects based on comparisons between features extracted from the one or more images at least partially depicting the one or more objects and features extracted from rendered images of the one or more objects generated using corresponding three-dimensional (3D) models of the objects;
- adjust the orientation of the one or more objects to achieve an alignment between features extracted from the one or more images and the features extracted from the rendered images;
- and
- cause a robotic system to use the orientation to manipulate the one or more objects.
4 more elements
Dependencies are inferred from the claim text.