公开(公告)号：US12183037B2

公开(公告)日：2024-12-31

申请号：US18487800

申请日：2023-10-16

Applicant: NVIDIA Corporation

Inventor： Sravya Nimmagadda ,  David Weikersdorfer

IPC: G06T7/73 ,  B25J9/16 ,  G06N3/045 ,  G06N3/08 ,  G06V20/10

Abstract: An autoencoder may be trained to predict 3D pose labels using simulation data extracted from a simulated environment, which may be configured to represent an environment in which the 3D pose estimator is to be deployed. Assets may be used to mimic the deployment environment such as 3D models or textures and parameters used to define deployment scenarios and/or conditions that the 3D pose estimator will operate under in the environment. The autoencoder may be trained to predict a segmentation image from an input image that is invariant to occlusions. Further, the autoencoder may be trained to exclude areas of the input image from the object that correspond to one or more appendages of the object. The 3D pose may be adapted to unlabeled real-world data using a GAN, which predicts whether output of the 3D pose estimator was generated from real-world data or simulated data.

公开(公告)号：US20220284624A1

公开(公告)日：2022-09-08

申请号：US17191543

申请日：2021-03-03

Applicant: NVIDIA Corporation

Inventor： Sravya Nimmagadda ,  David Weikersdorfer

IPC: G06T7/73 ,  G06K9/00 ,  G06N3/08 ,  G06N3/04 ,  B25J9/16

Abstract: An autoencoder may be trained to predict 3D pose labels using simulation data extracted from a simulated environment, which may be configured to represent an environment in which the 3D pose estimator is to be deployed. Assets may be used to mimic the deployment environment such as 3D models or textures and parameters used to define deployment scenarios and/or conditions that the 3D pose estimator will operate under in the environment. The autoencoder may be trained to predict a segmentation image from an input image that is invariant to occlusions. Further, the autoencoder may be trained to exclude areas of the input image from the object that correspond to one or more appendages of the object. The 3D pose may be adapted to unlabeled real-world data using a GAN, which predicts whether output of the 3D pose estimator was generated from real-world data or simulated data.

公开(公告)号：US20240037788A1

公开(公告)日：2024-02-01

申请号：US18487800

申请日：2023-10-16

Applicant: Nvidia Corporation

Inventor： Sravya Nimmagadda ,  David Weikersdorfer

IPC: G06T7/73 ,  B25J9/16 ,  G06N3/08 ,  G06V20/10 ,  G06N3/045

CPC classification number: G06T7/75 ,  B25J9/1697 ,  G06N3/08 ,  G06V20/10 ,  G06N3/045

Abstract: An autoencoder may be trained to predict 3D pose labels using simulation data extracted from a simulated environment, which may be configured to represent an environment in which the 3D pose estimator is to be deployed. Assets may be used to mimic the deployment environment such as 3D models or textures and parameters used to define deployment scenarios and/or conditions that the 3D pose estimator will operate under in the environment. The autoencoder may be trained to predict a segmentation image from an input image that is invariant to occlusions. Further, the autoencoder may be trained to exclude areas of the input image from the object that correspond to one or more appendages of the object. The 3D pose may be adapted to unlabeled real-world data using a GAN, which predicts whether output of the 3D pose estimator was generated from real-world data or simulated data.

公开(公告)号：US20250139827A1

公开(公告)日：2025-05-01

申请号：US19005645

申请日：2024-12-30

Applicant: NVIDIA Corporation

Inventor： Sravya Nimmagadda ,  David Weikersdorfer

IPC: G06T7/73 ,  B25J9/16 ,  G06N3/045 ,  G06N3/08 ,  G06V20/10

Abstract: An autoencoder may be trained to predict 3D pose labels using simulation data extracted from a simulated environment, which may be configured to represent an environment in which the 3D pose estimator is to be deployed. Assets may be used to mimic the deployment environment such as 3D models or textures and parameters used to define deployment scenarios and/or conditions that the 3D pose estimator will operate under in the environment. The autoencoder may be trained to predict a segmentation image from an input image that is invariant to occlusions. Further, the autoencoder may be trained to exclude areas of the input image from the object that correspond to one or more appendages of the object. The 3D pose may be adapted to unlabeled real-world data using a GAN, which predicts whether output of the 3D pose estimator was generated from real-world data or simulated data.

公开(公告)号：US11823415B2

公开(公告)日：2023-11-21

申请号：US17191543

申请日：2021-03-03

Applicant: NVIDIA Corporation

Inventor： Sravya Nimmagadda ,  David Weikersdorfer

IPC: G06T7/73 ,  B25J9/16 ,  G06N3/08 ,  G06V20/10 ,  G06N3/045

CPC classification number: G06T7/75 ,  B25J9/1697 ,  G06N3/045 ,  G06N3/08 ,  G06V20/10

Abstract: An autoencoder may be trained to predict 3D pose labels using simulation data extracted from a simulated environment, which may be configured to represent an environment in which the 3D pose estimator is to be deployed. Assets may be used to mimic the deployment environment such as 3D models or textures and parameters used to define deployment scenarios and/or conditions that the 3D pose estimator will operate under in the environment. The autoencoder may be trained to predict a segmentation image from an input image that is invariant to occlusions. Further, the autoencoder may be trained to exclude areas of the input image from the object that correspond to one or more appendages of the object. The 3D pose may be adapted to unlabeled real-world data using a GAN, which predicts whether output of the 3D pose estimator was generated from real-world data or simulated data.