公开(公告)号：US11741662B2

公开(公告)日：2023-08-29

申请号：US16174115

申请日：2018-10-29

Applicant: Autodesk, Inc.

Inventor： Thomas Davies ,  Michael Haley ,  Ara Danielyan ,  Morgan Fabian

IPC: G06T15/20 ,  G06N3/08 ,  G06N3/04 ,  G06F18/22 ,  G06F18/23 ,  G06V10/75 ,  G06V10/764 ,  G06V10/82 ,  G06V20/64

CPC classification number: G06T15/20 ,  G06F18/22 ,  G06F18/23 ,  G06N3/04 ,  G06N3/08 ,  G06V10/76 ,  G06V10/764 ,  G06V10/82 ,  G06V20/64 ,  G06T2200/24

Abstract: In various embodiments, a training application generates a trained encoder that automatically generates shape embeddings having a first size and representing three-dimensional (3D) geometry shapes. First, the training application generates a different view activation for each of multiple views associated with a first 3D geometry based on a first convolutional neural network (CNN) block. The training application then aggregates the view activations to generate a tiled activation. Subsequently, the training application generates a first shape embedding having the first size based on the tiled activation and a second CNN block. The training application then generates multiple re-constructed views based on the first shape embedding. The training application performs training operation(s) on at least one of the first CNN block and the second CNN block based on the views and the re-constructed views to generate the trained encoder.

公开(公告)号：US11380045B2

公开(公告)日：2022-07-05

申请号：US16174110

申请日：2018-10-29

Applicant: Autodesk, Inc.

Inventor： Thomas Davies ,  Michael Haley ,  Ara Danielyan ,  Morgan Fabian

IPC: G06T15/20 ,  G06N3/08 ,  G06N3/04 ,  G06F30/00

Abstract: In various embodiments, a training application generates a trained encoder that automatically generates shape embeddings having a first size and representing three-dimensional (3D) geometry shapes. First, the training application generates a different view activation for each of multiple views associated with a first 3D geometry based on a first convolutional neural network (CNN) block. The training application then aggregates the view activations to generate a tiled activation. Subsequently, the training application generates a first shape embedding having the first size based on the tiled activation and a second CNN block. The training application then generates multiple re-constructed views based on the first shape embedding. The training application performs training operation(s) on at least one of the first CNN block and the second CNN block based on the views and the re-constructed views to generate the trained encoder.

公开(公告)号：US11126330B2

公开(公告)日：2021-09-21

申请号：US16174119

申请日：2018-10-29

Applicant: Autodesk, Inc.

Inventor： Thomas Davies ,  Michael Haley ,  Ara Danielyan ,  Morgan Fabian

IPC: G06F3/0482 ,  G06F3/0481 ,  G06F3/0484 ,  G06N3/04 ,  G06N3/08 ,  G06T11/20 ,  G06T15/20 ,  G06K9/62 ,  G06F30/00 ,  G06F3/048

Abstract: In various embodiments, a training application generates a trained encoder that automatically generates shape embeddings having a first size and representing three-dimensional (3D) geometry shapes, First, the training application generates a different view activation for each of multiple views associated with a first 3D geometry based on a first convolutional neural network (CNN) block. The training application then aggregates the view activations to generate a tiled activation. Subsequently, the training application generates a first shape embedding having the first size based on the tiled activation and a second CNN block. The training application then generates multiple re-constructed views based on the first shape embedding. The training application performs training operation(s) on at least one of the first CNN block and the second CNN block based on the views and the re-constructed views to generate the trained encoder.