Autodesk, Inc.

Morgan Fabian

Thomas Davies

Michael Haley

Ara Danielyan

Lee Morris Taylor

Ran Zhang

Ebot Etchu Ndip-Agbor

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.

Shaped-based techniques for exploring design spaces

In various embodiments, a training application trains a machine learning model to modify portions of shapes when designing 3D objects. The training application converts first structural analysis data having a first resolution to first coarse structural analysis data having a second resolution that is lower than the first resolution. Subsequently, the training application generates one or more training sets based on a first shape, the first coarse structural analysis data, and a second shape that is derived from the first shape. Each training set is associated with a different portion of the first shape. The training application then performs one or more machine learning operations on the machine learning model using the training set(s) to generate a trained machine learning model. The trained machine learning model modifies at least a portion of a shape having the first resolution based on coarse structural analysis data having the second resolution.

Techniques for training a machine learning model to modify portions of shapes when generating designs for three-dimensional objects

In various embodiments, a topology optimization application solves a topology optimization problem associated with designing a three-dimensional ('3D') object. The topology optimization application coverts a first shape having a first resolution and representing the 3D object to a coarse shape having a second resolution that is lower than the first resolution. Subsequently, the topology optimization application computes coarse structural analysis data based on the coarse shape. The topology optimization application then uses a trained machine learning model to generate a second shape having the first resolution and representing the 3D object based on the first shape and the coarse structural analysis data. The trained machine learning model modifies a portion of a shape having the first resolution based on structural analysis data having the second resolution. Advantageously, generating the second shape based on structural analysis data having a lower resolution reduces computational complexity relative to prior art techniques.

Machine learning techniques for generating designs for three-dimensional objects

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.

Growing community of inventors

San Francisco, CA, United States of America

Morgan Fabian