公开(公告)号：US20240242483A1

公开(公告)日：2024-07-18

申请号：US18097946

申请日：2023-01-17

Applicant: ADOBE INC.

Inventor： Prafull Sharma ,  Valentin Mathieu Deschaintre ,  Michaël Yanis Gharbi ,  Julien Olivier Victor Philip

IPC: G06V10/774 ,  G06V10/74

CPC classification number: G06V10/774 ,  G06V10/761

Abstract: A model is trained to predict pixels from an image that correspond to a material of a selected pixel using contrastive loss. A training dataset comprising a training image and material information for the training image is received. An anchor pixel in the training image is identified. To train the model using the training dataset, the model generates embeddings for pixels of the training image, including the anchor pixel and a plurality of other pixels. A contrastive loss is determined based on a comparison of the embeddings to the material information. The model is updated based on the loss.

公开(公告)号：US20250078408A1

公开(公告)日：2025-03-06

申请号：US18458032

申请日：2023-08-29

Applicant: Adobe Inc.

Inventor： Valentin Mathieu Deschaintre ,  Vladimir Kim ,  Thibault Groueix ,  Julien Philip

IPC: G06T17/20 ,  G06T7/40 ,  H04N13/279

Abstract: Implementations of systems and methods for determining viewpoints suitable for performing one or more digital operations on a three-dimensional object are disclosed. Accordingly, a set of candidate viewpoints is established. The subset of candidate viewpoints provides views of an outer surface of a three-dimensional object and those views provide overlapping surface data. A subset of activated viewpoints is determined from the set of candidate viewpoints, the subset of activated viewpoints providing less of the overlapping surface data. The subset of activated viewpoints is used to perform one or more digital operation on the three-dimensional object.

公开(公告)号：US20240404244A1

公开(公告)日：2024-12-05

申请号：US18329385

申请日：2023-06-05

Applicant: Adobe Inc.

Inventor： Valentin Mathieu Deschaintre ,  Yiwei Hu ,  Paul Guerrero ,  Milos Hasan

IPC: G06V10/74 ,  G06V10/44 ,  G06V10/54

Abstract: Conditional procedural model generation techniques are described that enable generation of procedural models that are usable to recreate a visual appearance of an input image. A content processing system, for instance, receives a training dataset that includes a plurality of training pairs. The content processing system trains a machine learning model to generate procedural models based on input images. The content processing system then receives an input image that has a particular visual appearance. The content processing system leverages the trained machine learning model to generate a procedural model that is usable to recreate the particular visual appearance of the input digital image.