公开(公告)号：US20220292765A1

公开(公告)日：2022-09-15

申请号：US17201783

申请日：2021-03-15

Applicant: ADOBE INC.

Inventor： Eric-Tuan Le ,  Duygu Ceylan Aksit ,  Tamy Boubekeur ,  Radomir Mech ,  Niloy Mitra ,  Minhyuk Sung

IPC: G06T17/00 ,  G06N3/04

Abstract: Embodiments provide systems, methods, and computer storage media for fitting 3D primitives to a 3D point cloud. In an example embodiment, 3D primitives are fit to a 3D point cloud using a global primitive fitting network that evaluates the entire 3D point cloud and a local primitive fitting network that evaluates local patches of the 3D point cloud. The global primitive fitting network regresses a representation of larger (global) primitives that fit the global structure. To identify smaller 3D primitives for regions with fine detail, local patches are constructed by sampling from a pool of points likely to contain fine detail, and the local primitive fitting network regresses a representation of smaller (local) primitives that fit the local structure of each of the local patches. The global and local primitives are merged into a combined, multi-scale set of fitted primitives, and representative primitive parameters are computed for each fitted primitive.

公开(公告)号：US11488356B2

公开(公告)日：2022-11-01

申请号：US17195099

申请日：2021-03-08

Applicant: ADOBE INC.

Inventor： Radomir Mech ,  Minhyuk Sung ,  Minghua Liu

IPC: G06T11/60 ,  G06T17/20 ,  G06N3/04

Abstract: This disclosure includes technologies for deformation of 3D shapes using meta-handles. The disclosed 3D conditional generative system takes control points with biharmonic coordinates as deformation handles for a shape to train a network to learn a set of meta-handles for the shape. Further, each deformation axis of the latent space of deformation is explicitly associated with a meta-handle from a set of disentangled meta-handles, and the disentangled meta-handles factorize plausible deformations of the shape. Advantageously, an intuitive deformation of the shape may be generated by manipulating coefficients of the meta-handles, e.g., via a user interface.

公开(公告)号：US11682166B2

公开(公告)日：2023-06-20

申请号：US17201783

申请日：2021-03-15

Applicant: ADOBE INC.

Inventor： Eric-Tuan Le ,  Duygu Ceylan Aksit ,  Tamy Boubekeur ,  Radomir Meeh ,  Niloy Mitra ,  Minhyuk Sung

IPC: G06T17/00 ,  G06N3/045

CPC classification number: G06T17/00 ,  G06N3/045

Abstract: Embodiments provide systems, methods, and computer storage media for fitting 3D primitives to a 3D point cloud. In an example embodiment, 3D primitives are fit to a 3D point cloud using a global primitive fitting network that evaluates the entire 3D point cloud and a local primitive fitting network that evaluates local patches of the 3D point cloud. The global primitive fitting network regresses a representation of larger (global) primitives that fit the global structure. To identify smaller 3D primitives for regions with fine detail, local patches are constructed by sampling from a pool of points likely to contain fine detail, and the local primitive fitting network regresses a representation of smaller (local) primitives that fit the local structure of each of the local patches. The global and local primitives are merged into a combined, multi-scale set of fitted primitives, and representative primitive parameters are computed for each fitted primitive.

公开(公告)号：US20220284677A1

公开(公告)日：2022-09-08

申请号：US17195099

申请日：2021-03-08

Applicant: ADOBE INC.

Inventor： Radomir Mech ,  Minhyuk Sung ,  Minghua Liu

IPC: G06T17/20 ,  G06N3/04

Abstract: This disclosure includes technologies for deformation of 3D shapes using meta-handles. The disclosed 3D conditional generative system takes control points with biharmonic coordinates as deformation handles for a shape to train a network to learn a set of meta-handles for the shape. Further, each deformation axis of the latent space of deformation is explicitly associated with a meta-handle from a set of disentangled meta-handles, and the disentangled meta-handles factorize plausible deformations of the shape. Advantageously, an intuitive deformation of the shape may be generated by manipulating coefficients of the meta-handles, e.g., via a user interface.

公开(公告)号：US10204423B2

公开(公告)日：2019-02-12

申请号：US15430659

申请日：2017-02-13

Applicant: Adobe Inc.

Inventor： Vladimir Kim ,  Oliver Wang ,  Minhyuk Sung ,  Mehmet Ersin Yumer

IPC: G06T7/73 ,  G06T7/246 ,  G06T5/00

Abstract: Disclosed are techniques for more accurately estimating the pose of a camera used to capture a three-dimensional scene. Accuracy is enhanced by leveraging three-dimensional object priors extracted from a large-scale three-dimensional shape database. This allows existing feature matching techniques to be augmented by generic three-dimensional object priors, thereby providing robust information about object orientations across multiple images or frames. More specifically, the three-dimensional object priors provide a unit that is easier and more reliably tracked between images than a single feature point. By adding object pose estimates across images, drift is reduced and the resulting visual odometry techniques are more robust and accurate. This eliminates the need for three-dimensional object templates that are specifically generated for the imaged object, training data obtained for a specific environment, and other tedious preprocessing steps. Entire object classes identified in a three-dimensional shape database can be used to train an object detector.