Joint shape and appearance optimization by topological sampling
Bond point shape and appearance optimization through topological sampling is disclosed. The proposed systems and methods enable optimization of 3D model representations including the shape and appearance of a particular 3D scene or object. The opaque 3D mesh (e.g., vertex positions and corresponding...
Saved in:
Main Authors | , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
06.12.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Bond point shape and appearance optimization through topological sampling is disclosed. The proposed systems and methods enable optimization of 3D model representations including the shape and appearance of a particular 3D scene or object. The opaque 3D mesh (e.g., vertex positions and corresponding topologies) and spatially varying material properties are jointly optimized based on image spatial loss to match multiple image observations (e.g., reference images referring to the 3D scene or object). Geometric topologies define faces and/or cells that are visible in the opaque 3D mesh and may be randomly initialized and optimized by training based on image space loss. Applying a geometric topology to the opaque 3D mesh to learn the shape improves the accuracy and performance of the contour edge as compared to using a transparent mesh representation. The 3D model representation is learned based on image spatial differences, without the need for initial guessing, compared to methods that require initial guessing |
---|---|
Bibliography: | Application Number: CN202210558436 |