公开(公告)号：US20210165934A1

公开(公告)日：2021-06-03

申请号：US17107540

申请日：2020-11-30

Applicant: DASSAULT SYSTEMES

Inventor： Nicolas MONTANA ,  Andre LIEUTIER ,  Marc MONTEIL

IPC: G06F30/20 ,  G06T11/20 ,  G06T7/155

Abstract: A computer-implemented method for civil engineering is described that includes obtaining a watershed segmentation of a terrain. The watershed segmentation includes basins. The method further includes merging first basins of the watershed segmentation that each verify a smallness criterion, each with a second basin downstream to the first basin.

公开(公告)号：US20210165935A1

公开(公告)日：2021-06-03

申请号：US17107434

申请日：2020-11-30

Applicant: DASSAULT SYSTEMES

Inventor： Nicolas MONTANA ,  Andre LIEUTIER ,  Marc MONTEIL

IPC: G06F30/23 ,  G06F30/13 ,  G06T17/20 ,  G06T7/10

Abstract: A computer-implemented method for civil engineering including obtaining a mesh representing a terrain and a polyline on the mesh, the method further includes computing a contributor of the polyline. The computing of the contributor includes modifying the mesh by determining, based on the polyline, a trench below the polyline. The computing of the contributor further includes computing a watershed segmentation of the terrain based on the modified mesh. The computing of the contributor further includes, based on the computed watershed segmentation, identifying, on the modified mesh, a basin comprising the trench. The contributor corresponds to the identified basin.

公开(公告)号：US20200285907A1

公开(公告)日：2020-09-10

申请号：US16879507

申请日：2020-05-20

Applicant: Dassault Systemes

Inventor： Eloi MEHR ,  Andre LIEUTIER

IPC: G06K9/62 ,  G06K9/66 ,  G06K9/00 ,  G06N3/08 ,  G06N3/04 ,  G06T7/30 ,  G06T3/00 ,  G06F30/00

Abstract: A computer-implemented method for learning an autoencoder notably is provided. The method includes obtaining a dataset of images. Each image includes a respective object representation. The method also includes learning the autoencoder based on the dataset. The learning includes minimization of a reconstruction loss. The reconstruction loss includes a term that penalizes a distance for each respective image. The penalized distance is between the result of applying the autoencoder to the respective image and the set of results of applying at least part of a group of transformations to the object representation of the respective image. Such a method provides an improved solution to learn an autoencoder.

公开(公告)号：US20210166398A1

公开(公告)日：2021-06-03

申请号：US17107032

申请日：2020-11-30

Applicant: DASSAULT SYSTEMES

Inventor： Nicolas MONTANA ,  Andre LIEUTIER ,  Marc MONTEIL

IPC: G06T7/155 ,  G06T7/187 ,  G06T17/20 ,  G06T17/05

Abstract: Computer-implemented method for civil engineering including obtaining a mesh representing a terrain and computing a watershed segmentation of the terrain based on the mesh, the computing of the watershed segmentation including identifying one or more saddle points on the mesh, for each identified saddle point, identifying paths each ascending from the saddle point according to a direction of a local maximal slope around the saddle point, and a path descending from the saddle point according to a direction of a steepest slope around the saddle point, the identified ascending paths dividing the mesh into connected components, and, for each identified saddle point, merging each connected component of which bottom point is the saddle point with a connected component including the identified path descending from the saddle point according to a direction of a steepest slope. The merging yields at least a part of a basin.

公开(公告)号：US20200327727A1

公开(公告)日：2020-10-15

申请号：US16846139

申请日：2020-04-10

Applicant: DASSAULT SYSTEMES

Inventor： Andre LIEUTIER ,  Julien VUILLAMY ,  David COHEN-STEINER

IPC: G06T17/20 ,  G06T7/50

Abstract: A computer-implemented method for 3D reconstruction of a structure of a real scene including obtaining a 3D point cloud representing the structure. The method further includes determining a closed triangulated surface representing a skin of the structure. The determining explores candidate closed triangulated surfaces each meshing at least a respective part of the 3D point cloud. The determining penalizes a high rank of the candidate closed triangulated surfaces according to a lexicographic order. The lexicographic order is based on a triangle order and orders a first candidate closed triangulated surface having first triangles which are ordered according to a decreasing rank of the triangle order, relative to a second candidate closed triangulated surface having second triangles which are ordered according to a decreasing rank of the triangle order. The triangle order penalizes a triangle size. This constitutes an improved method for 3D reconstruction of a structure of a real scene.