Dassault Systemes

Dassault Systems

Eloi Mehr

Vincent Guitteny

Fernando Manuel Sanchez Bermudez

Andre Lieutier

Mourad Boufarguine

Serban Alexandru State

Yoan Souty

David Leo Bonner

Guillaume Randon

André Lieutier

Adil Baaj

Patrick Johnson

Ariane Jourdan

Mathieu Brus

David Bonner

A computer-implemented method of machine-learning including obtaining a dataset of training samples. Each training sample includes a pair of 3D modeled object portions labelled with a respective value. The respective value indicates whether or not the two portions belong to a same segment of a 3D modeled object. The method further includes learning a neural network based on the dataset. The neural network takes as input two portions of a 3D modeled object representing a mechanical part and outputs a respective value. The respective value indicates an extent to which the two portions belong to a same segment of the 3D modeled object. The neural network is thereby usable for 3D segmentation. The method constitutes an improved solution for 3D segmentation.

Machine-learning for 3D segmentation

A computer-implemented method for 3D reconstruction including obtaining 2D images and, for each 2D image, camera parameters which define a perspective projection. The 2D images all represent a same real object. The real object is fixed. The method also includes obtaining, for each 2D image, a smooth map. The smooth map has pixel values, and each pixel value represents a measurement of contour presence. The method also includes determining a 3D modeled object that represents the real object. The determining iteratively optimizes energy. The energy rewards, for each smooth map, projections of silhouette vertices of the 3D modeled object having pixel values representing a high measurement of contour presence. This forms an improved solution for 3D reconstruction.

3D reconstruction with smooth maps

A method for processing a shape attribute 3D signal including providing a graph having nodes and arcs, each node representing a point of a 3D discrete representation, each arc representing neighboring points of the representation, providing a set of values representing a distribution of the shape attribute, each value being associated to a node and representing the shape attribute at the point represented by the node, minimizing energy on a Markov Random Field on the graph, the energy penalizing, for each arc connecting a first node associated to a first value to a second node associated to a second value, highness of an increasing function of a distance between the first and second value, a distance between a first point, represented by the first node, and a medial geometrical element of the representation, and a distance between a second point, represented by the second node, and the medial geometrical element.

Processing a 3D signal of a shape attribute over a real object

A computer-implemented method of augmented reality includes capturing the video flux with a video camera, extracting, from the video flux, one or more 2D images each representing the real object, and obtaining a 3D model representing the real object. The method also includes determining a pose of the 3D model relative to the video flux, among candidate poses. The determining rewards a mutual information, for at least one 2D image and for each given candidate pose, which represents a mutual dependence between a virtual 2D rendering and the at least one 2D image. The method also includes augmenting the video flux based on the pose. This forms an improved solution of augmented reality for augmenting a video flux of a real scene including a real object.

Augmenting a video flux of a real scene

The disclosure notably relates to a computer-implemented method of machine-learning. The method includes obtaining a dataset including 3D modeled objects which each represent a respective mechanical part and further includes providing a set of neural networks. Each neural network has respective weights. Each neural network is configured for inference of 3D modeled objects. The method further includes modifying respective weights of the neural networks by minimizing a loss. For each 3D modeled object, the loss selects a term among a plurality of terms. Each term penalizes a disparity between the 3D modeled object and a respective 3D modeled object inferred by a respective neural network of the set. The selected term is a term among the plurality of terms for which the disparity is the least penalized. This constitutes an improved method of machine-learning with a dataset including 3D modeled objects which each represent a respective mechanical part.

Set of neural networks

A computer-implemented method for designing a 3D modeled object via user-interaction. The method includes obtaining the 3D modeled object and a machine-learnt decoder. The machine-learnt decoder is a differentiable function taking values in a latent space and outputting values in a 3D modeled object space. The method further includes defining a deformation constraint for a part of the 3D modeled object. The method further comprises determining an optimal vector. The optimal vector minimizes an energy. The energy explores latent vectors. The energy comprises a term which penalizes, for each explored latent vector, non-respect of the deformation constraint by the result of applying the decoder to the explored latent vector. The method further includes applying the decoder to the optimal latent vector. This constitutes an improved method for designing a 3D modeled object via user-interaction.

Designing a 3D modeled object via user-interaction

A computer-implemented method for forming a dataset configured for learning a neural network. The neural network is configured for inference, from a freehand drawing representing a 3D shape, of a solid CAD feature representing the 3D shape. The method includes generating one or more solid CAD feature includes each representing a respective 3D shape. The method also includes, for each solid CAD feature, determining one or more respective freehand drawings each representing the respective 3D shape, and inserting in the dataset, one or more training samples. Each training sample includes the solid CAD feature and a respective freehand drawing. The method forms an improved solution for inference, from a freehand drawing representing a 3D shape, of a 3D modeled object representing the 3D shape.

Forming a dataset for inference of solid CAD features

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.

Learning an autoencoder

The disclosure notably relates to a computer-implemented method of machine-learning. The method includes obtaining a dataset including 3D modeled objects which each represent a respective mechanical part. The dataset has one or more sub-datasets. Each sub-dataset forms at least a part of the dataset. The method further includes, for each respective sub-dataset, determining a base template and learning a neural network configured for inference of deformations of the base template each into a respective 3D modeled object. The base template is a 3D modeled object which represents a centroid of the 3D modeled objects of the sub-dataset. The learning includes a training based on the sub-dataset. This constitutes an improved method of machine-learning with a dataset including 3D modeled objects which each represent a respective mechanical part.

Machine-learning for 3D modeled object inference

The disclosure notably relates to a computer-implemented method for learning a neural network configured for inference, from a discrete geometrical representation of a 3D shape, of an editable feature tree representing the 3D shape. The editable feature tree includes a tree arrangement of geometrical operations applied to leaf geometrical shapes. The method includes obtaining a dataset including discrete geometrical representations each of a respective 3D shape, and obtaining a candidate set of leaf geometrical shapes. The method also includes learning the neural network based on the dataset and on the candidate set. The candidate set includes at least one continuous subset of leaf geometrical shapes. The method forms an improved solution for digitization.

Learning a neural network for inference of editable feature trees

The disclosure notably relates to a computer-implemented method for forming a dataset configured for learning a neural network. The neural network is configured for inference, from a discrete geometrical representation of a 3D shape, of an editable feature tree representing the 3D shape. The editable feature tree comprises a tree arrangement of geometrical operations applied to leaf geometrical shapes. The method includes obtaining respective data pieces, and inserting a part of the data pieces in the dataset each as a respective training sample. The respective 3D shape of each of one or more first data pieces inserted in the dataset is identical to the respective 3D shape of respective one or more second data pieces not inserted in the dataset. The method forms an improved solution for digitization.

Forming a dataset for inference of editable feature trees

A computer-implemented method for learning an autoencoder notably is provided. The method comprises providing a dataset of images. Each image includes a respective object representation. The method also comprises 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.

A computer-implemented method for determining an architectural layout. The method comprises providing a cycle of points that represents a planar cross section of a cycle of walls, and, assigned to each respective point, a respective first datum that represents a direction normal to the cycle of points at the respective point. The method also comprises minimizing a Markov Random Field energy thereby assigning, to each respective point, a respective one of the set of second data. The method also comprises identifying maximal sets of consecutive points to which a same second datum is assigned, and a cycle of vertices bounding a cycle of segments which represents the architectural layout. Such a method constitutes an improved solution for determining an architectural layout.

Determining an architectural layout

A computer-implemented method of computer vision in a scene that includes one or more transparent objects and/or one or more reflecting objects comprises obtaining a plurality of images of the scene, each image corresponding to a respective acquisition of a physical signal, the plurality of images including at least two images corresponding to different physical signals; and generating a segmented image of the scene based on the plurality of images. This improves the field of computer vision.

Growing community of inventors

Velizy Villacoublay, France

Eloi Mehr