Abstract: A method and system provide for annotating a computer-aided design (CAD) drawing. Existing drawings are obtained and include annotations and geometries that serve as hosts. A machine learning (ML) model is trained on the extracted geometries and annotations. A new drawing is obtained. First user input selecting a first geometry in the new drawing is received and the first annotation is created. The ML model generates potential new hosts and annotations. The potential new annotations are displayed in the new drawing and second user input selects one of the potential new annotations to utilize as one or more new annotations.

Abstract: In various embodiments, a training application trains machine learning models to perform tasks associated with 3D CAD objects that are represented using B-reps. In operation, the training application computes a preliminary result via a machine learning model based on a representation of a 3D CAD object that includes a graph and multiple 2D UV-grids. Based on the preliminary result, the training application performs one or more operations to determine that the machine learning model has not been trained to perform a first task. The training application updates at least one parameter of a graph neural network included in the machine learning model based on the preliminary result to generate a modified machine learning model. The training application performs one or more operations to determine that the modified machine learning model has been trained to perform the first task.

Abstract: In various embodiments, a style comparison application compares geometric styles of different three dimensional (3D) computer-aided design (CAD) objects. In operation, the style comparison application executes a trained neural network one or more times to map 3D CAD objects to feature map sets. The style comparison application computes a first set of style signals based on a first feature set included in the feature map sets. The style comparison application computes a second set of style signals based on a second feature set included in the feature map sets. Based on the first set of style signals and the second set of style signals, the style comparison application determines a value for a style comparison metric. The value for the style comparison metric quantifies a similarity or a dissimilarity in geometric style between a first 3D CAD object and a second 3D CAD object.

Abstract: The disclosed method for generating virtual objects includes generating, based on object data, compressed object data, performing, based on the object data and scales, operations to train a first untrained machine learning model to generate a first trained machine learning model comprising a trained codebook and a trained decoder, wherein the first trained machine learning model is trained to generate a reconstruction of the compressed object data, generating, based on the compressed object data and the scales and using the first trained machine learning model, token maps data, performing, based on the token maps data and conditions, operations to train a second untrained machine learning model to generate a second trained machine learning model comprising a trained autoregressive model, wherein the second trained machine learning model is trained to generate predicted token maps, and generating, based on the scales, conditions, and using both trained models, a virtual object.

Abstract: Generative constraining and dimensioning of CAD sketches includes generating one or more candidate constraint sequences using a constraint generation model, generating one or more quality scores for each of the candidate constraint sequences, and performing alignment training on the constraint generation model based on the one or more quality scores and the one or more candidate constraint sequences.

Abstract: Generative constraining and dimensioning of CAD sketches receiving an input sketch, the input sketch including geometric entities; processing the input sketch to determine one or more properties of each of the geometric entities, the one or more properties of a first geometric entity including a plurality of points along the first geometric entity, the points capturing a shape of the first geometric entity; generating embedded tokens from the properties of each of the geometric entities; generating contextualized geometry and constraint embeddings from the embedded tokens using a first transformer; gathering the contextualized geometry and constraint embeddings to generate a plurality of gathered constraints; processing the gathered constraints using a second transformer to generate pointers; and processing the pointers and the geometry and constraint embeddings using a pointer network to autoregressively generate a constraint sequence.

Abstract: Generative constraining and dimensioning of CAD sketches includes receiving training data comprising a plurality of training data elements, each training data element comprising an input sketch and a ground truth constraint sequence, selecting a first training data element from the plurality of training data elements, generating a variable length prompt from the first training data element, presenting the variable length prompt to a constraint generation model to generate a first constraint sequence, generating a loss based on the first constraint sequence, and updating the constraint generation model based on the loss.

Abstract: Techniques are disclosed for generating training datasets and training generative artificial intelligence (AI) models for mechanical assembly designs. A method includes receiving a catalog of mechanical parts and generating a parts grammar that defines compatibility relationships between the parts. Using the parts grammar, one or more combined mechanical assemblies are generated, each comprising compatible mechanical parts. Assembly metrics are then generated by applying one or more physics simulations to the combined mechanical assemblies. A dataset is created based on the assemblies and corresponding assembly metrics, and used to train a generative AI model.

Abstract: Techniques are disclosed for generating training datasets and training generative artificial intelligence (AI) models for mechanical assembly designs. A method includes receiving a catalog of mechanical parts and generating a parts grammar that defines compatibility relationships between the parts. Using the parts grammar, one or more combined mechanical assemblies are generated, each comprising compatible mechanical parts. Assembly metrics are then generated by applying one or more physics simulations to the combined mechanical assemblies. A dataset is created based on the assemblies and corresponding assembly metrics, and used to train a generative AI model.

Abstract: A computer-implemented method is disclosed for generating mechanical assemblies using iterative optimization and generative artificial intelligence (AI). The method includes receiving a mechanical parts catalog and assembly requirements, and executing an iterative generation process. The process comprises generating, via limited sampling, at least one combined mechanical assembly that may satisfy the requirements; generating, via a generative AI model, at least one complete mechanical assembly based on the combined assembly and the requirements; and generating assembly metrics by applying at least one physics simulation to the complete assembly. A reward score is generated based on the assembly metrics, and the iterative generation process is repeated based on the reward score until a convergence threshold is satisfied. The method further includes performing at least one operation associated with the complete mechanical assembly.

Abstract: A method and system provide for annotating a computer-aided design (CAD) drawing. Existing drawings are obtained and include annotations and geometries that serve as annotation points. The geometries are extracted and an Autoregressive Transformer model is trained on the extracted geometries and annotations. A new drawing is obtained. First user input selecting a first geometry in the new drawing is received and the first annotation is created. The new geometries are extracted. The Autoregressive Transformer model generates potential new annotation points and annotations. The potential new annotation points are connected to the selected first geometry and fall within a defined spatial boundary. The potential new annotations are displayed in the new drawing and second user input selects one of the potential new annotations to utilize as one or more new annotations.

Abstract: Various embodiments set forth techniques for generating computer-aided design (CAD) models that include generating a plurality of geometric prompts based on a plurality of inputs, wherein the plurality of inputs indicate at least one geometric value by which at least one CAD model to be generated is to be constrained, and the at least one geometric value is characterized by at least one mathematical inequality, executing a trained machine learning model on the geometric prompts to generate CAD data, and generating the at least one CAD model based on the CAD data, wherein the at least one CAD model is constrained in accordance with the at least one geometric value. Advantageously, the disclosed techniques can substantially facilitate the overall process of designing CAD objects and CAD models of differing levels of complexity, thereby increasing the accessibility of CAD software and applications to a wider array of users.

Abstract: A method and system provide for annotating a computer-aided design (CAD) drawing. Existing drawings are obtained and include annotations and geometries that serve as annotation points. The geometries are extracted and an Autoregressive Transformer model is trained on the extracted geometries and annotations. A new drawing is obtained. First user input selecting a first geometry in the new drawing is received and the first annotation is created. The new geometries are extracted. The Autoregressive Transformer model generates potential new annotation points and annotations. The potential new annotation points are connected to the selected first geometry and fall within a defined spatial boundary. The potential new annotations are displayed in the new drawing and second user input selects one of the potential new annotations to utilize as one or more new annotations.

Abstract: In various embodiments, a parameter domain graph application generates UV-net representations of 3D CAD objects for machine learning models. In operation, the parameter domain graph application generates a graph based on a B-rep of a 3D CAD object. The parameter domain graph application discretizes a parameter domain of a parametric surface associated with the B-rep into a 2D grid. The parameter domain graph application computes at least one feature at a grid point included in the 2D grid based on the parametric surface to generate a 2D UV-grid. Based on the graph and the 2D UV-grid, the parameter domain graph application generates a UV-net representation of the 3D CAD object. Advantageously, generating UV-net representations of 3D CAD objects that are represented using B-reps enables the 3D CAD objects to be processed efficiently using neural networks.

Abstract: In various embodiments, a style comparison metric application generates a style comparison metric for pairs of different three dimensional (3D) computer-aided design (CAD) objects. In operation, the style comparison metric application executes a trained neural network any number of times to map 3D CAD objects to feature maps. Based on the feature maps, the style comparison metric application computes style signals. The style comparison metric application determines values for weights based on the style signals. The style comparison metric application generates the style comparison metric based on the weights and a parameterized style comparison metric.

Abstract: In various embodiments, a parameter domain graph application generates UV-net representations of 3D CAD objects for machine learning models. In operation, the parameter domain graph application generates a graph based on a B-rep of a 3D CAD object. The parameter domain graph application discretizes a parameter domain of a parametric surface associated with the B-rep into a 2D grid. The parameter domain graph application computes at least one feature at a grid point included in the 2D grid based on the parametric surface to generate a 2D UV-grid. Based on the graph and the 2D UV-grid, the parameter domain graph application generates a UV-net representation of the 3D CAD object. Advantageously, generating UV-net representations of 3D CAD objects that are represented using B-reps enables the 3D CAD objects to be processed efficiently using neural networks.

Abstract: Techniques for generative design include a computer-implemented method for solving a design problem comprising initializing values for one or more categorical and continuous design variables, and performing a design iteration by generating sample vectors for each of one or more categorical design variables based on the categorical design variable probabilities, solving one or more governing equations for the design problem based on values of the continuous design variables and the sample vectors, computing a value of one or more constraint functions and an objective function, computing first gradients of the objective function and the constraint functions with respect to each of the continuous design variables, computing second gradients of the objective function and the constraint functions with respect to the categorical design variable probabilities, and updating values for the continuous design variables based on the first gradients and values for the categorical design variable probabilities based on the

Abstract: Techniques for interactive generative design with sensitivity analysis and probability visualization for categorical design variables include a computer-implemented method for evaluating an impact of categorical design variables on a design problem solution comprises receiving information regarding choices for one or more categorical design variables associated with each of a plurality of design members of a design problem, determining a respective sensitivity of an objective function to the choices for the one or more categorical design variables for each design member of the plurality of design members, determining a respective visual aspect for each design member based on the respective sensitivity, displaying, on a user interface, a graphical depiction of the plurality of design members, wherein each design member is displayed using the respective visual aspect, and displaying, on the user interface, a key for interpreting the respective visual aspects.

Abstract: Techniques for generative design include a computer-implemented method for solving a design problem comprising initializing values for one or more categorical design variable probabilities and one or more continuous design variables, and performing a design iteration by performing one or more iterations to update the categorial design variable probabilities by generating sample vectors for each of one or more categorical design variables based on the categorical design variable probabilities, computing first gradients of an objective function and one or more constraint functions with respect to the categorical design variable probabilities, and updating values for the categorical design variable probabilities based on the first gradients, then updating the sample vectors based on the updated categorical design variable probability values, computing second gradients of the objective and constraint functions with respect to each of the continuous design variables, and updating values for the continuous design v

Abstract: One embodiment of the present invention sets forth a technique for performing 3D shape generation. This technique includes generating semantic features associated with an input sketch. The technique also includes generating, using a generative machine learning model, a plurality of predicted shape embeddings from a set of fully masked shape embeddings based on the semantic features associated with the input sketch. The technique further includes converting the predicted shape embeddings into one or more 3D shapes. The input sketch may be a casual doodle, a professional illustration, or a 2D CAD software rendering. Each of the one or more 3D shapes may be a voxel representation, an implicit representation, or a 3D CAD software representation.