Differentiable 3D CAD Programs for Bidirectional Editing

  title={Differentiable 3D CAD Programs for Bidirectional Editing},
  author={Dan Cașcaval and Mira Shalah and Phillip Quinn and Rastislav Bod{\'i}k and Maneesh Agrawala and Adriana Schulz},
  journal={Computer Graphics Forum},
Modern CAD tools represent 3D designs not only as geometry, but also as a program composed of geometric operations, each of which depends on a set of parameters. Program representations enable meaningful and controlled shape variations via parameter changes. However, achieving desired modifications solely through parameter editing is challenging when CAD models have not been explicitly authored to expose select degrees of freedom in advance. We introduce a novel bidirectional editing system for… 

Automatic Differentiable Procedural Modeling

An auto‐differentiable representation of the procedural model that significantly accelerates optimization and allows for efficient derivative computation and enables interactivity, and designed a new optimizer to solve for inverse procedural modeling.

ComplexGen: CAD Reconstruction by B-Rep Chain Complex Generation

A novel neural framework is proposed that consists of a sparse CNN encoder for input point cloud processing and a tri-path transformer decoder for generating geometric primitives and their mutual relationships with estimated probabilities, leading to structurally more faithful and complete CAD B-Rep models than previous results.

HybridSDF: Combining Deep Implicit Shapes and Geometric Primitives for 3D Shape Representation and Manipulation

This paper proposes a representation combining latent and explicit parameters that can be decoded into a set of deep implicit and geometric shapes that are consistent with each other, which can effectively model both complex and highly regular shapes that coexist in manufactured objects.

Certifiable Robot Design Optimization using Differentiable Programming

This paper combines differentiable programming for robot design optimization with a novel statistical framework for certifying the robustness of optimized designs, and shows how the tool can be used to solve practical problems in robotics.

Neural Shape Compiler: A Unified Framework for Transforming between Text, Point Cloud, and Program

The proposed Neural Shape Compiler converts 3D shapes of three abstract types into unified discrete shape code, transforms each shape code into code of other abstract types through the proposed ShapeCode Transformer, and decodes them to output the target shape abstraction.

Robust Counterexample-guided Optimization for Planning from Differentiable Temporal Logic

  • Charles DawsonChuchu Fan
  • Computer Science
    2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
  • 2022
This paper presents an algorithm for finding robust plans that satisfy STL specifications, using automatically differentiable temporal logic to iteratively optimize its plan in response to counterexamples found during the falsification process.

Differentiable Hybrid Traffic Simulation

This is the first differentiable traffic simulator for macroscopic and hybrid models that can compute gradients for traffic states across time steps and inhomogeneous lanes and can provide more efficient and scalable solutions for complex learning and control problems posed in traffic engineering than other existing algorithms.



Interactive design space exploration and optimization for CAD models

This work uses precomputation on an adaptively sampled grid and proposes a novel scheme for interpolating in this domain where each sample is a mesh with different combinatorics, which can be used to interactively visualize and optimize objects with a variety of physical properties.

Local Editing of Procedural Models

This work introduces a bottom‐up algorithm for finding GELs directly from the attributes in procedural models, without special annotations, and introduces an algorithm to transfer meta‐locators to a different procedural model.

Synthesizing structured CAD models with equality saturation and inverse transformations

A second decompilation stage is proposed to robustly "shrink" unstructured CSG expressions into more editable programs with map and fold operators and presents Szalinski, a tool that uses Equality Saturation with semantics-preserving CAD rewrites to efficiently search for smaller equivalent programs.

X-CAD: Optimizing CAD Models with Extended Finite Elements

The XFEM method works directly on the intersection between the model and a regular simulation grid, without the need to mesh or remesh, thus removing a bottleneck of classical shape optimization strategies.

InverseCSG: automatic conversion of 3D models to CSG trees

This paper presents a method for reverse-engineering the process by which 3D models may have been generated, in the language of constructive solid geometry (CSG), using an algorithm that couples geometric processing with state-of-the-art program synthesis techniques.

Component‐wise Controllers for Structure‐Preserving Shape Manipulation

A component‐wise propagation algorithm is proposed to automatically preserve the established inter‐relations while maintaining the defining characteristics of individual controllers and respecting the user‐specified modeling constraints in a hierarchical setup.

Functional programming for compiling and decompiling computer-aided design

A purely functional language for CAD called LambdaCAD and a polygon surface-mesh intermediate representation is defined and denotational semantics of both languages to 3D solids and a compiler from CAD to mesh accompanied by a proof of semantics preservation are defined.

Generative 3D models: a key to more information within less bandwidth at higher quality

A novel, yet extremely compact shape representation method that 3D shapes are represented in terms of functions instead of geometric primitives to provide a semantic level-of-detail superior to any low-level polygon reduction scheme.

Fusion 360 gallery

The Fusion 360 Gallery is presented, consisting of a simple language with just the sketch and extrude modeling operations, and a dataset of 8,625 human design sequences expressed in this language, and an interactive environment called the Fusion 360 Gym, which exposes the sequential construction of a CAD program as a Markov decision process, making it amendable to machine learning approaches.

ShapeMOD: Macro Operation Discovery for 3D Shape Programs

ShapeMOD is proposed, an algorithm which takes as input a collection of 3D shape programs and makes them more compact by automatically discovering common macros which can be re-used across the collection, and finds that generative models which train on refactored programs containing these macros produce more plausible output shapes than those trained on the original programs.