Voxel Synthesis for Generative Design

Khokhlov, Matvey; Koh, Immanuel; Huang, Jeffrey

doi:10.1007/978-3-030-05363-5_13

Cited by 4 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides, other structured design representations include 3D voxelized geometry data and parametric design data. 3D voxelized geometries are represented as 3D grids of voxels, which typically take Boolean values (Khokhlov et al, 2019). Researchers also use the corresponding key design parameters often structured in a tabular form to represent a design (Kulfan, 2012).…”

Section: Design Representationsmentioning

confidence: 99%

“…Ideally, a design representation should capture the nature of the design to support formal and functional reasoning. The common design representation methods include images (Burnap et al, 2016), voxels (Khokhlov et al, 2019), point clouds (Park et al, 2019), meshes (Zhang et al, 2020), tabular parameters (Kulfan, 2012), design grammars (Stump et al, 2019), graphs (Cao et al, 2020), and so on. For the same design, different representations can capture the elements of the design and the interactions between pairs of elements in different ways.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Assessing Machine Learnability of Image and Graph Representations for Drone Performance Prediction

2022

View full text Add to dashboard Cite

Deep learning (DL) from various representations have succeeded in many fields. However, we know little about the machine learnability of distinct design representations when using DL to predict design performance. This paper proposes a graph representation for designs and compares it to the common image representation. We employ graph neural networks (GNNs) and convolutional neural networks (CNNs) respectively to learn them to predict drone performance. GCNs outperform CNNs by 2.6-8.1% in predictive validity. We argue that graph learning is a powerful and generalizable method for such tasks.

show abstract

Section: Design Representationsmentioning

confidence: 99%