LGAICVJan 1, 2024

An attempt to generate new bridge types from latent space of generative adversarial network

arXiv:2401.00700v15 citationsh-index: 4
Originality Synthesis-oriented
AI Analysis

This work addresses bridge design innovation for civil engineers, though it appears incremental as it applies existing GAN methods to a new dataset.

The researchers tackled the problem of generating novel bridge designs by training a generative adversarial network on symmetric bridge images, resulting in the creation of new asymmetric bridge types through latent space sampling.

Try to generate new bridge types using generative artificial intelligence technology. Symmetric structured image dataset of three-span beam bridge, arch bridge, cable-stayed bridge and suspension bridge are used . Based on Python programming language, TensorFlow and Keras deep learning platform framework , as well as Wasserstein loss function and Lipschitz constraints, generative adversarial network is constructed and trained. From the obtained low dimensional bridge-type latent space sampling, new bridge types with asymmetric structures can be generated. Generative adversarial network can create new bridge types by organically combining different structural components on the basis of human original bridge types. It has a certain degree of human original ability. Generative artificial intelligence technology can open up imagination space and inspire humanity.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes