Based on Research From: Case Study 1 of the Master's Thesis, "Structural Embodiment - Unified Workflow and Toolkit for Form-finding, Materialisation and Visualisation via Deep Learning Methods" by Tao Sun, conducted under the professorship of Structural Design and Chair of Architectural Informatics at the Technical University of Munich.

This LoRA model is specifically trained on a dataset comprising 100 varied renderings produced during the first case study of the aforementioned thesis. These renderings serve as the foundational dataset, facilitating the model's ability to generate model-like views of bridge structures with high fidelity.

Optimal Settings:

LoRA weight: 0.8

Depth ControlNet Unit Weight: 0.6

Canny ControlNet Unit Weight: 0.3

Utilising both Depth and Canny ControlNet Units simultaneously with the specified weights enhances the model's effectiveness, producing detailed and context-aware visualisations of bridge structures.

You should use physical model to trigger the image generation.

You should use arch bridge to trigger the image generation.

You should use suspension bridge to trigger the image generation.

You should use concrete to trigger the image generation.

You should use site to trigger the image generation.

You should use timber to trigger the image generation.

You should use wire trees to trigger the image generation.

You should use dark acrylic to trigger the image generation.

You should use reflection to trigger the image generation.

You should use studio lighting to trigger the image generation.

You should use SETLKT to trigger the image generation.

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.