Architecture Engineers Award

  Architectural plan generation using pix2pix-series algorithms presents significant challenges, particularly regarding the absence of domain-specific evaluation standards and limited understanding of how training configurations jointly influence performance. Current architectural AI workflows focus heavily on visual output, often overlooking principle adherence, vectorization quality, and the structural logic of predicted plans. Addressing this gap, the proposed framework introduces a systematic experimental design and a dual-aspect evaluation approach specifically tailored for architectural-like plans, establishing a rigorous foundation for AI-enabled generative design in architecture. Limitations of Existing pix2pix Adaptations in Architectural Design Although pix2pix architectures are widely adopted for map translation and image-to-image tasks, their adaptation to architectural plan generation remains underdeveloped due to the lack of domain-centered evaluation benchmarks. A...

The evolution of architectural design is increasingly shaped by computational tools that support complex form exploration and performance-driven decision-making. Traditional form-finding methods are often limited by the designer’s manual iteration speed and the inability of existing machine learning models to generate structurally coherent, user-guided architectural forms. Emerging techniques, such as Stable Diffusion and Low-Rank Adaptation (LoRA), enable more efficient training and controlled generation of 3D morphological structures. By integrating heat-map-based geometry control with diffusion models, architects gain a powerful workflow for producing diverse design alternatives and achieving seamless transitions from conceptual forms to realistic renderings. This research situates itself within the broader framework of AI-assisted design and aims to expand the creative and technical capabilities of form-finding practices. Limitations of Existing Machine Learning Approaches in Form...

The rapid evolution of artificial intelligence has opened new possibilities for integrating deep learning into architectural workflows, particularly in the transformation of conceptual sketches into structured 3D models. This study introduces an intelligent modeling framework that merges architectural domain knowledge with advanced computer vision systems to overcome the long-standing challenge of the 2D-to-3D domain gap. By embedding the phased, selective, and iterative characteristics of architectural design directly into the workflow, the proposed model aligns AI-driven generation with the precision and logic expected in traditional design practices. This integration forms the foundation for a more adaptive and intelligent design-to-model pipeline within architecture. Deep Learning Integration in Architectural Design The research demonstrates how deep learning tools such as Stable Diffusion, CycleGAN, and Pixel2Mesh can be strategically aligned with architectural reasoning. These mo...

  Generative design has rapidly become a cornerstone of modern architectural and engineering innovation, enabling automated exploration of spatial layouts through computational intelligence. However, while these models efficiently generate diverse design options, they often lack integrated systems for evaluating performance metrics such as energy use, carbon impact, and spatial efficiency. This study bridges that gap by introducing a fully automated, image-based performance evaluation framework that accelerates the transition from concept to simulation-ready layouts. The approach redefines the workflow of early-stage design, providing instant sustainability insights that support low-carbon and high-performance outcomes. Image-Based Performance Evaluation Framework Traditional performance evaluation methods rely heavily on manual modeling, which is both time-consuming and prone to human error. To overcome this, the research introduces an automated image-to-simulation (Image2Sim)...