Architecture Engineers Award

Building Information Modeling (BIM) has become a central workflow in the architecture, engineering, and construction (AEC) industry, yet 2-dimensional (2D) architectural drawings remain indispensable for documentation, communication, and construction execution. Despite BIM’s advantages, generating 2D drawings from BIM models still requires substantial manual effort, with current practices demanding significant additional time for conversion. This research addresses this inefficiency by proposing an automated, deep learning–based framework capable of transforming BIM models into accurate and stylistically diverse 2D architectural drawings. Persistent Role of 2D Drawings in BIM-Based Workflows Although BIM offers rich parametric and information-embedded models, 2D drawings continue to dominate regulatory submissions, on-site coordination, and professional communication. The reliance on manual or semi-automated conversion processes introduces redundancy and inefficiency into BIM workfl...

  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...

  Data scarcity in the construction domain limits the performance of deep neural networks, especially for computer vision tasks such as object detection and activity recognition. While synthetic data offers a scalable alternative to real-world data collection, its lack of photorealism often results in a noticeable “reality gap,” reducing the effectiveness of trained models in practical scenarios. Addressing this challenge requires innovative frameworks that can enhance the visual realism and diversity of synthetic datasets while retaining precise annotations. Synthetic Data Generation Challenges in Construction Generating high-quality synthetic datasets for construction environments presents several challenges, including complex lighting conditions, occlusions, material textures, and dynamic scenes involving workers and machinery. Traditional 3D rendering techniques fail to fully capture these contextual details, leading to unrealistic visual outputs. The lack of domain-specific re...

Thermal bridges in building envelopes are localized weaknesses that significantly affect energy performance and occupant comfort. These areas of elevated heat transfer can increase energy consumption by up to 40 %, highlighting the need for precise identification and mitigation strategies. Recent advancements in remote sensing and machine learning offer promising avenues for non-destructive monitoring of material performance in urban buildings. By integrating aerial multi-modality imaging, which combines visible, thermal, and LiDAR data, researchers can achieve real-time, city-scale diagnostics, providing a foundation for energy-efficient retrofitting and sustainable urban development. Development of a Deep Learning Framework This study presents a customized deep learning (DL) approach based on YOLOv9, tailored for thermal bridge detection using a multimodal dataset of 6,927 images. By leveraging the complementary strengths of visible, thermal, and LiDAR data, the model effectively ...