TITLE BLOCK DETECTION AND INFORMATION EXTRACTION FOR ENHANCED BUILDING DRAWINGS SEARCH

Oct 23, 2026

Upload by Fishy

The architecture, engineering, and construction (AEC) industry depends on information stored in building drawings, yet information extraction from title blocks is often slow and costly, especially for historical scans with poor quality and inconsistent standards. The work compares existing methods for title-block extraction, then proposes a novel detection and IE pipeline combining a lightweight CNN and GPT-4o. The pipeline achieves high-accuracy metadata extraction for both CAD and hand-drawn drawings, enabling faster drawing search. A deployed UI demonstrates significant time savings, and an extensible expert-annotated dataset is created via an AEC-friendly workflow.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

Problem context: historical building drawings

Title block as drawing metadata

Existing methods for title block extraction

Limitations of classical approaches

2025 European Conference on Computing in Construction CIB W78 Conference on IT in Construction Porto, Portugal July 14-17, 2025 TITLE BLOCK DETECTION AND INFORMATION EXTRACTION FOR ENHANCED BUILDING DRAWINGS SEARCH Alessio Lombardi 1 , Li Duan2 , Ahmed Elnagar 1 , Ahmed Zaalouk2 , Khalid Ismail2 , Edlira Vakaj2 1Buro Happold Ltd., London, United Kingdom 2Birmingham City University, Birmingham, United Kingdom Abstract The architecture, engineering, and construction (AEC) industry still heavily relies on information stored in drawings for building construction, maintenance, compliance and error checks. However, information extraction (IE) from building drawings is often time-consuming and costly, especially when dealing with historical buildings. Drawing search can be simplified by leveraging the information stored in the title block portion of the drawing, which can be seen as drawing metadata. However, title block IE can be complex especially when dealing with historical drawings which do not follow existing standards for uniformity. This work performs a comparison of existing methods for this kind of IE task, and then proposesa novel title block detection and IE pipeline which outperforms existing methods, in particular when dealing with complex, noisy historical drawings. The pipeline is obtained by combining a lightweight Convolutional Neural Network and GPT-4o, the proposed inference pipeline detects building engineering title blocks with high accuracy, and then extract structured drawing metadata from the title blocks, which can be used for drawing search, filtering and grouping. The work demonstrates high accuracy and efficiency in IE for both vector (CAD) and hand-drawn (historical) drawings. A user interface (UI) that leverages the extracted metadata for drawing search is established and deployed on real projects, which demonstrates significant time savings. Additionally, an extensible domain-expertannotated dataset for title block detection is developed, via an efficient AEC-friendly annotation workflow that lays the foundation for future work. Introduction Building drawings are a key element in the AEC industry. Historical drawings, in particular, are important for maintenance and refurbishment projects, although their quality is often poor and they are rarely aligned to modern conventions. Managing and retrieving information from technical drawings is time-consuming. Additionally, while several modern solutions have been proposed to address compliance checking on buildings and infrastructure, in many cases these processes still rely on exchange and manual commenting of drawings (Soliman-Junior et al. (2021)) . Therefore, information extraction (IE) applied to building drawings is still very important in the AEC industry: a fast and effective method to search through a large number of drawings is crucial in the daily workflow of engineers and Figure 1: An example of an engineering drawing demonstrating the different elements: title block (red), main drawing content (orange), legend (blue) and notes (green) . architects. However, drawing search is complex. Drawings aren’t archived with conventions that allow simple retrieval. This varies depending on the project, and it’s particularly true for building projects involving historical, pre-existing structures, such as refurbishment or extension projects. In these cases, the drawings are often only available as scans (of either computer-produced or hand-drawn drawings) and can be of poor quality. Searching specific drawings in such contexts can be a very tedious, manual task that can take several man-hours of work and is error-prone. On certain projects, in particular large refurbishment ones, a significant amount of the project budget (up to 40%) maybe dedicated to this task alone. An example is the Barbican refurbishment project in London, where upwards of 42 thousand hand-drawn historical drawings produced in the 1970s need to be scoured for

FAQs

Why is title block information extraction important in the AEC industry?

Because engineers rely on drawing metadata during construction, maintenance, compliance checks, and error verification. Extracting that metadata enables faster searching across large drawing archives.

What challenges make title block IE difficult for historical buildings?

Historical drawings are often scanned, of poor quality, and may not follow uniform standards. Variations in size, rotation, and layout increase detection and OCR errors.

How does the proposed pipeline improve performance over existing methods?

It combines a lightweight Convolutional Neural Network with GPT-4o. The approach detects engineering title blocks accurately and extracts structured metadata for use in drawing search, filtering, and grouping.

Oct 23, 2026

Upload by Fishy