SkeySpot: Automating Service Key Detection for Digital Electrical Layout Plans in the Construction Industry

Dhruv Dosi, Rohit Meena, Param Rajpura, Yogesh Kumar Meena · 2025

Read SkeySpot: Automating Service Key Detection for Digital Electrical Layout Plans in the Construction Industry by Dhruv Dosi, Rohit Meena, Param Rajpura, Yogesh Kumar Meena (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-10449.txt.

Introduction

Dhruv Dosi and co-authors build a tool that spots service key marks on power plans. The work dates to 2025. Old floor plans oft live just as scanned sheets. With no machine-read form, large reads turn slow and fault prone for build teams.

The team made a tagged set of forty-five scanned power plans. It holds near two and a half thousand marks in more than thirty classes. Past trained spot models were set on the set in a fair frame. The best model hit a strong mean score. Their light tool then finds, names, and counts the marks in real time. Its output feeds cost and care flows.

Catalog record

CREATORDhruv Dosi, Rohit Meena, Param Rajpura, Yogesh Kumar Meena
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE24 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-10449

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

SkeySpot: Automating Service Key Detection for Digital Electrical Layout Plans in the Construction Industry Dhruv Dosi, Rohit Meena, Param Rajpura, Yogesh Kumar Meena Abstract Legacy floor plans, often preserved only as scanned documents, remain essential resources for architecture, urban planning, and facility management in the construction industry. However, the lack of machine-readable floor plans render large-scale interpretation both time-consuming and error-prone. Automated symbol spotting offers a scalable solution by enabling the identification of service key symbols directly from floor plans, supporting workflows such as cost estimation, infrastructure maintenance, and regulatory compliance. This work introduces a labelled Digitised Electrical Layout Plans (DELP) dataset comprising 45 scanned electrical layout plans annotated with 2,450 instances across 34 distinct service key classes. A systematic evaluation framework is proposed using pretrained object detection models for DELP dataset. Among the models benchmarked, YOLOv8 achieves the highest performance with a mean Average Precision (mAP) of 82.5%. Using YOLOv8, we develop SkeySpot, a lightweight, open-source toolkit for real-time detection, classification, and quantification of electrical symbols. SkeySpot produces structured, standardised outputs that can be scaled up for interoperable building information workflows, ultimately enabling compatibility across downstream applications and regulatory platforms. By lowering dependency on proprietary CAD systems and reducing manual annotation effort, this approach makes the digitisation of electrical layouts more accessible to small and medium-sized enterprises (SMEs) in the construction industry, while supporting broader goals of standardisation, interoperability, and sustainability in the built environment. I INTRODUCTION Digital electrical layout plans are central to construction and facilities management workflows, representing the design and placement of electrical and heating, ventilation, and air conditioning (HVAC) components within construction environments [ 1 ] . These plans encode spatial and architectural information using standardised graphical symbols, referred as service keys, to represent components like sockets, switches, outlets, fixtures, etc. Generated via Computer-Aided Design (CAD) software, these diagrams are used to estimate costs, plan installations, coordinate maintenance and assess compliance. However, in practice, CAD files are often inaccessible to many stakeholders, particularly small and medium-sized enterprises (SMEs) in the construction industry, who instead rely on scanned floor plans or PDFs. The high cost and technical complexity of tools like AutoCAD 1 1 1 AutoCAD Floor Plan Software further limit their adoption in such contexts, creating a significant gap in accessible, automated annotation solutions. Manual detection and annotation of symbols in scanned plans remain the prevailing method but are slow, error-prone, and require domain expertise [ 1 ] . The task is further complicated by densely cluttered diagrams, visual occlusions, and subtle inter-class variations in symbol shapes, sizes, and orientations. These challenges motivate the need for robust, scalable, and easy-to-use automation tools that do not depend on proprietary design software. Fig. 1 : Proposed SkeySpot pipeline, outlining the workflow from dataset preparation and annotation to training, benchmarking, and deployment using Streamlit, an open-source Python framework Symbol spotting in architectural diagrams has been an open research topic over the past decade, especially in the context of scanned or rasterised floor plans [ 1 ] . To the best of our knowledge, there have not been significant contributions toward symbol detection and its real-time deployment for electrical layout plans. To address automatic symbol detection, prior non-deep learning approaches for symbol spotting have primarily relied on two methodological paradigms: pixel-based and vector-based techniques. For instance, [ 2 ] proposes a two-phase approach using low-level descriptors and template matching. Pixel-based methods such as F-signature [ 3 ] , Blurred Shape Model (BSM) [ 4 ] , and Circular BSM (CBSM) [ 5 ] operate on raster data, while vector-based approaches like SCIP and ESCIP [ 6 ] exploit vectorial representations. Structural methods leveraging graphs—such as Attributed Relational Graphs [ 7 ] , Region Adjacency Graphs [ 8 ] , graph serialization [ 9 ] , graph embedding [ 10 ] , and relational indexing [ 11 ] , attempt to capture the topological structure of symbols. Other works have used connected components, constituent primitives, and contour maps [ 12 ] . Despite their algorithmic sophistication, these approaches remain sensitive to noise, symbol distortion, and layout heterogeneity. Further, there are multiple symbols that are very similar in color and shape, that are not efficiently classified by rule-based methods [ 13 ] . In recent years, deep learning-based methods have shown notable progress in digital document analysis. Dey et al. [ 14 ] employ a simple LeNet for handwritten symbol recognition, while Riba et al. [ 15 ] utilize message-passing neural networks to operate on structural graphs. Semantic segmentation models have been applied to architectural diagrams for wall and door detection [ 16 , 17 ] . GAN-based methods have also been explored for symbol spotting [ 18 ] . However, many of these approaches are tested on simplified datasets and are not designed for real-time deployment in high-clutter environments. Notably, [ 19 , 20 ] utilize public datasets of about 1300 images, but these lack the symbol diversity and visual complexity encountered in practical settings. Additionally, previous works [ 21 , 22 ] also focus on piping and instrumentation diagrams using Faster R-CNN, which involves slower inference and is not optimised for deployment. Rezvanifar et al. [ 23 ] address only 12 architectural symbols and do not generalize to more complex electrical layouts. By combining dataset creation, model evaluation, and deployment in an accessible toolkit, this work fills a significant gap in automated symbol spotting for scanned electrical layouts and offers a robust baseline for further research in the automatic symbol spotting domain. The main contributions of this work are summarised as follows: 1. Introduced a novel dataset of 45 annotated floorplans, covering 34 unique service key classes and a total of 2450 instances across the service keys. 2. Benchmarked Faster R-CNN and YOLOv8 models fine-tuned for the service key detection and classification, offering a comprehensive performance evaluation of their strengths and limitations. 3. Developed the SkeySpot Interactive Interface Toolkit, a real-time web-based tool for small and medium-sized enterprises in the construction industry, automating service key spotting, detection summaries, and analysis to enhance decision-making and operational efficiency with an end-to-end scalable solution. II Materials and methods This section outlines the methodology used to create the dataset, design the model architecture, and train the deep learning model for detecting service keys in digitised electrical layout plans. An overview of the development and deployment framework for SkeySpot is illustrated in Figure 1 . II-A Dataset Preparation The Digital Electrical Layout Plans (DELP) dataset comprises 45 curated floor plans annotated with 34 distinct classes of service keys relevant to electrical systems. This dataset presents numerous challenges for automated symbol detection due to factors such as occlusion, clutter, color variations, and textual elements. The floor plans were obtained from UK-based SMEs engaged in housing estate development. Representative examples of symbol variability are shown in Figure 2 , which includes radiators with differing shapes and varying water entry positions, as well as low-energy pendant lights that, while similar in shape, differ in color. These intra-class variations and inter-class similarities add to the complexity of the classification task, particularly in instances where symbols overlap spatially. Each image in the dataset includes labeled annotations indicating the location and class of the service keys. A complete list of

Read the complete text