Intelligent Detection of Mechanical, Electrical, and Plumbing (MEP) Metrics Based on 2D Floor Plans

Oct 28, 2026

Upload by Stanley

This research develops a neural-network model that extracts key Mechanical, Electrical, and Plumbing (MEP) information from 2D floor plans. It detects lighting symbols, classifies the appropriate light type, and extracts the associated text to support efficient floor design. The approach converts PDF floor plans into images, manually annotates a private dataset, converts annotations to COCO format, and trains a Mask R-CNN-based model. Reported results reach bbox_mAP 0.7596 and segm_mAP 0.7111, with strong performance at multiple IoU thresholds.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Related work

Technical approach

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000. Digital Object Identifier 10.1109/ACCESS.2023.0322000 Intelligent Detection of Mechanical, Electrical, and Plumbing (MEP) Metrics Based on 2D Floor Plans TARANDEEP SINGH MANDHIRATTA 1, ANK ZAMAN2,(Member, IEEE) and ABDUL-RAHMAN MAWLOOD-YUNIS3,(Member, IEEE) 1Dept. of Physics & Computer Science, Wilfrid Laurier University (WLU), Waterloo, ON, Canada ([mand4710@mylaurier.ca](mand4710@mylaurier.ca)) 2Dept. of Physics & Computer Science, Wilfrid Laurier University (WLU), Waterloo, ON, Canada (e-mail: [azaman@wlu.ca](azaman@wlu.ca)) 3Dept. of Physics & Computer Science, Wilfrid Laurier University (WLU), Waterloo, ON, Canada (e-mail: [amawloodyunis@wlu.ca](amawloodyunis@wlu.ca)) Corresponding author: ANK Zaman (e-mail: [azaman@wlu.ca](azaman@wlu.ca)). [ cs .CV] 14 Aug 2026 ABSTRACT This research developed a neural network-based model to extract various information from 2D floor plans. We detect lighting symbols, identify the appropriate type of light, and extract the associated texts with lights. The study aims to enable efficient floor designing and determining the number and typeof lights needed per floor, i.e., allow efficient design and estimate the power requirement of the floor plan. The model was developed using Mask RCNN as the base. The images were annotated and converted into a Coco data format for training the model. The model achieved bbox_mAP and segm_mAP values of 0.7596 and 0.7111, respectively. It also performed well at different IoU thresholds, i.e., with bbox_mAP 50 and segm_mAP 75 values of 0.9850 and 0.9219, respectively. The developed model will help various industries, such as architecture and construction, to improve design time and create efficient workflows by automatically detecting Mechanical, Electrical, and Plumbing (MEP) objects from floor plans, and it isthe first step towards building tools that will help energy-efficient building design. INDEX TERMS Intelligent detection, MEP metrices, 2D floor plans, Mask RCNN, Coco data format, Workflow efficiency, Energy efficiency, Engineering design process. I. INTRODUCTION TODAY’S technology has made collecting and process ing information in indoor environments possible. This has positively impacted facilities management, indoor localization, and indoor data models, but it has also increased the demand for creating actual databases of indoor information. One area still underdeveloped [1], [2] is the analysis of the Mechanical, Electrical, and Plumbing (MEP) matrices of buildings, which is crucial for efficient engineering design processes. Various approaches have been used to analyze MEP matrices of buildings, including those that use 3D laser scanning, CAD plans, architectural sketches, and mobile applications. This study focuses on using 2D floor plans, as they provide the most fundamental source of information for existing buildings. Recent efforts have been put into using machine learning models for extracting and labelling data from floor plans. However, this is a challenging problem, as the notation of different floor plans varies significantly according to the institutions and the design offices. Furthermore, floor plans archived as complex, fuzzy architectural drawings usually characterize raster images. arXiv :2608 . 14317v1 This research work aims to intelligently collect MEPrelated metrics from floor plans to aid in improving the efficiency of the engineering design process. As a first step, we developed a neural network model to collect three types of information from 2D floor plans: a) detect lighting symbols, b) identify the appropriate type of light, and c) extract the associated texts with lights. This work converted PDF files of 2D floor plans into image format to achieve these objectives. We annotated the images manually. This is because we use a private dataset provided by the sponsor, and the data isnot annotated. Once the annotations we

FAQs

What information does the model extract from 2D floor plans?

The model detects lighting symbols, identifies the appropriate type of light, and extracts associated texts tied to the lights.

How is the training data prepared for the proposed approach?

PDF floor plans are converted to images, the images are manually annotated, and the annotations are converted into COCO data format; images are also sliced into 1000×1000 grids with overlap.

Which model architecture is used and what performance metrics are reported?

Mask R-CNN is used as the base model. The study reports bbox_mAP 0.7596 and segm_mAP 0.7111, with additional results at bbox_mAP 50 and segm_mAP 75 IoU thresholds.

Oct 28, 2026

Upload by Stanley