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

Tarandeep Singh Mandhiratta; ANK Zaman; Abdul-Rahman Mawlood-Yunis

cs.CVAug 14, 2026

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

Authors: Tarandeep Singh Mandhiratta, ANK Zaman, Abdul-Rahman Mawlood-Yunis

Organizations: Dept. of Physics & Computer Science, Wilfrid Laurier University (WLU), Waterloo, ON, Canada

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 type of 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 is the first step towards building tools that will help energy-efficient building design.

Figures & tables

Fig. 1: An architectural floor plan section, specifically a residential unit’s kitchen and bathroom areas [ 6 ]

Fig. 3: Neural Network Model Development Methodology

Fig. 4: Data Collection and Preprocessing

Fig. 7: Sliced Component of the Original Image

Model	Advantages	Disadvantages
Mask R-CNN [ 3 ] , [ 4 ]	Object detection and instance segmentation in a single model, with high accuracy, end-to-end training, open-source implementation	Computationally expensive, slower inference speed compared to other models
Faster R-CNN [ 33 ]	Good accuracy, faster inference speed compared to Mask R-CNN	Requires twostage detection, may miss small objects
YOLO [ 33 ]	Fast inference speed, real-time object detection	Lower accuracy compared to other models, may miss small objects
SSD [ 34 ]	Fast inference speed, real-time object detection, good accuracy for small objects	Lower accuracy compared to other models for larger objects

TABLE I: Object Detection Model Comparison

Fig. 10: Training Accuracy and Loss Graphs

Epoch	Learning Rate	Training Time	Memory	Loss_rpn_cls	Loss_rpn_bbox	Loss_cls	Accuracy	Loss_bbox	Loss_mask	Loss
1	2.500e-03	0:40:42	2543	0.1327	0.0245	0.3299	86.1719	0.2994	0.5616	1.3480
10	2.500e-04	0:10:28	2904	0.0019	0.0125	0.0716	97.0508	0.1685	0.1259	0.3804
20	2.500e-05	0:00:02	2904	0.0014	0.0105	0.0570	97.5586	0.1559	0.1161	0.3409

TABLE II: Training Metrics

Fig. 11: Segmentation Evaluation Results

Fig. 12: Evaluation Results for Epoch 20

Fig. 13: Detected Lighting Symbols in Test Image

Fig. 14: Extraction of Lighting Symbols and Associated Text

Object Detection

Object detection locates and classifies objects in images or video, balancing localization accuracy, robustness, and inference speed.

See all 624 papers

SceneAligner: 3D-Grounded Floorplan Localization in the Wild
May 21, 2026Junhyeong Cho, Ruojin Cai, Hadar Averbuch-ElorFloor-Plan-Based Visual Localization Map-Based Localization
Cornell University · Kempner Institute, Harvard University
Benchmarking Deep Learning Approaches for AEC Engineering Drawing Layout Detection and Information Extraction
Jul 21, 2026Tianyang Huang, Alessio Lombardi, Ahmed Elnagar +7Document Understanding Document Information Extraction
Buro Happold Ltd., London, United Kingdom · Birmingham City University, Birmingham, United Kingdom
When Should a Network Emit Geometry, and When Should It Detect It? Readout, Reconciliation, and Representation in Floorplan Vectorization
Aug 26, 2026He ZhangIndoor Scene Reconstruction Autoregressive Decoding
Independent Researcher

May 21, 2026J/K move · Enter open · S save

SceneAligner: 3D-Grounded Floorplan Localization in the Wild

Junhyeong Cho, Ruojin Cai, Hadar Averbuch-Elor

Cornell University · Kempner Institute, Harvard University

Floor-Plan-Based Visual Localization Map-Based Localization

Open paper AskPDF arXiv

Many public buildings provide floorplans with a "you are here" indicator to help visitors orient themselves. Floorplan localization seeks to computationally replicate this capability by determining where visual observations were captured within a floorplan. However, existing methods typically assume controlled small-scale environments and precise vectorized floorplans, limiting their ability to operate in large-scale buildings and rasterized floorplans. In this work, we present an approach for performing floorplan localization in the wild by grounding the task in a reconstructed 3D representation of the scene. Given an unconstrained image collection, our method reconstructs a gravity-aligned 3D scene and projects it into a 2D density map that serves as a floorplan proxy. Floorplan localization is then formulated as aligning this proxy with the input floorplan via a 2D similarity transform. To bridge the appearance gap between density maps and architectural floorplans, we adapt a 2D foundation model to learn cross-modal correspondences, introducing a fine-tuning scheme that encourages semantically aligned matches while preserving structural consistency. Extensive experiments demonstrate substantial improvements over prior methods, including in extremely sparse settings with as little as a single input image. Our code and data will be publicly available.

Abstract

Object Detection

Explore similar work

SceneAligner: 3D-Grounded Floorplan Localization in the Wild

Benchmarking Deep Learning Approaches for AEC Engineering Drawing Layout Detection and Information Extraction

When Should a Network Emit Geometry, and When Should It Detect It? Readout, Reconciliation, and Representation in Floorplan Vectorization