Skip to content

Research

MSc dissertation

The dissertation benchmarks three YOLO-seg models and a U-Net (ResNet-50) for crack segmentation, and an FPN–EfficientNet-B4 for 18-class defect segmentation on DACL10k. The models are validated on industry site imagery and set against a 21-respondent practitioner survey. U-Net gave the most accurate crack masks, YOLO11x-seg the most robust field detection, and the multi-class model broader coverage, which together support a hybrid inspection pipeline integrated with BIM and digital twins.

September 2025

Towards Digital Transformation of Structural Health Monitoring (SHM): Comparative YOLO and U-Net Deep Learning for Multi-Defect Detection

Institution
Belfast School of Architecture and the Built Environment, Ulster University
Supervisor
Prof. Ibrahim Motawa
Industry partner
Amphora Consulting, Belfast (689 site images and professional input)
Module result
74% (60-credit dissertation module)
Publication status: A journal manuscript based on this work, focused on field robustness, is in submission (Built Environment Project and Asset Management, Emerald).
Read the case study →

Practitioner survey

21 professionals · UK · Saudi Arabia · UAE · Egypt

had used CV tools for inspection
9%
had used CV tools for inspection
open to adopting AI tools
≈75%
open to adopting AI tools
typical inspection, medium building
4–7 days
typical inspection, medium building

Source: dissertation §4.1

YOLO-seg on the Crack-Seg test split

Box vs mask mAP50. Box localisation is strong; pixel masks are harder.

0.000.250.500.751.000.807YOLOv8x-seg · Box mAP50: 0.8070.674YOLOv8x-seg · Mask mAP50: 0.674YOLOv8x-seg0.804YOLO11x-seg · Box mAP50: 0.8040.639YOLO11x-seg · Mask mAP50: 0.639YOLO11x-seg0.792YOLO11n-seg · Box mAP50: 0.7920.658YOLO11n-seg · Mask mAP50: 0.658YOLO11n-seg
  • Box mAP50
  • Mask mAP50
View data table
CategoryBox mAP50Mask mAP50
YOLOv8x-seg0.8070.674
YOLO11x-seg0.8040.639
YOLO11n-seg0.7920.658

Source: MSc dissertation (Ulster University, 2025), Table 6

Throughput per configuration

Seconds per image on an RTX 4060 (lower is faster).

0.000.651.301.952.60Crack U-Net0.53Crack U-Net · s / image: 0.53Crack-only YOLO0.77Crack-only YOLO · s / image: 0.77Multi-class segmentation2.04Multi-class segmentation · s / image: 2.04Full classification + report2.38Full classification + report · s / image: 2.38
View data table
Categorys / image
Crack U-Net0.53
Crack-only YOLO0.77
Multi-class segmentation2.04
Full classification + report2.38

Source: MSc dissertation (Ulster University, 2025), Table 8

Education

Degrees

2024 – 2025

MSc Digital Construction Analytics & Building Information Modelling

Ulster University, Belfast

Distinction

Module marks
ModuleMark
Introduction to Data Science81%
Industry Project (AECOM)80%
Data Validation & Visualisation75%
Research Design & Dissertation74%
Building Information Modelling73%
Digital Construction: Technology, Strategy & Management67%
Business Intelligence & Analytics63%

Source: Ulster University statement of academic record. Distinction threshold 70%.

2018 – 2023

BSc Civil Engineering

Higher Technological Institute, 10th of Ramadan, Egypt

GPA 3.24 / 4.00 · Graduation project A+ (steel structures)

Certifications & memberships

Continuing development