StructDamage:A Large Scale Unified Crack and Surface Defect Dataset for Robust Structural Damage Detection
Authors: Misbah Ijaz, Saif Ur Rehman Khan, Abd Ur Rehman, Sebastian Vollmer, Andreas Dengel, Muhammad Nabeel Asim
Organizations: Department of Computer Science, University of Gujrat, Gujrat, 51700, Pakistan. · Department of Computer Science, Rhineland-Palatinate Technical University of Kaiserslautern-Landau, orgaddressKaiserslautern, 67663, Germany. · BiogentX (biogentx.com, Renala Khurd, District Okara, Punjab, Pakistan. · German Research Center for Artificial Intelligence, Kaiserslautern, 67663, Germany. · Intelligentx GmbH (intelligentx.com), Kaiserslautern, Germany. · Department of Core Informatics, Graduate School of Informatics ,Osaka Metropolitan University, Saka, 599-8531, Japan.
Automated detection and classification of structural cracks and surface defects is a critical challenge in civil engineering, infrastructure maintenance, and heritage preservation. Recent advances in Computer Vision (CV) and Deep Learning (DL) have significantly improved automatic crack detection. However, these methods rely heavily on large, diverse, and carefully curated datasets that include various crack types across different surface materials. Many existing public crack datasets lack geographic diversity, surface types, scale, and labeling consistency, making it challenging for trained algorithms to generalize effectively in real world conditions. We provide a novel dataset, StructDamage, a curated collection of approximately 78,093 images spanning nine surface types: walls, tile, stone, road, pavement, deck, concrete, and brick. The dataset was constructed by systematically aggregating, harmonizing, and reannotating images from 32 publicly available datasets covering concrete structures, asphalt pavements, masonry walls, bridges, and historic buildings. All images are organized in a folder level classification hierarchy suitable for training Convolutional Neural Networks (CNNs) and Vision Transformers. To highlight the practical value of the dataset, we present baseline classification results using fifteen DL architectures from six model families, with twelve achieving macro F1-scores over 0.96. The best performing model DenseNet201 achieves 98.62% accuracy. The proposed dataset provides a comprehensive and versatile resource suitable for classification tasks. With thorough documentation and a standard structure, it is designed to promote reproducible research and support the development and fair evaluation of robust crack damage detection approaches.
Figures & tables
Author
Year
Country
Dataset Name
Surface
Resolution
Images
Karimi et al. (2024)
2024
Iran & Portugal
Crack in Historic Buildings
Masonry materials
3456 × 3456
4374
Özgenel and Sorguç (2018)
2018
Turkey
METU Concrete Crack
Concrete
227 x 227
40,000
Guzmán-Torres et al. (2023)
2023
Mexico
CONCORNET2023
Reinforced concrete
Variable
790
Dorafshan et al. (2018)
2018
USA
SDNET2018
Concrete
256 x 256
56,000
Xu et al. (2022)
2022
China
CrackNJ156
Pavement
512 × 512
156
Guo et al. (2023)
2023
United States
CrackSC
Pavement
320 × 480
197
Table 1: Comparison of existing datasets with total samples and unique classes
Figure 1: Distribution of calss count categories of source datasets
Figure 2: StructDamage dataset construction workflow
Figure 3: Number of images per source dataset included in StructDamage dataset
Figure 4: Geographic distribution of source datasets included in StructDamage dataset
Source datasets
Collection methods
Captures devices
Karimi et al. (2024) ; Özgenel and Sorguç (2018) ; Guzmán-Torres et al. (2023) ; Dorafshan et al. (2018) ; Xu et al. (2022) ; Guo et al. (2023) ; Sabouri and Sepidbar (2023) ; Long et al. (2025) ; Garita-Durán et al. (2025) ; Ye et al. (2021) ; Flotzinger et al. (2024) ; Benz and Rodehorst (2022) ; Oluwaseun (2023) ; Mundt et al. (2019) ; Oliveira and Correia (2014) ; Xu et al. (2019)
Ground survey
Smartphone camera, DSLR camera, digital camera, area array CCD camera, drones
Eisenbach et al. (2017) ; Stricker et al. (2019) ; Alkalbani et al. (2025) ; Li et al. (2024) ; Maeda et al. (2018) ; Maeda et al. (2021) ; Zou et al. (2012)
Vehicle mounted
CCD cameras, Smartphone (Android)
Dai et al. (2025)
CCTV / borehole camera
CCTV camera
Katsigiannis et al. (2023)
Ground survey + online sources
Smartphone & Manufacturer websites
Ren et al. (2024)
Street view platform
Street-view camera
Bai et al. (2021)
Post event survey
Smartphone / DSLR
Table 2: Licensing framework and redistribution terms for source datasets
Source datasets
Annotation type
Number of datasets
Karimi et al. (2024) ; Özgenel and Sorguç (2018) ; Dorafshan et al. (2018) ; Dai et al. (2025) ; Katsigiannis et al. (2023) ; Xu et al. (2019) ; Oluwaseun (2023)
Image level label
7
Guzmán-Torres et al. (2023) ; Eisenbach et al. (2017) ; Stricker et al. (2019) ; Long et al. (2025) ; Ren et al. (2024) ; Alkalbani et al. (2025) ; Li et al. (2024) ; Maeda et al. (2018) ; Maeda et al. (2021) ; Arya et al. (2021) ; Arya et al. (2024)
Bounding box annotations
11
Xu et al. (2022) ; Guo et al. (2023) ; Zou et al. (2012) ; Garita-Durán et al. (2025) ; Ye et al. (2021) ; Benz and Rodehorst (2022) ; Goo et al. (2024) ; Oliveira and Correia (2014)
Table 3: Summary of annotation format used across the source datasets
Figure 5: Annotation types across the source datasets
License type
Redistribution terms
Source datasets
CC BY 4.0
Attribution required; commercial and derivative use permitted
Özgenel and Sorguç (2018) ; Dorafshan et al. (2018) ; Xu et al. (2022) ; Sabouri and Sepidbar (2023) ; Katsigiannis et al. (2023) ; Ren et al. (2024) ; Li et al. (2024) ; Arya et al. (2024)
CC BY-NC 3.0
Attribution required; non-commercial use only
Arya et al. (2021) ; Flotzinger et al. (2024)
CC BY-SA 4.0
Attribution required; commercial and derivative use permitted; redistribution allowed under the same license
Oluwaseun (2023)
CC BY-NC-SA 4.0
Attribution required; non commercial use only; derivative works permitted; redistribution under the same license (ShareAlike)
Garita-Durán et al. (2025) ; Gao and Mosalam (2020)
MIT license
Attribution required (license notice); commercial use, modification, and redistribution permitted with no copyleft requirement.
Guzmán-Torres et al. (2023) ; Guo et al. (2023) ; Bai et al. (2021) ; Maeda et al. (2018) ; Maeda et al. (2021)
Apache-2.0 license
Attribution required (license and NOTICE file); commercial use, modification, and redistribution permitted; patent rights granted.
Alkalbani et al. (2025)
Table 4: Licensing framework and redistribution terms for source datasets
Figure 6: Preprocessing operations
Figure 7: Class distribution of StructDamage dataset
Figure 8: Balanced data distribution
Methods
Accuracy (%)
Precision (%)
Recall (%)
F1-Score (%)
DenseNet121
98.4300
0.98326
0.98378
0.98339
DenseNet169
98.4300
0.98286
0.98334
0.98302
DenseNet201
98.6203
0.98526
0.98580
0.98534
EfficientNetV2B0
96.2654
0.96060
0.96047
0.96000
EfficientNetV2B1
93.9105
0.93746
0.93820
0.93670
MobileNetV2
98.3349
0.98220
0.98236
0.98214
Table 5: Baseline classification results on StructDamage (balanced validation subset)
Cracks are a critical indicator of building health, and early stage identification is fundamental to prevent harmful damages. Advances in deep learning (DL), particularly convolutional neural networks (CNNs), have enabled scalable solutions for automated crack detection. However, CNN performance strongly depends on the availability of large and diverse datasets, which is particularly challenging for complex surfaces such as masonry. Collecting sufficient real data is time-consuming, while publicly available datasets may not be adequate. To address this limitation, we explored generating synthetic crack data, which complements real data and improves training effectiveness. The real dataset consists of masonry crack images collected from buildings in Bologna and surrounding areas. In contrast, the synthetic dataset was generated using a crack overlay tool that adds cracks to background images in a controlled orientation and placement. The real dataset was used to train several DL architectures, to identify the best-performing model (InceptionV4) employed for experiments with generated data. Six training scenarios were tested in InceptionV4 by varying the ratio of real and synthetic data, with evaluation performed on a test set composed of real images using the F1-score and mean Intersection over Union (mIoU) metrics. Results show that training on synthetic data plus a modest addition of 20% real data achieves results comparable to training on real data only. Moreover, the 20/80 scenario (synthetic/real) achieved an 76% F1-score and 80% mean IoU, outperforming the real-only case. As can be seen, the method demonstrates the potential of synthetic data to reduce collection efforts while enhancing crack detection accuracy.
Mattia Forlesi, Alfonso Esposito, Ivan Zyrianoff +2
Department of Computer Science and Engineering, University of Bologna, Bologna, Italy · Department of Civil, Chemical, Environmental, and Materials Engineering, University of Bologna, Bologna, Italy
Structural damage detection is essential for maintaining the safety and reliability of civil infrastructure. However, accurately identifying different types of structural damage from images remains challenging due to variations in damage patterns and environmental conditions. To address these challenges, this paper proposes MS-SSE-Net, a novel deep learning (DL) framework for structural damage classification. The proposed model is built upon the DenseNet201 backbone and integrates novel multi-scale feature extraction with channel and spatial attention mechanisms (MS-SSE-Net). Specifically, parallel depthwise convolutions capture both local and contextual features, while squeeze-and-excitation style channel attention and spatial attention emphasize informative regions and suppress irrelevant noise. The refined features are then processed through global average pooling and a fully connected classification layer to generate the final predictions. Experiments are conducted on the StructDamage dataset containing multiple structural damage categories. The proposed MS-SSE-Net demonstrates superior performance compared with the baseline DenseNet201 and other comparative approaches. Specifically, the proposed method achieves 99.31% precision, 99.25% recall, 99.27% F1-score, and 99.26% accuracy, outperforming the baseline model which achieved 98.62% precision, 98.53% recall, 98.58% F1-score, and 98.53% accuracy.
Saif ur Rehman Khan, Imad Ahmed Waqar, Arooj Zaib +4
Department of Computer Science, Rhineland-Palatinate Technical2026 University of Kaiserslautern-Landau, orgaddressKaiserslautern, 67663, Germany.Apr · 4BiogentX (biogentx.com, Renala Khurd, District Okara, Punjab, Pakistan. · 2German Research Center for Artificial Intelligence, Kaiserslautern, 67663, Germany. +2
Automated structural health monitoring is essential to prevent catastrophic infrastructure failures. Precise, pixel-level defect segmentation is needed to accurately assess structural integrity, but progress in defect segmentation for civil infrastructures has been held back by an extreme scarcity of data, which requires costly expert annotation. The need for data is accentuated by algorithmic hurdles intrinsic to the problem, including center-bias and the need to rely more on shape when inspecting nearly textureless building materials. To remove the bottleneck, we introduce Cracks in the Foundation (CiF), the largest and most detailed civil infrastructure (instance) segmentation dataset to date, comprising ≈150,000 high-resolution images meticulously curated over five years in collaboration with civil engineering experts. With the help of this unprecedented data source, we expose a blind spot of current visual AI: despite the advent of promptable Foundation Models (FMs) and Vision Language Models (VLMs), and despite the impressive abilities of today's specialised segmentation models, it turns out that dense image understanding in the built environment is nowhere near solved. Our evaluations indicate that even the most recent zero-shot FMs face significant challenges when deployed on real-world infrastructure and even the performance of specialised models with domain-specific supervision plateaus at ≈25% mAP. CiF establishes inspection of civil infrastructure, an elementary and seemingly easy perceptual task, as an open challenge that reveals fundamental weaknesses of present-day models trained predominantly on internet images, literally and figuratively highlighting cracks in the current foundation model paradigm.
Nicola Farronato, Niccolo Avogaro, Thomas Frick +6