Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Data-driven Detection and Evaluation of Damages in Concrete Structures: Using Deep Learning and Computer Vision

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that YOLO-v7 instance segmentation detects cracks and spalls in concrete more accurately and faster than Mask R-CNN, reporting 96.1% mAP@0.5 at 40 FPS versus 92.1% at 18 FPS, and recommends YOLO-v7 for real-time…

desk verdict Routine benchmark with a likely split-after-augmentation leak; the specific mAP numbers are not reliable until the protocol and inconsistent metrics are fixed. read the letter →

arxiv 2501.11836 v1 pith:4V427PVR submitted 2025-01-21 cs.CV cs.AIcs.LG

Saeid Ataei , Saeed Adibnazari , Seyyed Taghi Ataei This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords concretestructuresdamagedetectiondeeplearninginstancesegmentationYOLO-v7MaskR-CNNcrackstructuralhealthmonitoring
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper aims to show that a modern one-stage instance-segmentation network can automate the detection of cracks and spalls in concrete surfaces well enough for real-time structural health monitoring. Using 400 annotated images expanded by augmentation to 10,995, it trained two models and reports that YOLO-v7 instance segmentation reaches 96.1% mean average precision at 40 frames per second, while Mask R-CNN reaches 92.1% at 18 frames per second. The authors conclude that YOLO-v7 is the better choice for live, high-throughput inspection and that Mask R-CNN remains an option for offline analysis. If the comparison holds, the practical payoff is automated, low-cost, continuous monitoring of bridges, tunnels, and walls.

What carries the argument

The carrying object is instance segmentation, a vision task that labels each individual defect with its own pixel mask rather than only classifying whole images; this is what lets the system localize separate cracks and spalls. The comparison is carried by a fixed experimental pipeline: combining three public datasets into 400 images, expanding them to 10,995 via geometric and color augmentations plus model-specific transforms, fine-tuning both models from pre-trained weights, and scoring with mAP@0.5 and FPS. The mAP numbers are the bridge between model choice and deployment recommendation.

What would settle it

A reader could rerun the same training protocol with the 400-image set split into training and test before any augmentation, then augment only the training side. If YOLO-v7's mAP@0.5 falls well below 96.1% or the gap with Mask R-CNN shrinks, the reported superiority is an artifact of data leakage rather than a property of the model.

Watch

Extended reading notes

Core claim

The authors' central discovery is that, on their curated concrete-damage dataset, YOLO-v7 instance segmentation outperforms Mask R-CNN on every headline metric: mAP@0.5 of 96.1% versus 92.1%, precision of 94.9% versus 92.0%, recall of 94.3% versus 62.8%, and inference speed of 40 FPS versus 18 FPS. They interpret this as evidence that one-stage instance segmentation is the right architecture for real-time structural health monitoring, especially for catching spalls (mAP 99.5%) and, to a lesser degree, cracks (mAP 92.7%). The paper also reports qualitative generalization tests on random internet photos and videos, which the authors take as further evidence of robustness.

Load-bearing premise

The load-bearing premise is that the 90/10 training/test split of the augmented dataset keeps each original concrete image on one side only; if augmented copies of the same crack appear in both training and test sets, the reported mAP mostly measures recognition of near-duplicates rather than true generalization.

Editorial extensions

If this is right

  • YOLO-v7 instance segmentation can process concrete inspection video at 40 frames per second, so it could run on drone or wall-climbing camera feeds without waiting for offline batch analysis.
  • Mask R-CNN's recall of 62.8% on this dataset implies that roughly a third of defects would be missed in a live setting, making it a poorer fit for safety-critical screening despite its 92.1% mAP.
  • Both models detect spalls more reliably than cracks, so crack recall (88.6% for YOLO-v7) is the bottleneck that future datasets and architectures should target.
  • Transfer learning from a large pre-trained vision model allows a domain dataset of only 400 source images to reach near-96% detection performance after augmentation.
  • Deploying YOLO-v7 on edge devices with IoT cameras would enable on-site, real-time damage assessment without sending images to a central server, a direct corollary of the reported speed.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The reported accuracy is likely to be an upper bound on real-world generalization: the 90/10 split was applied after augmentation, so augmented copies of the same original image can appear in both training and test sets; a split-before-augmentation retest would reveal how much of the score comes from near-duplicate recognition.
  • If the speed-accuracy gap persists on a held-out, pre-augmentation test set, the practical consequence extends beyond concrete inspection: any instance-segmentation task with small annotated datasets and real-time requirements should prefer carefully tuned one-stage detectors over two-stage ones.
  • A natural testable extension is to add fine-crack classes at higher imaging resolution, because the paper's own limitation discussion predicts YOLO-v7 will lose accuracy on extremely fine cracks as resolution is constrained.
  • The qualitative internet-photo and video tests suggest some out-of-distribution robustness, but a quantitative evaluation with established crack benchmarks would be needed to claim the model generalizes across cameras and lighting.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. This paper benchmarks two instance segmentation models, Mask R-CNN and YOLO-v7, on a combined dataset of 400 concrete crack/spall images augmented to 10,995 images. The models are trained with transfer learning and evaluated with precision, recall, mAP@0.5, and FPS. The central claim is that YOLO-v7 achieves higher mAP@0.5 (96.1%) and higher throughput (40 FPS) than Mask R-CNN (92.1%, 18 FPS), making it preferable for real-time structural health monitoring. The paper also discusses limitations and future directions.

Significance. If the quantitative comparison is valid, the result is a useful practical benchmark for infrastructure inspection: it aligns with the general observation that one-stage detectors such as YOLO-v7 offer better speed-accuracy trade-offs than two-stage Mask R-CNN on similar segmentation tasks. The manuscript's strengths are the use of three publicly available datasets, the comparison of two standard architectures, and the explicit acknowledgment of limitations (Section 3.11). However, the validity cannot currently be assessed due to the unclear data-split point, inconsistent dataset counts, contradictory metric tables, and unreported statistical variability; these issues must be resolved before the comparison can be used to support the recommendation.

major comments (4)
  1. [§2.1–2.2, Table 1] The manuscript does not state whether the 90/10 split was performed on the 400 original images before augmentation or on the augmented set. Because the augmentation pipeline generates many near-duplicate variants of each source image, a post-augmentation split can place sibling images of the same crack in both training and test partitions; with 400 source images and a 1,000-image test partition, this is a concrete risk that would inflate mAP by near-duplicate retrieval. Additionally, the reported counts are inconsistent: the text states 10,995 augmented images, while Table 1 sums to 10,000 (9,000 + 1,000). Please specify the split point and reconcile the counts; if the split was post-augmentation, re-run the comparison with a split performed before augmentation.
  2. [§3.2–3.3, Tables 6, 7 and 9] The mAP comparison for Mask R-CNN is internally inconsistent. Table 9 reports Mask R-CNN mAP@0.5 = 92.1%, but Table 7 gives class-specific AP values of 37.3% (crack) and 79.3% (spall), whose average cannot equal 92.1%; Table 6 also reports a 92.1% 'classification' AP50 alongside a 95.5% bounding-box AP50, without defining how these relate to the mAP used in Table 9. Please clarify which metric is reported and recompute the class-averaged mAP, since the central accuracy comparison rests on this number.
  3. [§3.3 and §3.7, Tables 8 and 11] All accuracy and speed metrics appear to be single-run measurements with no error bars or number of repeated trials. The 4-point mAP gap between YOLO-v7 and Mask R-CNN may be within run-to-run variance for models trained with stochastic gradient descent; at minimum, report metrics across multiple seeds with means and standard deviations, and state the number of runs.
  4. [§3.7, Tables 9 and 11] The FPS comparison is not reproducible as reported because the manuscript does not state the GPU model, batch size at inference, input resolution after preprocessing (the text says all images resized to 640×640 but does not state whether this was used at test time for both models), or the measurement protocol for FPS. Please provide this information so the speed advantage can be independently verified.
minor comments (4)
  1. [Title] The title contains a typographical error: 'DATA-D RIVEN' should be 'DATA-DRIVEN'.
  2. [§3.4, second bullet] The statement 'Mask R-CNN: Superior precision for crack classification' is not supported by the tables: Table 10 shows lower overall precision for Mask R-CNN (92.0% vs 94.9%), and no crack-specific precision for Mask R-CNN is provided in Table 7.
  3. [§2.1, Table 1] The phrase 'Validation, Test' is ambiguous; it is unclear whether the 10% is a separate test set or a combined validation/test partition, and the paper does not state how the reported test metrics were computed.
  4. [§2.2, Tables 2 and 3] The word 'Cutting' in Table 2 should be 'Cropping', and it is unclear whether Mask R-CNN received any model-specific augmentation or only the Roboflow augmentation; please clarify for a fair comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical benchmark; the only concern is a data-leakage risk from the post-augmentation split, not a circular derivation.

full rationale

This paper is an empirical benchmark comparing YOLO-v7 instance segmentation with Mask R-CNN on a concrete damage dataset. The central claims—mAP@0.5 of 96.1% at 40 FPS versus 92.1% at 18 FPS—are measured experimental results, not quantities defined in terms of one another. No parameter is fitted to a subset and then reported as a prediction of a closely related quantity, and no load-bearing appeal is made to the authors' own prior theorems or uniqueness results. The only near-circularity concern is the dataset-handling description in Section 2.1: the paper says the 400 images were 'expanded through augmentation to 10,995 images' and then gives the 9,000/1,000 split in Table 1, which suggests augmentation occurred before the train/test split. If so, augmented siblings of the same original image could appear in both training and test partitions, making the test set partially a near-duplicate retrieval task rather than a true generalization test. This is a genuine evaluation-validity risk, and the paper should have disclosed whether the split was performed before augmentation. However, this is data leakage or experimental-design ambiguity, not circular reasoning: the reported accuracies are not 'equivalent to their inputs by construction' in the sense of a self-definitional derivation, a fitted parameter renamed as a prediction, or an imported uniqueness claim. The comparison itself is externally grounded in two standard off-the-shelf model architectures and public datasets, so the paper is not self-referential in its derivation chain. Accordingly, no specific circular step can be exhibited under the required standard, and the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper is an empirical benchmark, so the ledger captures the hand-chosen training and evaluation settings that the central comparison depends on, plus domain assumptions about transfer learning, augmentation, and metric validity. No new theoretical entities are introduced.

free parameters (6)
  • Mask R-CNN initial learning rate = 0.00025
    Section 2.3 sets the learning rate without reported tuning; the comparison depends on it.
  • Mask R-CNN batch size = 2
    Section 2.3 sets batch size to 2; no sensitivity analysis is reported.
  • YOLO-v7 initial and end learning rates = 0.01 to 0.1
    Section 2.4 and Table 4 set SGD learning rates; no grid search is reported.
  • YOLO-v7 augmentation hyperparameters = H 0.015, S 0.7, V 0.4, translation 0.2, zoom 0.5, rotation 0.5, mosaic 1
    Table 3 gives hand-chosen augmentation strengths that affect the training distribution and hence the reported mAP.
  • IoU positive threshold = 0.2
    Table 4 sets IoU threshold 0.2 for YOLO-v7; choosing a different positive threshold changes precision and mAP.
  • Training and test split ratio = 90% training, 10% validation and test
    Section 2.1 Table 1 states the split; because the split is applied after augmentation, the effective train and test overlap is uncontrolled.
assumptions (5)
  • domain assumption Pre-trained COCO weights transfer useful features to concrete crack and spall segmentation.
    Sections 2.3 and 2.4 state transfer learning from COCO pre-trained weights without validating that COCO features are sufficient for this domain.
  • domain assumption Geometric and color augmentation preserves the validity of damage labels in the augmented images.
    Section 2.2 describes flips, rotations, zoom, HSV, and mosaic augmentation; if labels are stretched or color-shifted inconsistently, the annotations become noisy.
  • domain assumption The mAP@0.5 metric is a faithful proxy for practical damage detection quality.
    Section 3.1 defines mAP and IoU threshold, but no field validation against manual inspection or severity is provided.
  • domain assumption The three combined public datasets cover a representative distribution of concrete damage.
    Section 2.1 assembles 400 images from three sources; representativeness for bridges, tunnels, and walls is asserted, not measured.
  • domain assumption Internet-sourced images and videos used in Section 3.8 are genuinely out-of-distribution test cases.
    The paper provides no provenance or quantitative comparison for these samples, only example figures.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-driven Detection and Evaluation of Damages in Concrete Structures: Using Deep Learning and Computer Vision." pith.science (2026). https://pith.science/paper/4V427PVR

@misc{pith2026250111836,
  author       = {Pith},
  title        = {Pith review of: Data-driven Detection and Evaluation of Damages in Concrete Structures: Using Deep Learning and Computer Vision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4V427PVR}},
  note         = {Machine review of arXiv:2501.11836}
}
read the original abstract

Structural integrity is vital for maintaining the safety and longevity of concrete infrastructures such as bridges, tunnels, and walls. Traditional methods for detecting damages like cracks and spalls are labor-intensive, time-consuming, and prone to human error. To address these challenges, this study explores advanced data-driven techniques using deep learning for automated damage detection and analysis. Two state-of-the-art instance segmentation models, YOLO-v7 instance segmentation and Mask R-CNN, were evaluated using a dataset comprising 400 images, augmented to 10,995 images through geometric and color-based transformations to enhance robustness. The models were trained and validated using a dataset split into 90% training set, validation and test set 10%. Performance metrics such as precision, recall, mean average precision (mAP@0.5), and frames per second (FPS) were used for evaluation. YOLO-v7 achieved a superior mAP@0.5 of 96.1% and processed 40 FPS, outperforming Mask R-CNN, which achieved a mAP@0.5 of 92.1% with a slower processing speed of 18 FPS. The findings recommend YOLO-v7 instance segmentation model for real-time, high-speed structural health monitoring, while Mask R-CNN is better suited for detailed offline assessments. This study demonstrates the potential of deep learning to revolutionize infrastructure maintenance, offering a scalable and efficient solution for automated damage detection.

Figures

Figures reproduced from arXiv: 2501.11836 by the authors.

Figure 1
Figure 1. A Sample of Labeled Data through augmentation to 10,995 images. All images were resized to a uniform 640×640 resolution for preprocessing [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Sample Mask R-CNN Output for Spall 7 [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. YOLO-v7 Instance Segmentation Model Predictions [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Ground Truth (above) vs. Predicted Labels (below) [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: YOLO-v7 Instance Segmentation Model Training Performance [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: YOLO-v7 Instance Segmentation Confusion Matrix [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Comparison of the Accuracy of the Two Models [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Comparing Speed (orange frames/sec) and Runtime (milliseconds) [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Testing the Accuracy of the Model with a Random Photo Taken from the Internet [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Testing the Accuracy of the Model with a Random Video Taken from the Internet [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Vision-based autonomous structural damage detection using data-driven methods

    cs.CV 2025-01 conditional novelty 3.0 of 10

    On a public wind-turbine surface-image dataset, YOLOv7 reaches 82.4% mAP@50 at 11 ms per image, beating YOLOv7-tiny and Faster R-CNN.

Reference graph

Works this paper leans on

30 extracted references · 28 canonical work pages · cited by 1 Pith paper

  1. [1]

    C. F. Özgenel. Concrete crack segmentation dataset. Mendeley Data 2019

    work page 2019

  2. [2]

    Zhang, C

    C. Zhang, C. Chang, and M. Jamshidi. Simultaneous pixel-level concrete defect detection and grouping using a fully convolutional model. Structural Health Monitoring 2021, 20, 2199–2215

    work page 2021

  3. [3]

    Y . Shawn. FCN for crack recognition. Available online:https://github.com/OnionDoctor/FCN_for_crack_ recognition (accessed on 13 March 2018)

    work page 2018

  4. [4]

    M. Gatti. Structural health monitoring of an operational bridge: A case study. Eng. Struct. 2019, 195, 200–209

    work page 2019

  5. [5]

    H. Kim, E. Ahn, M. Shin, and S.-H. Sim. Crack and noncrack classification from concrete surface images using machine learning. Struct. Health Monit. 2019, 18, 725–738

    work page 2019

  6. [6]

    Dorafshan, M

    S. Dorafshan, M. Maguire, N. V . Hoffer, and C. Coopmans. Fatigue crack detection using unmanned aerial systems in under-bridge inspection. Ida. Transp. Dep. 2017, 2, 1–120

    work page 2017

  7. [7]

    Dorafshan, R

    S. Dorafshan, R. J. Thomas, and M. Maguire. Fatigue crack detection using unmanned aerial systems in fracture critical inspection of steel bridges. J. Bridge Eng. 2018, 23, 04018078

    work page 2018

  8. [8]

    M. R. Jahanshahi, J. S. Kelly, S. F. Masri, and G. S. Sukhatme. A survey and evaluation of promising approaches for automatic vision-based defect detection of bridge structures. Struct. Infrastruct. Eng. 2009, 5, 455–486

    work page 2009

Show all 30 references
  1. [9]

    M. R. Jahanshahi and S. F. Masri. A new methodology for non-contact accurate crack width measurement through photogrammetry for automated structural safety evaluation. Smart Mater . Struct.2013, 22, 035019

    2013

  2. [10]

    C. M. Yeum and S. J. Dyke. Vision-based automated crack detection for bridge inspection.Comput.-Aided Civ. Infrastruct. Eng. 2015, 30, 759–770

    2015

  3. [11]

    Badrinarayanan, A

    V . Badrinarayanan, A. Kendall, and R. Cipolla. Segnet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 2017, 39, 2481–2495

    2017

  4. [12]

    J. Long, E. Shelhamer, and T. Darrell. Fully convolutional networks for semantic segmentation. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition , Boston, MA, USA, 7–12 June 2015; pp. 3431–3440

    2015

  5. [13]

    Y . J. Cha, W. Choi, G. Suh, S. Mahmoudkhani, and O. Büyüköztürk. Autonomous structural visual inspection using region-based deep learning for detecting multiple damage types. Comput.-Aided Civ. Infrastruct. Eng. 2018, 33, 731–747

    2018

  6. [14]

    C. V . Dung and L. D. Anh. Autonomous concrete crack detection using deep fully convolutional neural network. Autom. Constr .2019, 99, 52–58

    2019

  7. [15]

    Bigdeli, H

    N. Bigdeli, H. Jabbari, and M. Shojaei. An intelligent method for classification of cracks in concrete structures based on deep neural networks. Amirkabir Journal of Civil Engineering 2020, 53, No. 8

    2020

  8. [16]

    Heidari, M

    A. Heidari, M. R. Pourtabari, and M. R. Khalilianpour. Evaluation of cracking in concrete dams by artificial neural networks. In 2nd International Conference on New Research Findings in Civil Engineering, Architecture and Urban Management, Tehran, 2016. 16 A PREPRINT - JANUARY...

    2016

  9. [17]

    Azimi, A

    M. Azimi, A. D. Eslamlou, and G. Pekcan. Data-driven structural health monitoring and damage detection through deep learning: State-of-the-art review. Sensors 2020, 20(10), 2778

    2020

  10. [18]

    Y . Xu, Y . Bao, Y . Zhang, and H. Li. Attribute-based structural damage identification by few-shot meta learning with inter-class knowledge transfer. Structural Health Monitoring 2020

    2020

  11. [19]

    W. R. L. d. Silva and D. S. d. Lucena. Concrete cracks detection based on deep learning image classification. In Proceedings 2018, 2, 489

    2018

  12. [20]

    Y . J. Cha and W. Choi. Deep learning-based crack damage detection using convolutional neural networks. Comput.-Aided Civ. Infrastruct. Eng. 2017, 32(5), 361–378

    2017

  13. [21]

    A. R. Majdi Flah, M. L. Suleiman, and M. L. Nehdi. Classification and quantification of cracks in concrete structures using deep learning vision-based techniques. Cement and Concrete Composites 2020, 114, 103781

    2020

  14. [22]

    Ahmed, S

    B. Ahmed, S. Mangalathu, and J. S. Jeon. Seismic damage state predictions of reinforced concrete structures using stacked long short-term memory neural networks. Journal of Building Engineering 2022, 46, 103737

    2022

  15. [23]

    Harirchian, T

    E. Harirchian, T. Lahmer, and S. Rasulzade. Earthquake hazard safety assessment of existing buildings using optimized multi-layer perceptron neural network. Energies 2020, 13(8), 2060

    2020

  16. [24]

    Harirchian, K

    E. Harirchian, K. Jadhav, V . Kumari, and T. Lahmer. ML-EHSAPP: a prototype for machine learning-based earthquake hazard safety assessment of structures by using a smartphone app. Eur . J. Environ. Civ. Eng.2021, 1–21

    2021

  17. [25]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , Las Vegas, NV , USA, 26–30 June 2016; pp. 770–778

    2016

  18. [26]

    Y . Xu, S. Wei, Y . Bao, and H. Li. Automatic seismic damage identification of reinforced concrete columns from images by a region-based deep convolutional neural network. Structural Control and Health Monitoring 2019, e2313

    2019

  19. [27]

    Maeda, Y

    H. Maeda, Y . Sekimoto, T. Seto, T. Kashiyama, and H. Omata. Road damage detection and classification using deep neural networks with smartphone images. Computer-Aided Civil and Infrastructure Engineering 2018, 33(12), 1127–1141

    2018

  20. [28]

    G. Li, B. Ma, S. He, X. Ren, and Q. Liu. Automatic tunnel crack detection based on U-Net and a convolutional neural network with alternately updated clique. Sensors 2020, 20, 717

    2020

  21. [29]

    Guo et al

    M.-H. Guo et al. Segnext: Rethinking convolutional attention design for semantic segmentation. arXiv preprint arXiv:2209.08575 (2022)

    2022 arXiv

  22. [30]

    Mousavi and A

    M. Mousavi and A. Bakhshi. Crack detection in concrete members using encoder-decoder models based on deep learning. Journal of Civil Engineering 2022, 38.2(2.2), 79–88. 17

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.