Intelligent Detection of Underwater Defects in Concrete Dams Based on YOLOv8s-UEC

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1) The article is well-structured and follows a logical flow. However, there are several instances where the language could be made more concise, and certain sections could be clearer. For example, the description of the YOLOv8s-UEC network and the Efficient-RepGFPN structure (Section 2.3) is quite technical but lacks sufficient explanatory depth, especially for readers unfamiliar with these advanced machine learning models.

2) The contribution of establishing a dataset of underwater defect images is notable, especially given the difficulty in obtaining such data. However, more emphasis should be placed on the limitations of the dataset, such as its size and generalizability to different dam environments. Additionally, while the authors propose various improvements to the YOLOv8s network, there is limited comparison to other state-of-the-art methods beyond their brief mention in the literature review.

3) The ablation study is thorough and appropriately addresses the incremental improvements to the network, but the choice of hyperparameters and training setup could benefit from more detailed justification. For example, it would be beneficial to explain why certain values for the learning rate, batch size, or epoch number were selected and how these may affect the reproducibility of the results. The training conditions and hardware specifications (e.g., GPU/CPU setup) should also be clearly stated for reproducibility.

4) The figures included, such as Figure 7 (proposed YOLOv8s-UEC network architecture), are helpful for understanding the structural modifications. However, some figures (like Figure 14, showing precision-recall curves) could use more detailed explanations in the text, highlighting key takeaways and what differentiates each network. Figure 15, which visualizes detection results, provides an excellent demonstration of the improvements. However, adding more examples, particularly of failure cases, would enrich the analysis.

5) The literature review is comprehensive but lacks depth in discussing more recent advancements in underwater imaging technologies and deep learning for similar detection tasks. The authors should include a more in-depth discussion of recent work on underwater defect detection and clarify how their improvements compare quantitatively and qualitatively to previous methods.

6) There is no mention of possible improvements or future directions, such as incorporating transfer learning for more diverse environments or adapting the model for real-time implementation with lower resource consumption.

Comments on the Quality of English Language

There are several grammatical errors throughout the manuscript, especially in the abstract and introduction. A thorough proofreading would improve readability and reduce confusion.

Author Response

非常感谢您花时间审阅此手稿。请在重新提交的文件中找到下面的详细回复和突出显示的相应修订/更正/跟踪更改。

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The study proposes the use of YOLOv8s-UCE model for detecting defects in underwater concrete structures. The study is interesting and has certain results. Before publication, the following points should be considered for improvement:

1.       Introduction: The limitations of DL application studies in detecting defects in aquatic environments should be further analyzed to highlight the novelty and necessity of this study. The strengths and improvements of the proposed method should also be mentioned to fill the research gaps.

2.       The title of section 2 should be changed to be more appropriate.

3.       Further comments on the choice of YOLOv8s? How is it better suited for this study? Is there a need to compare with other sizes?

4.       Formulas should be cited

 

5.       The conclusion should be more clearly written about the contributions and novelties of the research.

Comments on the Quality of English Language

Minor editing of English language required.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

 

In this manuscript, a concrete dam underwater apparent defect detection algorithm named YOLOv8s-UEC for intelligent identification of underwater defects is presented. Here are some comments:

1. Fig. 2: The quality and font size are poor. I recommend upgrading them. A similar issue exists with Figs. 3, 7, and 9.
2. Dataset: I don’t understand how this data allows for proper model generation. How many images were recorded? How many scenarios were considered, and why these scenarios? How can this case be generalized, i.e., can similar defects be detected even in scenarios different from those in the training? My concern is that the defects in the dataset were manually generated. Is data augmentation really useful in this case? More discussion is required.
3. Implementation Details: I see "good" experimental results, but in a real-world scenario, will an underwater agent capture the data, and then online processing be carried out? I think a smart camera implementation could be more feasible and suitable for real-world scenarios, but in this context, does the model allow for embedded implementation using dedicated hardware (GPGPU/FPGA/ASIC)? If not, could embedded sequential processors such as ARM provide enough computational resources? Discussion on this would be a nice complement to the current manuscript.
4. Results: Information regarding processing speed and computational resource usage is missing. Including both would be a nice complement to the current manuscript.

 

Comments on the Quality of English Language

1 – There are some minor grammatical and style errors. I suggest a detailed revision of the English language.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The update has been improved