Refined Deformable-DETR for SAR Target Detection and Radio Signal Detection
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe paper proposes a "Refined Deformable-DETR designed to enhance detection performance in SAR and signal processing scenarios." The paper's contributions and novelties may be considered a variation of a known concept. The paper's readability requires improvement. Also, the paper contains typos and grammar errors. Additionally, the paper presentation is good. Also, the paper is technically sound, even though its readability may jeopardize this analysis. Experimental evaluations corroborate the proposed method. However, the papers require some more discussion and analysis. As suggestions to improve the paper's quality, the authors should consider:
- The authors should better motivate and contextualize their proposal. The authors should consider enhancing the state-of-the-art discussion regarding detection transformers for the evaluated application.
- The authors should improve the discussion regarding the implementation aspects and the selection of network configuration/parameters. What are their impacts on the performance of the proposed technique?
- The authors should present and discuss the evaluated datasets better.
- As can be observed, the performance comparison shows that the proposed technique achieves marginal gains in some scenarios. Thus, to better motivate the publication, the authors should include a statistical analysis regarding the performance metrics. Are the presented performance gains statistically significant?
- The authors should present qualitative analyses and justify the observed performance behaviors.
- The discussion of the observed performance results should be improved overall. For instance, the results show that the proposed method achieves different performance gains for specific metrics in some data sets. Additionally, the technique was outperformed by others in the literature in some scenarios. What motivates this performance behavior? What are the situations where this behavior is more expected?
- The authors should present an analysis of computational costs.
- The authors should improve the discussion and present more analyses for the ablation studies.
Minor Comments:
- The authors should consider proofreading.
- There are some undefined symbols and acronyms.
- The authors should adequately use equation formalism.
Comments on the Quality of English LanguageThe paper needs proofreading.
Author Response
We thank the reviewer for this crucial feedback.
The revised manuscript now explicitly differentiates our method through methodological comparisons and experimental validation. We have added critical metadata: spatial resolution, polarization mode, and scene complexity metrics. These details better contextualize the dataset's diversity and validate our method's robustness under heterogeneous SAR imaging conditions. We acknowledge the importance of elaborating on the implementation aspects and the selection of network configurations/parameters. In our revised manuscript, we will provide a more detailed discussion on how these factors influence the performance of our proposed technique, including their impact on detection accuracy, computational efficiency, and robustness in real-world scenarios. We will enhance our discussion on the observed performance results by analyzing the factors contributing to the varying performance gains across different metrics and datasets. Additionally, we will provide a detailed explanation of the scenarios where our method is outperformed by others in the literature, discussing the underlying reasons and conditions where such behavior is expected. To provide a more comprehensive evaluation, we will include an analysis of the computational costs, including inference time and memory consumption, to assess the trade-off between performance and efficiency."
We will carefully review the manuscript and make necessary edits to improve the clarity and readability of the English language.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors- The novelty is not very clear, and it should be clarified further. The differences between the proposed approach and the existing method should be explicitly enlightened.
- Half-window filter is not well presented, either in theory or experiments. It should be enhanced from the principle, visual illustration and visual comparison.
- Why ATSS, YOLOv3 and Faster-RCNN rather than other methods are used for auxiliary feature extractors ?
- How alpha and beta in 19 are set? How do they affect the performances?
- The performance in Tab. 2 lacks detailed analysis. Especially, why mAP_50 is worse than the other two methods?
- Ablation study is not convincing, and it lacks visual result comparison.
English could be improved by the native speaker.
Author Response
Comments 1:The novelty is not very clear, and it should be clarified further. The differences between the proposed approach and the existing method should be explicitly enlightened.
Response 1: We thank the reviewer for this crucial feedback. We have explicitly highlighted the key innovations of our approach, particularly and provided direct comparisons with existing methods
Comments 2:Half-window filter is not well presented, either in theory or experiments. It should be enhanced from the principle, visual illustration and visual comparison.
Response 2:We thank the reviewer for the valuable feedback. In the revised manuscript (Section 3.1, pages 5-6), we have significantly enhanced the Half-Window Filter (HWF) presentation.
Comments 3:Why ATSS, YOLOv3 and Faster-RCNN rather than other methods are used for auxiliary feature extractors?
Response 3: Thank you for raising this critical question. These methods were selected due to their diverse architectures—YOLOv3 for real-time detection, Faster-RCNN for region-based accuracy, and ATSS for anchor-free robustness—ensuring a comprehensive evaluation of our proposed framework.
Comments 4:How alpha and beta in 19 are set? How do they affect the performance?
Response 4: Thank you for raising this important point. We have added a detailed explanation of the parameter settings and their impact on performance in the revised manuscript.
Comments 5:The performance in Tab. 2 lacks detailed analysis. Especially, why mAP_50 is worse than the other two methods?
Response 5: While the mAP50 is slightly lower, this is due to our model's stricter localization criteria (IoU > 0.6) designed to reduce false positives in complex SAR environments. This trade-off results in a superior overall mAP (0.682 vs. 0.660-0.669). Such an approach aligns with real-world maritime surveillance needs, where precise target delineation is more critical than a loose count of detected objects.
Comments 6: Ablation study is not convincing, and it lacks visual result comparison.
Response 6: Thank you for your valuable comment. We have expanded the ablation study with a more detailed analysis of different configurations and their impact on detection performance (Tables \ref{tab2} and \ref{tab3}). While we acknowledge the value of visual comparisons, our focus remains on quantitative analysis for objective evaluation. The extended results provide clearer insights into model behavior, and we will consider adding visual comparisons in future work.
We will carefully review the manuscript and make necessary edits to improve the clarity and readability of the English language.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper proposes a well-motivated improvement to Deformable DETR, demonstrating strong quantitative performance on SAR and radio signal detection tasks. However, significant issues remain, particularly regarding model justification, dataset description, and failure case analysis. Additionally, formatting, citation issues, and unclear wording need to be addressed before publication.
1、The matrix shape for the symbol representations is not explained on page 5.
2、The derivation of the coefficients in Equations 2-5 lacks an explanation.
3、Section 3.2 and Figure 2 provide unclear descriptions of the multi-scale feature extraction method.
4、The content of Figure 3(b) seems inconsistent with its description.
5、Figure 4 needs to include an explanation of the visualization method and a detailed analysis of the experimental phenomena.
Comments on the Quality of English LanguageMinor editing of English language required.
Author Response
Comments 1: The matrix shape for the symbol representations is not explained on page 5.
Response 1: Thank you for your feedback. We will clarify the matrix shape for the symbol representations on page 5 in the revised manuscript to ensure better understanding.
Comments 2: The derivation of the coefficients in Equations 2-5 lacks an explanation.
Response 2: Thank you for your feedback. We acknowledge the need for a clearer explanation of the coefficient derivation in Equations 2–5. In the revised manuscript, we will provide a detailed explanation of their mathematical foundation and derivation process to enhance clarity.
Comments 3: Section 3.2 and Figure 2 provide unclear descriptions of the multi-scale feature extraction method.
Response 3: Thank you for your valuable feedback. We acknowledge that the description of the multi-scale feature extraction method in Section 3.2 and Figure 2 could be clearer. In the revised manuscript, we will refine the explanation and provide additional details to improve clarity.
Comments 4: The content of Figure 3(b) seems inconsistent with its description.
Response 4: Thank you for your feedback. We will carefully review Figure 3(b) and its corresponding description to ensure consistency. If any discrepancies are found, we will revise either the figure or the text accordingly in the revised manuscript.
Comments 5: Figure 4 needs to include an explanation of the visualization method and a detailed analysis of the experimental phenomena.
Response 5: Thank you for your suggestion. In the revised manuscript, we will provide a clearer explanation of the reason for using Figure 4 and the chosen visualization method to enhance clarity and interpretability.
We will carefully review the manuscript and make necessary edits to improve the clarity and readability of the English language.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe majority of my previous comments were properly addressed. However, there are some issues regarding spacing, writing, and the equation's formalism, which should be addressed for the final paper version.
Author Response
Thank you for your valuable feedback. We appreciate your careful review and insightful comments. We have carefully revised the manuscript to address the issues regarding spacing, writing, and equation formalism to improve the clarity and readability of the final version. Please refer to the revised manuscript for the detailed modifications.
