DASeg: A Domain-Adaptive Segmentation Pipeline Using Vision Foundation Models—Earthquake Damage Detection Use Case
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors- 1 white text is unreadable
- All embedded font should appear as 10pt or larger when image is in the manuscript. Many of yours are not. This makes them unreadable. Please change them.
- Use of lightblue text in caption 1 is unreadable
- 1 would have been much better if oriented vertically. Then we could actually see what is happening in your images. If you are going to use this style of flow chart such thumbnails must be legible
- Methodology section – please specify whether the imagery data sets were take from the ground, mobile terrestrial, uav, helicopter, fixed wing or satellite. Als provide the range of resolutions of the images. This would really benefit from a table where other meta data could be presented. Having it in the text makes it hard to follow as there are 5 data sets
- Results- please report the computation expense as part of your discussion of fig. 1
- Please add subsection to the discussion where you discuss the objects and scenes that were most commonly mis-classified. Please report what they were misclassified is and then check some of the approaches that you used for benchmarking and see if they also missed those same types of objects and scenes. Doing so could be the most valuable part of the paper
Author Response
Dear Reviewer 1,
We sincerely thank you for your helpful suggestions on our paper. We have addressed all your comments, and detailed responses can be found in the attached document.
Best regards,
Huili Huang
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsComments
Abstract:
The authors need to rewrite the abstract completely. The first sentence is confusing and needs to be reformulated. The objective must be clearly stated (e.g., “the objective is to develop, perform, evaluate…”). Please be straightforward. The methodology should be described with greater clarity. In the results section, the authors must highlight the main findings. Finally, a concluding sentence must be included.
Therefore, I recommend the authors follow this structure: one introductory sentence, a clear statement of the objective, a brief methodology description, main results, and a concluding sentence.
1. Introduction:
In general, the introduction of a scientific article should not contain images. I ask the authors to remove them from this section and instead incorporate the data through contextualization in the text.
The authors should explain the consequences of not performing disaster assessment promptly after the event.
The second paragraph lacks context; please revise to ensure logical flow between paragraphs—this issue recurs throughout the text.
The importance of using new models should be explained before stating the objective.
The objective is currently unclear and appears in an inappropriate place. I suggest reorganizing the introduction to present the objective as a closing paragraph. As it stands, it disrupts the structure and coherence.
Lines 80 to 93 resemble a conclusion rather than an introduction and should be deleted.
2. Related Work:
This section provides broad and thematically organized coverage of relevant prior work.
I suggest the authors emphasize the practical limitations of model generalization in real-world disaster scenarios.
Expanding on the methodological and computational challenges of using segmentation and interpretability in real-time settings would also be beneficial.
The authors should establish clearer connections between the reviewed methods and this paper's specific focus: social media images in the context of disasters.
Additionally, the integration of segmentation and explainability methods could be further explored as a promising research direction to enhance model trustworthiness in critical situations.
3. Materials and Methods:
I suggest including a flowchart to visually represent the entire methodological pipeline.
3.1.3. Bounding Box Annotation:
Please justify the choice of the parameters used in this stage.
4. Results:
Lines 346 to 366 describe methodological procedures not in the results section. These lines should either be removed or relocated to the methodology section.
5. Discussion:
This section lacks direct comparison with the existing literature. Although some alternative approaches (e.g., Knowledge Distillation, open-set detection) are briefly mentioned, there is no direct contrast between DASeg-Quake results and those from previously cited methods, such as Deep-Disaster [21], Grad-CAM [29], DAV [22], or the DSS dataset benchmarks.
I suggest adding paragraphs with quantitative or qualitative comparisons, highlighting how DASeg outperforms or complements existing approaches in terms of accuracy, speed, or scalability.
The authors mention that the results were "excellent," but no specific metrics (e.g., mIoU, F1-score, inference time) are presented or compared to prior benchmarks. I recommend reinforcing the discussion with these key metrics to position the method in relation to the state-of-the-art properly.
Furthermore, several models are mentioned as potential technical complements (e.g., ViT-CX, LRP, AlphaCLIP), but there is no critical discussion on how they would influence DASeg’s performance.
I also recommend that the authors include a discussion of the limitations of their current work.
6. Conclusions:
The authors should present at least one key result to directly respond to the stated objective. It is also important to include clear recommendations for future work.
The article requires a major revision in English.
Author Response
Dear Reviewer 2,
We sincerely thank you for your valuable suggestions on our paper. We have carefully addressed all your comments and have reorganized the abstract, introduction, and conclusion sections accordingly. Detailed responses are provided in the attached document.
Best regards,
Huili Huang
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI think the authors have done an adequate job responding to the reviewers' concerns. The only additional thing that I would like to see done to the manuscript is a light re-writing so that it is no longer in the first person (e.g. "we", "our", "ours", etc.)
Author Response
Comment1:
I think the authors have done an adequate job responding to the reviewers' concerns. The only additional thing that I would like to see done to the manuscript is a light re-writing so that it is no longer in the first person (e.g. "we", "our", "ours", etc.)
Response1:
We authors thank Reviewer 1 for the suggestion related to the writing style. We had intentionally used the first person since that is the protocol used in all other manuscripts published in the special issue: Machine Learning at the Object: Fine-Grained Extraction and Analysis in Remote Sensing. We will defer to the Editor for guidance if we need to further address this.
Reviewer 2 Report
Comments and Suggestions for AuthorsAbstract
The first sentence must be rewritten. Why is it “crucial”? Not everything is necessarily crucial. The authors should be more explicit about the study’s objective, as requested in the first round of review. Additionally, the methodology must be presented in the abstract. The last sentence should be deleted.
Introduction
The second paragraph remains disconnected from the first, resulting in incoherent information. Overall, the introduction lacks logical flow between paragraphs and presents scattered ideas that have already been flagged and remain unchanged. The authors should remove the first sentence of the final paragraph, as it presents preliminary results that are inappropriate for this section.
Materials and Methods The authors should improve the organization of the methodology section. Figures are presented consecutively without accompanying explanatory text.
Results The first and second paragraphs should be removed, as they do not present results but instead describe methodology, an issue already highlighted in the first review.
Discussion The authors have not grasped the need to discuss the results in light of relevant literature. The section currently presented does not constitute a proper discussion. I recommend the authors study more about how to effectively write a scientific discussion in order to analyze their results properly.
Comments on the Quality of English LanguageI recommend that the authors seek assistance from a native English speaker to improve the clarity and fluency of the manuscript.
Author Response
We authors thank the reviewer for the comments. Please see the attachment.
Author Response File: Author Response.pdf
