A New Smoke Segmentation Method Based on Improved Adaptive Density Peak Clustering

Round 1

Reviewer 1 Report

This study proposes a smoke segmentation method. Its main contribution consists in using an improved adaptive density peak clustering.

The document is generally easy to read and follow.

The English needs minor spell checking.

The document is well supported with references.

The subject of the paper has great potential of application.

The proposed work main weakness is the application of the proposed algorithm to simple scenes. It lacks more complex backgrounds or other perspectives like aerial images.

 

In line 149 authors should justify the use of 500 as the initial superpixel parameter.

 

Figure 3 should be referenced in the text before it is shown. Please correct.

 

In Figure 4 authors reference several algorithms but Watershed and MIT were never referenced before in the document text contrarily to the other algorithms. Authors must also include a previous reference two these algorithms along with bibliographic references.

 

Authors should include Figure 5 after it is referenced in the text. Please correct.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

I have read the paper "A new smoke segmentation method based on improved adaptive density peak clustering." This study provided an innovative method to improve the fire image segmentation algorithm. The new method significantly improved the accuracy and effectiveness. 

The overall quality of this study is desirable. Furthermore, this study is urgently needed and it can largely benefit society. Therefore, I highly recommend accepting this paper as it is and getting it published as soon as possible!  

Author Response

Thank you for your recognition of this study!

Reviewer 3 Report

 A review of the paper "A new smoke segmentation method based on improved adaptive density peak clustering" has been carried out and have some suggestions:

- The references must be updated and some references with works related to this topic should be included. For example "Unsupervised segmentation of fire and smoke from infrared videos", "Smoke detection based on multi feature fusion", "Deep smoke segmentation" and others.

- In the Algorithm 2, ¿How do the user select the decision value?. ¿Is it the same value for different set of images? Explain in detail in the manuscript.

- Paper could include some different metrics such as Intersection over Union (mIoU) in order to validate the obtained results.

- In my opinion, additional experiments are required. Set of images with different conditions should be considered to test the robustness of the algorithm.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

In this work, authors propose smoke segmentation using improved double truncation distance self-adaptive density peak clustering (TSDPC). I have the following comments:

1. Authors do not show ablations. how does the proposed TSDPC algorithm improve over the baseline superpixel segmentation.

2. The test dataset proposed and used for this paper is too small to be conclusive. Should show results on other existing datasets.

3. Smoke segmentation is not a new problem. However, there is no comparison with state-of-the-art. The comparison table/image shown are generic and do not have any citations.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The comments made for paper "A new smoke segmentation method based on improved adaptive density peak clustering" were addressed correctly. Thank you.

Author Response

 Thank you for your recognition of this study!

Reviewer 4 Report

The authors have addressed all of my comments. Thank you including results on another dataset and showing comparison. I have some follow-up comments:

1. Thank for clarifying the importance of super pixel segmentation. My question was about ablations in general e.g. difference in performance before and after TSDPC.

Author Response

Please see the attachment.

Author Response File: Author Response.docx