Real-Time Semantic Segmentation of Point Clouds Based on an Attention Mechanism and a Sparse Tensor

Round 1

Reviewer 1 Report

1. The English can be improved

2. The paper is well written and clear

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

 

I have several questions that need further improvement in the manuscript, which proposes a real-time semantic segmentation method based on an attention mechanism and sparse tensor point cloud. The idea is intuitive and the experimental results also prove that the method is effective. However:

1. There is an error in Formula (13) in Section 3.4.2 Summary; please check.

2. The clarity of the pictures in the article is low, such as in Figures 2 and 3; please use photos with higher clarity.

3. It is suggested to add other algorithm detection effect pictures for comparison in Section 4.4.

4. In Section 4.4, it is mentioned that your model can pay more attention to small instances. Still, the accuracy of the experimental results in Table 2 does not seem to be very ideal.

5. There is an error in the format of the real-time index column in Table 3; please check.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

In the paper, a lightweight convolutional network based on an attention mechanism and a sparse tensor is proposed to better balance the accuracy and real-time performance of point cloud semantic segmentation.

The paper is well structured and the methodological process is adequate for the problem posed and the research conducted. The results show that the proposal is interesting for advancing the development of techniques for point cloud semantic segmentation.

I congratulate the authors for the work developed and recommend reviewing the following points of form before publication.

1. In lines 33-36 it is recommended to revise the wording as the intended idea is not clear.

2. In some parts of the document the citation is not separated from the text, please review the whole document.

3. Some words have been incorrectly divided with hyphen between lines, please review throughout the document.

Some of the commented aspects have been highlighted in the attached document.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

In this paper, the authors proposed a lightweight fully convolutional network based on an attention mechanism and sparse tensor to balance better the accuracy and real-time performance of point cloud semantic segmentation. Overall, the study and the topic are very interesting, the work is well written and organized, the experiments are well done, and several metrics have been considered. However, several points should be revised and considered before acceptance. We note:

-          Authors should define all acronyms.

-           The paper’s organization should be added in the last part of the introduction. 

-          Authors should highlight the practical applications of the studied problem.

-          The research gap for existing works should be highlighted to the readers in the last part of the Related Work.

-          Does this study have never been attempted before? Justify this statement and give an appropriate explanation.

-          Line 294, correct “chapter” ----> “Section”.

-          Authors should add some perspectives in the last part of the conclusion.

-          References should be rewritten according to the MDPI format.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have adequately addressed my concerns, and the paper is ready for acceptance in its current form.

Reviewer 4 Report

 The answers of the authors are very convincing. Therefore, the manuscript can be accepted in its present form.