An Efficient Point Cloud Semantic Segmentation Method Based on Bilateral Enhancement and Random Sampling

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

An Efficient Point Cloud Semantic Segmentation Method Based on Bilateral Enhancement and Random Sampling

 

This paper proposes a bilateral-based segmentation method for accuracy enhancement, meanwhile the runtime is improved by using random sampling. Experiments have demonstrated the efficiency and 

 

1. Please keep the font in Fig. 4 and 5 consistent with the main context. In Fig. 4, what is the unit of Loss? Please show it in the figure.

2. In table 2, compared to PointNet, the proposed method has around 5X number of parameters to train, but its training time is even less than the training time of PointNet. Could the author explain this?

3. It is kind of confusing, when the terms, mIoU in equation 5 and mIoU in Table 4, are the same but refer to different things. Could the author resolve this?

4. Please clean the reference format, such as missing pages. For example:

 

Li, Y.; Bu, R.; Sun, M.; Wu, W.; Di, X.; Chen, B. Pointcnn: Convolution on x-transformed points. In Proceedings of the Advances 411 in neural information processing systems, 2018, Vol. 31. 

 

Author Response

Dear reviewer,

thank you for your review comments. Now I am sending you the reply letter and the revised manuscript as attachments, hoping to get your approval.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. In addition to the statistical results, the abstract should also include conclusions and suggestions for future research.

2. The current research aimed to point the Cloud Semantic Segmentation Method in the right direction, but yet it did not highlight the problem that needs to be addressed.

3. The author of the Materials and Methods section explained the materials rather than providing the methods in order to explain them.

4. There must be a flowchart or a block diagram included in the methods.

5. There is a lot of information in the title of the figure 2 and figure 3. It needs to be more specific to the figure.

6. I would like to know why the section Multi-resolution Decoding Module is used in spite of the fact that it is not mentioned in the paper.

7. The datasets and evaluation metrics should not be included in the results of the experiment.

8. There was no good result related to the topic that I found.

9. The conclusion, which should be the conclusion to the main contribution, has not been shown to me.

10. Statistical accuracy should be presented in both the abstract as well as the conclusion of the paper.

11. It appears that the references have not been updated.

Comments on the Quality of English Language

 Extensive editing of English language required

Author Response

Dear reviewer,

thank you for your review comments. Now I am sending you the reply letter and the revised manuscript as attachments, hoping to get your approval.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I commend the efforts of the authors in incorporating the suggested changes. Nonetheless, minor comments and suggestions for refining this manuscript remain, and I recommend addressing them to enhance the manuscript's potential acceptance in this journal. I eagerly anticipate observing these modifications in the revised version.

Comments:

I have a question about the practical implementation and the computational novelty of this system. Authors need to convince readers about the applications of their system throughout the paper. The abstract is refined, but it should be reconsidered to briefly research gaps before explaining the summary of this work. You can take help from the article entitled deep multi-scale pyramidal features network for supervised video summarization. Similarly, authors should add some info about the utilized datasets. What is the main difficulty when applying the proposed method? Authors should clearly state the limitations of the proposed method in practical applications. You can take help from the article deep learning based speech emotion recognition for Parkinson patient to explain the suggested changes fluently. Current challenges are not clearly mentioned in the introduction section of this paper. I suggest adding a paragraph about the current challenges in this area followed by the authors’ contribution to overcoming those challenges. Authors can follow the introduction section of the following article, crowd counting using end-to-end semantic image segmentation for assistance. I appreciate writing contributions in bullet points, but the contributions of this research in the current manuscript are not crystal clear. I strongly recommend improving the contributions by adding the technical explanation. Captions given to tables and figures are not self-explanatory. I recommend updating all figures and captions.

Author Response

Dear reviewer,

thank you for your review comments. Now I am sending you the reply letter and the revised manuscript as attachments, hoping to get your approval.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

The authors successfully refined the revised version. There are no more comments from my side. I recommend the acceptance of the manuscript.