Self-Supervised Depth Completion Based on Multi-Modal Spatio-Temporal Consistency

Round 1

Reviewer 1 Report

This paper proposed a self-supervised network with a set of constraints based on multi-modal spatio-temporal consistency (MSC) for depth compeletion. The novel idea of auto-mask is intuitive and effective.  The proposed self-supervised method shows a new framework in multi-modal tasks and is promising to be promoted in more areas.

1. What are the advantages of the self-supervised networks, and why self-supervised networks have these advantages.

2. The language in the text needs to be polished, and there are some typos and grammar mistakes. The authors are encouraged to invite a native speaker to carefully proofread this paper and make it more readable.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

08/12/2022

Dear authors,

In the manuscript Self-supervised depth completion based on multi-modal spatio-temporal consistency you have shown the newly designed a self-supervised depth completion method is designed based on multi-modal spatio-temporal consistency (MSC). The most critical novelty of this work lies in that the self-supervised mechanism is designed to train the depth completion network by MSC constraint. With the self-supervised mechanism of MSC constraint, the overall system outperforms many other self-supervised networks, even exceeding partially supervised networks.

General comments

The study is interesting and have some potential in targets-detection applications. Furthermore, such manuscripts should be written in the third person. In many places (68) in the text, you have not done so. Below the title of the manuscript are your names, therefore, everything you write in it without references is your results. Because of that, this greatly, greatly irritates the reader.  So, you need to change this throughout the text.

In the Introduction you make a lot of statements that you do not substantiate with references.

You have provided many examples of the use of your method and compared it with existing methods. The results should be discussed in a separate chapter, and only the findings from that discussion should be interpreted in detail in the Conclusion.

In the Conclusion you need to interpret all the results in details and highlight what you have proven in your tests. As a rule, the results are not presented in the conclusion, but they are interpreted in detail.

 

Specific comments (are in the manuscript)

-          Line 2-3 – Nothing is designed in this paper. The method was designed in the research process, it is only described and presented in the paper. Please be precise in your expression.

-          Line 20-23 - Is this your knowledge or did you take it from other sources? If so, please provide references.

-          Line 80 - Such manuscripts should be written in the third person.

-          Line 271-272 - This text is redundant in the Conclusion. You already stated that earlier.

 

 

Best regards

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The paper is interesting to read, however, there are some problems: 

- Figure 1 or the framework description in the text should be given in more detail

- Figure 3 is too huge. It would be good to split it in two or simplify it if possible and move closer to the link on it

- Figure 4 goes before the link on it in the text

- There is no link to KITTI and KITTI2012 datasets in the "Datasets and Setup" section

- Section 3.2 Figures and Tables should be passed in the text after links on them are given.

- It would be better to give the information from lines 243 - 252 right after figure 5

- Table 4 should not be in the conclusion

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Authors have made a hard-working to improve the manucript which can be accepted now.