An Improved Traffic Congestion Monitoring System Based on Federated Learning

Round 1

Reviewer 1 Report

The article contains information technical and innovative. The problem addressed is current and has technical relevance, which makes it significant. The abstract is written concisely. The paper is well organized and convincing. The experimental methodology is described as comprehensively. Interpretations and conclusions are justified by the results. 
My recommendations are:
a) The abstract can be rewritten to be more meaningful. The authors should add more details about their final results in the abstract and how the proposed approach is validated;                                                              b)The paper does not explain clearly its advantages with respect to the literature: it is not clear what is the novelty and contributions of the proposed work: does it propose a new method? Or does the novelty only consists of the application? The advantage of the proposed method with respect to other methods in the literature should be clarified;
c) Quality of Figures is so important too. Please provide some high-resolution figures. Some figures have a poor resolution. The comparison of different methods using clear graphs should be explained;                                          d) Authors should explain the reason why they choose these neural networks algorithms. What are the limitations of this work? How can the rigour of this work be demonstrated?;                                                                            e) The paper does not provide significant experimental details needed to correctly assess its contribution: What is the validation procedure used?;     f)Results need more explanations. Additional analysis is required for each experiment to show its main purpose;                                                        g) Comparison with recent study and methods would be appreciated.      h)The conclusion should state the scope for future work;                              i) Discuss future plans with respect to the research state of progress and its limitations.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

At the beginning, I'll make a general remark on the whole work. Despite the fact that the issue is very interesting, the authors have included many parts that are unnecessary. The article should be formulated on a frame: to identify the research problem, describe the work of other scientists and indicate the scientific gap, describe their own solution and show the results. This work contains as many as 21 pages, where there are many parts that explain something that is already done and described in other articles such as CNN, etc. Can you please rebuild the article in such a way that you only score which methods you use and indicate what modifications have been applied? In this way the scientific contribution will be more visible.

Below I also provide suggestions for the individual parts of the article:

The abstract of the article should be rewritten. It should describe more clearly what new is proposed. Especially since the title indicates that this is an improved method. So, in relation to which specific method is this improvement?

Figure 1 should be removed because it is not directly related to the subject of the article.

The introduction describes the method that was adopted for scientific research. However, the aim of the article is not written. It is not sufficiently indicated whether this problem is so important that it should be solved.

At the end of the introduction, the structure of the further article should also be described, so that the reader knows where to look for information of interest.

The state of literature is very poorly described. Chapter 2.4 is crucial here, but there are only 2 articles with other solutions. Can you please focus on this chapter and expand it strongly? There is a lot of scientific work that deals with this topic. Please show them in a solid way and describe their concrete advantages and disadvantages. At the end of this chapter, please also indicate the research gap and how your work will help to fix it.

The model validation in chapter 5.1 is also poor. The sample of 2 images is not statistically significant. So can you please increase the sample significantly to estimate the accuracy of the prediction?
Such a test should also be carried out for different weather conditions and times of day and night. The conditions under which this method can be used should be discussed. Can you please also show a comparison with other methods? In the title you wrote that your method is improved and it has not been discussed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The objective of the research is interesting, but the methodology is inadequate and has many shortcomings. It is not clearly specified what the dataset is (Los Angeles dataset is very vague because there are a lot of datasets related to Los Angeles), it is not specified how the dataset is divided into training, validation and testing, so the accuracy is not interpretable without this information.

For all these reasons, the results are not correctly interpretable and cannot be evaluated.

I think it would be necessary to rewrite the article giving priority to the description of the data and the methodology, instead of describing the architectures used, that are already well known and described in their corresponding articles, and solved by standard software packages (PaddlePaddle, Tensorflow, Yolo...).

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors made the changes as requested. I am in favor of its publication

Author Response

Thank you for your suggestions and guidance on our work. I am very honored and grateful that you agree to the publication of our article.

Reviewer 2 Report

NC

Author Response

Thank you for your suggestions and guidance on our work. I am very honored and grateful that you agree to the publication of our article.

Reviewer 3 Report

The paper is improved a lot in the last version.

What do you consider true / false classification in your experiment must be clarified. ¿Number of car detected? ¿Bounding box similar to target? ¿How similar? ...

In figures 12-17 you must to scale error in another range, because 0-100 is good for accuracy but it's bad for error. This new scale will be placed at right in the figure.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf