Research on a Lightweight Panoramic Perception Algorithm for Electric Autonomous Mini-Buses

Round 1

Reviewer 1 Report

Thank you for your article!

Some important questions:

1. Lines 52-55: "Traditional convolutional neural networks view images as one-dimensional vectors, while SCNN treats image data as two-dimensional tensors using adaptive kernel sizes and depths, thereby better preserving the spatial relationships of images" -- this sentence looks incorrect for me as all CNN treats image as two (or three)dimensional vectors. The model that treats image as one-dimensional vector is a fully-connected NN.

2. For my mind, Formulas 1 and 2 don't mach the description in lines 119-121 (The difference between DSC and ordinary convolutions is that the process of ordinary convolutions is divided into two steps: depth-wise convolution and point-wise convolution). I do not see signs of these two-steps in formulas. Could you please explain this point?

3. Line 163: I'd suggest to explain differense between MBConv module and FMBConv in details. 

4. What is the meaning of Formula 7? And is the formula written correctly (upper and lower parts seems identical)?

5. Table 2: Could you please explain more clear why SCNN and ENet give such bad mIoU result?

6. Table 4: Could you please explain why loss function change gave significantly better fps? Is the computational complexity of the proposed loss function less than of the original one? 

 

Minor editorial issues:

1. There are many different abbreviations given in the article, and while some of them, like YOLO, are independent names that are difficult to decipher, for abbreviations like R-CNN, SSD or RSU, their full name (Region-based Convolutional Neural Network, Single Shot Detector, Roadside Unit) can be given.

2. Line 74: ... tasks, In this paper, the YOLOP ... -- capital letter in the middle of a sentence.

3. Text on Figure 1. is too small.

4. Line 101: "... and the parameter is hurge" -- probably better use "model size is very large" or "number of parameters is huge".

5. Line 102: "... meeting real-time requirements requires ..." -- better  use synonym instead of "requirements" or "requires"

6. Line 103: I'd suggest to use term "model size (number of parameters)" instead of "model parameter size" 

7. Lines 256-257: "and set 256 initial learning rates of β1, β2, and 0.001, 0.924, and 0.999 respectively" -- something is missed here, probably b3.

 

 

Author Response

I would like to express my gratitude  for the time and effort that had been spent in handling our manuscript. For your convenience, all of the corrections/changes are highlighted in GREEN in the revised manuscript.Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The article is related to a relevant issue in the mobility of autonomous vehicles in a certain context, for which it proposes the implementation of certain improvements to a widely used algorithm (YOLO). There are several aspects that need to be improved, so that the article can be considered for publication in the WEVJ journal. Below are the aspects mentioned:

1. In the abstract, the YOLOP-E algorithm is initially mentioned. I think it is important to initially mention the context of the problem that the research is trying to solve, which is mentioned later.

2. In the abstract it is mentioned that the YOLOP-E algorithm is used, but then 3 stages of development are mentioned (that are not very well understood), and that also do not mention when the version of the YOLOP-E algorithm was reached and, what does the extra E mean? It is important to improve this.

3. In the introduction, some acronyms are used that are not previously defined. It is important that any acronym used be previously defined.

4. It is important that in the introduction (at the end) it is expressly highlighted what is the novelty of the research, with respect to the current state of the art on the subject.

5. The results obtained should not be mentioned in the introduction, it is recommended to delete the last sentence, which mentions the results of the experiments performed.

6. It is recommended for the entire document to place references before a punctuation mark.

7. Throughout the document, the authors must take into account that all the figures that appear must be PRIORLY referenced. There are several figures that appear without being referenced anywhere in the document.

8. A similar situation occurs with the Tables, all the tables must be properly referenced in a PRIOR way. There are several tables that do not meet this requirement.

9. What is related to the experiments performed should be detailed much better. What was planned to be done, how it was done and what results were obtained.

10. It is recommended to create an independent section for the BROAD discussion of the results, not to do it in section 3.

11. Please remove the numbering of the conclusions in section 4, this is not commonly used in a scientific article.

12. The conclusions section should be considerably improved, part of the abstract should not be repeated again, conclusions should be generated regarding the results obtained.

13. The current section 5 is recommended to be deleted, since it does not apply to this document.

Regarding the use of English, it is important to improve the writing of some sentences. 

Author Response

I would like to express my gratitude to you  for the time and effort that had been spent in handling our manuscript. For your convenience, all of the corrections/changes are highlighted in YELLOW in the revised manuscript.Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have significantly corrected the document, following the recommendations made.

There are still some minor details to fix:

1. Keep in mind how to reference, when you have two references or more in a row, remember that you must use the comma, between the references.

2. In the final text of the article, do not forget to remove some marks that you placed in the corrections made (Point1, Point2, etc.).

3. I consider that the Conclusions section can be improved a little more.

Regarding the use of English, I recommend reviewing all the content again, to correct some typos.

Author Response

I would like to express my gratitude  for the time and effort that had been spent in handling our manuscript. For your convenience, all of the corrections/changes are highlighted in RED in the revised manuscript.Please see the attachment.

Author Response File: Author Response.docx