Research on an Intelligent Driving Algorithm Based on the Double Super-Resolution Network

Round 1

Reviewer 1 Report

1. The network design structure diagram is not detailed enough, the network structure is not fully reflected and is not clear enough;
2. The loss function is not introduced in detail and cannot correspond to the network structure. More details about the loss function are needed.
3. How to get the real segmentation graph is not mentioned in the paper;
4. Please check the incorrect formula label format;
5. The specific process of network data transmission is not introduced in detail, so we can't see how the network works. More details about the network data transmission are needed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a network model for training high-resolution altas, which improves the segmentation accuracy without adding additional computation, based on dual super-resolution learning network (DSRL). The proposed network consists of two parts, one part is the super-resolution network, the other is the high-resolution picture convolution network, and the internal convolution is used to replace the partial convolution, which not only reduces the network parameters but also improves the segmentation accuracy. My comments on this paper are summarized below:

• The network structure of the proposed design shown in Figure 5 is not easy to be understood. More description should be included in Figure 5 to illustrate the architecture of the proposed network, especially the components of the network. In addition, the readability of the network shown in Figure 5 is not good, which needs further revision.
• Different backbone models are adopted in the evaluation of the proposed network, and the CSPdarknet53 exhibits the best performance. It is suggested to describe the reason why.
• It is suggested to evaluate the processing speed of the proposed network when realizing in a computing platform to meet the real-time processing applications.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Paper covers interesting issues, paper contains interesting results. Here below I provide my remarks:

1. Paper misses scientific formulation of research task, therefore it is problematic to decide about achievement of the research aim.
2. Conclusions looks weak for such amount of results, lack of task makes them difficult to formulate. I would ask them to rewrite.
3. Fig. 2 unclear in the graphics form and its logic meaning. Please, remake it and explain or remove.
4. Fig. 3 needs scaling and explaining, because presents no illustration in present form.
5. Fig. 4 is unreadable - please, rethink graphical presentation and explain it. In present form it is not acceptable.
6. Fig. 5 needs the same things, legend, explanations, graphics quality.
7. Fig. 6 unclear totally, its meaning lost due to bad quality.
8. Plenty of terms and abbreviations in the text unexplained or explanation distributed in the text; nomenclature would solve that entirely.
9. Figures 7 and 8, containing key results, absolutely lost. Also, no explanation, connected with pictures, available.
10. Conclusions simply need to be rewritten.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Reviewer 2 Report

Since the authors have addressed the issues I listed in the review comments and modified the manuscript accordingly, I agree this paper to be published in the current form.