Thermodynamics-Inspired Multi-Feature Network for Infrared Small Target Detection

Round 1

Reviewer 1 Report

This paper proposes a thermodynamics-inspired multi-feature Network for the IRSTD task. This network uses a TSB module directly optimize the results of the encoder inspired by Hamming equation of the thermodynamic. The work of this paper has certain innovation and practical value. On the whole, I think this paper can be published after a major revision.

 1 In section 3.1, the introduction of the C1, C2, F1, F2, N1, N2 in equation 1-3 should be clearer.

 2 The Hamming method of thermodynamics should be explained in more detail. What thermodynamic problems is it mainly used to address? What functions can it realize and why can it be used as a reference in the network described in this article?

 3 In equation 19-21, the authors should explain why the difference between input and output results is used to replace f in equation 18, and why h is chosen as 1.

 4 In equation 23-25, the it needs to be pointed out what the letter N represents.

 5 In 4.1.3, How are the hyperparameters determined in the network training process?

 6 I suggest the authors to add some illustrative comparisons in ablation study.

 This article contains too many long sentences. It is suggested to revise the expression to improve the readability of the article.

Author Response

The authors extend our sincere gratitude for your approval and invaluable suggestions. Your insightful feedback has been thoroughly considered, and we have made every effort to respond to each query raised in the review report.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript proposes a novel network, TMNet, for  infrared small target detection. This work introduces the AFCE module, which incorporates a novel attention mechanism and multi-scale cross-layer feature interaction mechanism to aid the network in effectively extracting valuable target information. Then the TSB module is proposed, which utilizes a super-resolution branch that corresponds to each layer of the semantic segmentation network to assist the model in learning high-resolution details in infrared images. They further show that the TMNet outperforms existing models in terms of objective evaluation metrics and visual quality on the NUAA-SIRST dataset.

There are some problems, which must be solved before it is considered for publication. If the following problems are well-addressed,  the essential contributions of this paper are important for infrared small target detection.

 (1) The abstract should end with the paper's contribution to the field of study.

(2) The “CFF” and “DMA” on line 112 and 113 is not defined in the main text of the paper.

(3) The logic of writing in Section 2.3 of the paper is confusing, and it is recommended to reorganize the writing. For example, "thermal particles always tend to move from an unstable high temperature state to a more stable low temperature state"—how does it relate to the topic of this paper? How does it relate specifically to super-resolution?

(4) The “ODE” on line 216 is not defined in the main text of the paper.

(5) It is recommended to add the hardware information of the experiment, such as the graphics card model, in Section 4.1.3.

(6) It is recommended to add "(ours)" after TMNet to increase readability when comparing with different models in tables and figures, such as Table 1 and Figure 5.

(7) In Figure 5, the deep learning model used for comparison suggests choosing FC3-Net instead of ALCNet because of its better performance. Moreover, the paper below also achieves good performance for infrared small target detection through attention mechanism as well which should be added for the comparison experiments.

[1] APAFNet: Single-Frame Infrared Small Target Detection by Asymmetric Patch Attention Fusion

(8) There is such a sentence in the conclusion: “In addition, we also observe that the super-resolution features of infrared images contain detailed information that is not present in low-resolution features.” Please show "detailed information in the super-resolution features of infrared images” in a visual way in the results of the paper.

 Moderate editing of English language required.

Author Response

The authors extend our sincere gratitude for your approval and invaluable suggestions. Your insightful feedback has been thoroughly considered, and we have made every effort to respond to each query raised in the review report.

Author Response File: Author Response.docx

Reviewer 3 Report

See the attached file.

Comments for author File: Comments.pdf

Check typos and spells.

Author Response

The authors extend our sincere gratitude for your approval and invaluable suggestions. Your insightful feedback has been thoroughly considered, and we have made every effort to respond to each query raised in the review report.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I think the author has made revisions based on my initial review comments, and I recommend accepting it.

Author Response

Reviewer 2 Report

All of concerns have been revised and the paper is now suitable for the publication. Thank you.

 Minor editing of English language required

Author Response

Firstly, the authors extend our sincere gratitude for your approval and invaluable suggestions. Your insightful feedback has been thoroughly considered, and we have made every effort to respond to each query raised in the review report.

Author Response File: Author Response.docx

Reviewer 3 Report

You revised according to our comments.

However, the followings should be considered.

1. There is no clear reference in the hamming method. Your mentioned paper does not have the same equation.

2. The coefficients of the modified equation can affect to the results. Check this again. 

Just check typos.

Author Response

Firstly, the authors extend our sincere gratitude for your approval and invaluable suggestions. Your insightful feedback has been thoroughly considered, and we have made every effort to respond to each query raised in the review report.

Author Response File: Author Response.docx