STFTrack: Spatio-Temporal-Focused Siamese Network for Infrared UAV Tracking
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors The manuscript entitled "STFTrack: Spatio-temporal focused siamese network for infra-2red UAV trackin" has been investigated in detail. The topic addressed in the manuscript is potentially interesting and the manuscript contains some practical meanings, however, there are some issues which should be addressed by the authors:
1. In the first place, I would encourage the authors to extend the abstract more with the key results. As it is, the abstract is a little thin and does not quite convey the interesting results that follow in the main paper.
2. The Introduction section needs a major revision in terms of providing more accurate and informative literature review and the pros and cons of the available approaches and how the proposed method is different comparatively. Also, the motivation and contribution should be stated more clearly.
3. At the end of the introduction, clearly state the art of the work.
4. See more recent articles on drones.
5. Why is no validation for experimental and numerical results with other valid results provided in the text of the manuscript?
6. The analysis on the results is weak. Analyze the results more fully.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper proposes a two-stage tracking framework based on a two-level target focusing mechanism for applications of UAV tracking. The paper is well organized and written. I have the following minor comments, just for reference.
1. The authors should highlight the novelties of their work in the introduction. The last paragraph in the current version is more like a mix of novelties and paper organization. Using separate enumerations for novelties may be a good practice.
2. Please pay attention to some of the symbols. In Eq. (8) (and Eq. (16)), “dots” are used to connect variables. Since most of the other equations omit the “product” symbol, this “dot” here are more likely mean “dot product” – is this true? Otherwise, it is misleading to use “dot” here. Talking about “otherwise”, the reviewer feels that it is not necessary to use italic font for the condition words in Eq. (6), since they are just text instead of variables. Please also note, “\lambda” are written with mixed italic and non-italic formats.
3. How are the merits in Table 2, i.e., precision, success, and accuracy, calculated? The authors might as well give brief explanations or equations accordingly.
4. Tracking stability is of much concern as the context of this research. What can be an evaluation merit for it?
5. The conclusion can be enriched. Analogous to the first comment, the novelties should again be highlighted. It is also suggested that some of the important results are summarized quantitatively therein, as a proof to show that your method is superior.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for Authors1. Figure 1. In the SIAMESE FEATURE EXTRACTION section, the display of drone recognition results in the Query Image is the same as that of the Instance Discrimination results. It cannot reflect the results after the target tracking.
2. Figure 3 mainly shows the process of predicting the target location through the integration of space -time information. (D) According to the result of the author who wants to display the fusion, it can be seen that the second threshold chart area of (d) is more concentrated, but (but (but (but (but (but (but (but (but (but ((D), D) The first picture of the frame is just a decrease in the candidate area of the frame. The image has not actually seen the change, and the effect of fusion cannot be displayed.
3. The 3.3 and 3.4 parts of this article show the accuracy and success rate of the nine algorithms on the same data set, and only the tracking effect of five algorithms is displayed in the 3.5 part. It is recommended to complete the tracking effect of several other algorithms so that More intuitively compare the superiority of this experimental algorithm.
4. Figures in Figure 9 can only see the identification accuracy display of different algorithms on infrared imaging, and cannot reflect the continuous tracking effect. And in some diagrams, I cannot see that there are drones in the real area, and the same group of images lack coherence and cannot reflect the tracking process.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
Comments and Suggestions for Authors
The quality of the paper and the Images used for the explanation is good.
The paper proposes a new method called STFTrack for infrared UAV tracking using a spatiotemporal focused siamese network. The method utilizes a two-level focusing strategy from global to local to handle the issues of tracking instability generated by environmental thermal crossover and similar distractors. It also uses a spatiotemporal information-guided region proposal network (RPN) to focus on the search region and generate high-quality candidates. Additionally, it uses metric learning to design an instance discriminative RCNN to focus on the target UAV among candidates. The authors conduct extensive experiments on Anti-UAV and LSOTB_TIR datasets to demonstrate the effectiveness of the proposed method compared to other advanced trackers.
Overall, the paper is well-structured and clearly written, providing a suitable background and motivation for the proposed method. The authors clearly explain the challenges associated with infrared UAV tracking and how their proposed method addresses these challenges. The paper also provides a thorough review of related work and how the proposed method improves upon existing methods.
However, there is a scope for improvement. The abstract could benefit the readers with clear information concerning the results achieved through the proposed method. Further, the paper could benefit from more detailed information regarding the experiments conducted, such as the experimental setup and performance metrics used. Finally, the paper could benefit if the authors discuss the limitations of the proposed approach and areas for future work.
The conclusion needs to be a bit more extensive. The present conclusion is too short.
Overall, the paper is a good contribution to the field of anti-UAV technology and infrared UAV tracking, and the proposed approach shows promise for enhancing the accuracy and robustness of infrared UAV tracking.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAll my comments have been thoroughly addressed. It is acceptable in the present form.
