Constructing Adaptive Multi-Scale Feature via Transformer-Aware Patch for Occluded Person Re-Identification
Round 1
Reviewer 1 Report
I have carefully reviewed this manuscript and below is my decision.
- Reference and your citations don't appear in the paper.
- Conclusion section should be improved.
It can be published after corrections are made.
Author Response
Please see attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Q1. The title does not contain any flavour of Machine Learning. However keywords have the term Weakly-Supervised Learning.
Q2. What the difference of Weakly-Supervised Learning and Machine Learning?
Q3. “Transformer has achieved remarkable success in the Nature Language Processing”- Justify it.
Q4. How cropping of visible pedestrian is made?
Q5. ViT is a Transformer-based network structure commonly used for image classification and feature extraction.- Explanation with practical example is required.
Q6. “we think we can it assumed that the approximate distribution of the visible parts of pedestrians in the original input image can be firstly searched “- The sentence does not have any meaning.
Q7. Why collected attention points are required to observe?
Q8. Define weakly supervise the learning.
Q9. Discuss the disadvantages of the proposed method.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The main research question in this paper concerns Dual-Transformer symmetric architecture Authors proposed :
A Transformer-Aware Patch Searching (TAPS) module
An Adaptive Visible-Part Cropping (AVPC) Strategy
Methodology includes some experiments
The Paper is relevant and interesting
However,
1) Please, specify precisely in the title of section 3 , what approach will be considered. Do the same for section 3.1, please, do not start from Figures. Do the same for section 3.3.
2) Please, insert an introduction to section 3 , before the Figure
3) Please, make shorter the Figure 1 title , and all included there explanations, please, move to the text of section 3
4) Eq. 1 is not an equation
5) Please, reduce the Title of Table 1. All explanation on this comparison, please, insert in the section text (section 4.3) . The same for section 4.4.
6) Please, discuss weaknesses of the proposed approach (may be in Section Conclusions)
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The ppaer has been radically improved in comparison with the previous version.
Now the main question is well addressed. The paper is wel written, conclusions are consistent with the evidence and presented arguments. Therefore, I recommend this paper for publication.
