Micro-Expression Spotting Based on a Short-Duration Prior and Multi-Stage Feature Extraction
Round 1
Reviewer 1 Report
A Micro-expressions (MEs) spotting network called AEM-Net (Adaptive Enhanced ME Detection Network) is proposed in this paper. Multiple experiments are performed to show that it reaches better accuracy than state-of-the-art
1. The introduction of the proposed model is confusing as there are too many terms, e.g. which part is spatial attention mechanism and which part is channel attention mechanism?
Figure 1 can be optimized to highlight the four major parts and their key components.
2. More details about the dataset and evaluation process should be provided.
(1) line 298 how many samples are included in the dataset SAMM-L
(2) line 359 what is LOSO learning? leave one subject out?
3. Even though the model proposed by the authors largely improves the result, the overall performance is still limited: F1 < 0.5, the authors can add some discussion in the conclusion section
4. Paper needs to be reviewed to fix typos, and grammar issues e.g.
(1) line 297
(2) figure 2 caption 'crop proposal'
(3) line 302, 'figure 7 shows our dataset'
Author Response
Thank you for your thoughtful review of the manuscript. We appreciate the valuable comments and suggestions from the reviewers and the Associate Editor. These comments have helped us to greatly improve our manuscript, and provided important guidance for our future research. The authors try to revise the manuscript to address your concerns. The detailed responses to the reviewer1’ comments are as follows.
Comments 1
- The introduction of the proposed model is confusing as there are too many terms, e.g. which part is spatial attention mechanism and which part is channel attention mechanism?
Figure 1 can be optimized to highlight the four major parts and their key components.
Reply: Many thanks for the reviewer’s insightful comment. In Subsection 3.1 of the revised manuscript, according to your comment, authors add the spatial attention module and channel attention module in the Figure 2 and highlight the four major parts. Moreover, the detail structures of these parts are elaborated in following sections, which are explained in the title of Figure 2.
Comments 2
- More details about the dataset and evaluation process should be provided.
(1) line 298 how many samples are included in the dataset SAMM-L
(2) line 359 what is LOSO learning? leave one subject out?
Reply: (1) In subsection 4.1, authors compensate for the description on the samples in the dataset SAMM-LV:
Meanwhile, SAMM-LV includes 343 MaEs and 159 MEs in 147 long videos from 32 subjects with 200 FPS.
(2) In line 352 of the revised manuscript, the LOSO is explained by ‘the leave one subject out (LOSO) learning method’.
Comments 3
- Even though the model proposed by the authors largely improves the result, the overall performance is still limited: F1 < 0.5, the authors can add some discussion in the conclusion section.
Reply: In the section 5 of the revised manuscript, authors add some discussion:
However, due to the brief and subtle nature of ME, the F1 score of our ME spotting is not satisfactory yet for real-world applications. Powerful feature learning and action localization in an end-to-end ME spotting framework are promising directions in future research.
Comments 4
- Paper needs to be reviewed to fix typos, and grammar issues e.g.
(1) line 297
(2) figure 2 caption 'crop proposal'
(3) line 302, 'figure 7 shows our dataset'
Reply: A native English-speaking colleague checked the entire manuscript to help us improve the English expression. In addition, authors double checked these revisions to fix typos and grammar issues.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Figure 1 should be closer to the reference, e.g. around line 54.
In line 87 there should be an introduction sentence explaining the division of the review.
Figure 2 should be on page 5.
Figure 3 is not that large to justify its horizontal layout.
Please re-check your work for editing (Language errors):
Line 52: "an expression. the apex frame"
Line 129: "micro-expression contest, they also introduced" -> "micro-expression contest. They also introduced"
Line 163: "and then extract features" -> "and features are then extracted"
Line 189: "optical flow maps. fOn"
Line 261: "In parallel, of interest is the ..." -> At the same time, the channel relationship ..."
Line 276: "the 1D samples. a dropout operation"
Line 288: "Owing to this operation, our multi-scale localization process to locate MEs more accurately" -> "Through this operation, our multi-scalar localization process allows for a more accurate location of ME"
Line 302: "Figure 7 shows our" -> "Figure 8 shows our"
Line 356: "shown in Figure 8" -> "shown in Figure 9"
After reviewing in the library the article identified by the second reviewer, I can confirm that this article can be considered self-plagiarism. The same editorial errors are even made in both articles.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
There is reasonable suspicion from autoplagiarism.
https://link.springer.com/chapter/10.1007/978-3-031-20233-9_29
Since the overall merit of the article is practically the same as in case of previously published one by same authors (see link) I would recommend rejection of the manuscript.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Until authors provide permission from the publisher of an earlier article to republish some of the material on an open access basis, publication in this journal is not possible. This is clearly stated in the following rules:
"If you include already published figures or images, please obtain the necessary permission from the copyright holder to publish under the CC-BY license."
This is probably possible through the Rightslink system.
Also, the abstract and introduction should make it clear that this is an extension of a previous publication.
It is also good practice to have the earlier publication added for the reviewers, in "Non-published Material"
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
I would recommend to explicitly mention within the abstract and introduction, that the article is extended version of conference paper (with corresponding refference), since there is a large amount of exactly the same phrases and paragraphs, including figures as well.
In the Experiments chapter, especially within the Results and comparisons section, I would strongly recommend to highlight the changes and extensions in comparison with the original conference paper. Same comparison should be performed within the conclusion chapter.
In addition I would recommend minor changes:
- check the grammar to avoid typos (e.g. Fig. 1 - original video lenth ),
- rotate figure 3 for better readability,
- authors may reconsider splitting the graphs in fig.4, and present the investigation of suitable detector in one graph,
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 3
Reviewer 2 Report
The sentence introduced in the abstract (lines 4-5) is not linguistically correct.
Author Response
Response Letter
Dear Associate Editor and Reviewers
Thank you for your thoughtful review of the revised manuscript. The authors try to revise the manuscript to address your concerns. The detailed responses to the reviewer comments are as follows (The modification parts are marked using the Track Changes” in the revised LaTeX text):
Comments
The sentence introduced in the abstract (lines 4-5) is not linguistically correct.
Reply: We appreciate your careful comment. The sentence is corrected by:
A ME spotting network called AEM-Net (Adaptive Enhanced ME Detection Network) is proposed, which is an extension of the conference paper presented at the Chinese Conference on Biometric Recognition (CCBR).
Sincerely,
Jiangxi Science and Technology Normal University
Prof. Zhihua Xie
E-Mail: [email protected]
Author Response File: Author Response.pdf
