Masked Face Recognition System Based on Attention Mechanism
Round 1
Reviewer 1 Report
1. In the abstract section, the author must first elaborate on all the abbreviations.
2. Figure 4 is not clear and needs to revise.
3. Please improve Figure 7
4. There is no way to understand how the proposed method is performing from Figure 8. Please fix it.
5. Conclusion and future workers need to be more meaningful.
6. There are many statistical measurements, and I am not sure why author uses only one performance evaluation.
Author Response
1. In the abstract section, the author must first elaborate on all the abbreviations.
Answer：Thank you for your suggestions. We have corrected this part.
2. Figure 4 is not clear and needs to revise.
Answer：Thank you for your suggestions. We've redrawn figure 4.
3. Please improve Figure 7
Answer：Thank you for your suggestions. We've redrawn figure 7. Adding smooth curves made the image look cluttered, so we removed it.
4. There is no way to understand how the proposed method is performing from Figure 8. Please fix it.
Answer：Thank you for your suggestions. We've redrawn figure 8. We designed a frame diagram to make it easy to understand.
5. Conclusion and future workers need to be more meaningful.
Answer：Thank you for your suggestions. We have rewritten part of the conclusion.
6. There are many statistical measurements, and I am not sure why author uses only one performance evaluation.
Answer：Based on your suggestions, We added three new metrics, including Precision Recall and F1-score.
Reviewer 2 Report
Face recognition applications are very popular and useful, and the masked face recognition system is much more complex. Therefore, the authors rightly investigated the possibility of using an attention mechanism to improve the accuracy of face recognition. The Facenet network model was compared with the new ConvNeXt-T model which uses the ECA attention mechanism. In addition, a large set of faces wearing masks was constructed. In addition to these advantages, several issues need to be corrected as below.
1. The authors emphasize that through experiments, the model proved to be 99.76% accurate for real faces wearing masks. However, the values of other metrics: F1-score and AUC ROC were not provided. What was the average training time for the new model?
2. All mathematical variables should be written in italics, eg N and L (below equation (2)).
3. Spaces are often missing, e.g. "countries worldwide. It has advantages" => "countries worldwide. It has advantages" in Itroduction. This applies to the lack of spaces after a period in a sentence, as well as when quoting.
4. Conference names should be capitalized, e.g. page 11: "Proceedings of the IEEE/CVF conference on computer vision and pattern recognition," => "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,".
Only after the above deficiencies have been diligently corrected, the manuscript can be published.
Author Response
1. The authors emphasize that through experiments, the model proved to be 99.76% accurate for real faces wearing masks. However, the values of other metrics: F1-score and AUC ROC were not provided. What was the average training time for the new model?
Answer：Based on your suggestions, We added three new metrics, including Precision, Recall and F1-score.And we have filled in the training time of each epoch in the article.
2. All mathematical variables should be written in italics, eg N and L (below equation (2)).
Answer：Thank you for your suggestions. We have corrected this part.
3. Spaces are often missing, e.g. "countries worldwide. It has advantages" => "countries worldwide. It has advantages" in Itroduction. This applies to the lack of spaces after a period in a sentence, as well as when quoting.
Answer：Thank you for your suggestions. We have corrected this part and checked it at the same time.
4. Conference names should be capitalized, e.g. page 11: "Proceedings of the IEEE/CVF conference on computer vision and pattern recognition," => "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,".
Answer：Thank you for your suggestions. We have corrected this part and checked it at the same time.
Round 2
Reviewer 1 Report
The paper can be accepted in its current format.
