Real-Time Face Mask Detection to Ensure COVID-19 Precautionary Measures in the Developing Countries

Round 1

Reviewer 1 Report

For better outcomes evaluations, multiple deep learning architectures were used in this research. Following rigorous testing, we chose a bespoke model that has the best performance in determining whether or not people are wearing a face mask while still being simple to deploy on a small device like the Jetson Nano. After applying several masks to those photographs, the experimental evaluation is performed on the custom dataset created from the website (https://thispersondoesnotexist.com/). In compared to existing methods, the proposed model provided higher accuracy (99 percent for training accuracy and 99 percent for validation accuracy). Furthermore, the proposed method can be used on devices with limited resources. overall the paper is in good format. some comments are required to address:

The main contributions are not clear. Author need to write main contributions in introduction section. Further, the main challenges are also need to write.

The methodology needs more explanation according to the flow of the main architecture.

Author need to write a short paragraph in the literature review section that how the proposed work filled the research gap.

Need to write future research directions in the conclusion section. 

 

 

Author Response

Point #1: The main contributions are not clear. The author needs to write the main contributions in the introduction section. Further, the main challenges also need to write.

Response: We are thankful to the reviewer for the constructive feedback. We totally agree with the reviewer on this feedback concerning the contributions of the work. We incorporated the mentioned comments in the revised manuscript which can be observed by track changes and for ease of the reviewer, the changes are as follows.

Face-mask detection using ML and DL achieved promising results, however, the existing methods need to be more robust in a real-time environment. Furthermore, the lack of suitable datasets and DL-based models for facemask detection is still in progress [12], that can detect a person without a face-mask in real-time to avoid the spread of COVID-19. For this purpose, we create a large-scale dataset that consists of two classes i.e. face-mask and not-mask, and also developed a lightweight CNN model for efficient face-mask detection to provide a safe and secure environment for every individual by minimizing virus spread. The proposed work will currently be implemented in targeted areas such as schools, colleges, universities, mosques, and super-stores. It will be implemented on the main entry point, where the system will check each individual's face. If the system detected any person without a mask, they would not be allowed to enter. The contributions of the proposed work are as:

1. A new face dataset is developed containing images of different people generated by a generative adversarial network and then designed different face masks that are applied to these face images to create a custom dataset consisting of two classes (mask and non-mask).
2. A lightweight DL-based method consisting of four convolutional, one fully connected, and one output layer has been developed to accurately detect face masks.

• To validate the performance of the proposed lightweight DL model over the custom dataset, we trained different pre-trained DL models such as AlexNet, VGG16, VGG19, ResNet101, and NesNetMobile, MobileNet. The proposed model outperforms the existing DL-based models in terms of a better balance among time complexity, model size, and accuracy.

 

Point #2: The methodology needs more explanation according to the flow of the main architecture.

Response: According to the mentioned comment, we updated the manuscript and add some details to explain the flow of the proposed work, which can be observed by track changes, and for ease of the reviewer, the changes are as follows.

Face-mask detection using ML-based methods is tedious and time-consuming work due to ML requiring hand-crafted features engineering which required domain experts. Particularly, in ML-based methods, early face-mask detection and response generation are also challenging due to low detection accuracy. Considering these challenges, DL-based models are efficiently utilized for face-mask detection in a surveillance system. DL provides end to end feature extraction mechanism but it requires a huge amount of training data and high computational power. The main motive behind this work is to enhance the performance of the DL-based model, and deployability over edge devices, for which, we proposed a new framework as shown in Figure 1. The proposed framework is divided into two major steps such as 1) face cropping and 2) face-mask detection.

Point #3: The author needs to write a short paragraph in the literature review section that how the proposed work filled the research gap.

Response: We thank the reviewer for the constructive feedback concerning the literature review section that how the proposed work filled the research gap. We updated the literature review section of the revised manuscript which has significantly improved the quality of the work.

In the light of the literature review, several researchers used different mechanisms for face mask detection and also achieved superior results, however, the limited amount of training data and computational complexity restrict their systems from a real-time implementation. Therefore, firstly we create a novel large-scale dataset that consists of two classes i.e. face-mask and not-mask. Secondly, we developed a lightweight CNN model that can be easily deployed over resource-constrained (edge devices) to efficiently and accurately detect face masks in a real-time scenario.

 

Point #4: Need to write future research directions in the conclusion section.

Response: The conclusion section of the revised manuscript is updated by adding future research directions.

In the future, we can improve the proposed work by using a large amount of data and can also be extended to classify the type of mask and implement a facial recognition system, deployed at various workplaces to support person identification while wearing the mask. We are aiming to test the proposed model in a hazy and foggy environment in the future, where face detection can be challenging.  We will use a generative adversarial network to enhance the quality of a given image for better performance evaluation. Furthermore, the proposed work can also be extended by using efficient object detection algorithms such as the various versions of Yolo, Faster-RCNN, etc. which can efficiently crop faces from the given input images and recognize them correspondingly whether the person wears a face mask or not in one blended step.

Reviewer 2 Report

Please see attached document

Comments for author File: Comments.docx

Author Response

Point #1: In table 4 the authors need to look at the Results (Accuracy) parameter. According to the authors the accuracy ranges from 0.95% to 0.99% with the proposed being 0.98%. All of these numbers are less than 1%. I believe the authors meant to put 99%, however, this needs to be addressed and/or looked into. If this is true the accuracy is very low.

Response: We are thankful to the reviewer for identifying this point. We totally agree with the reviewer on this feedback concerning the accuracy parameters. We incorporated the mentioned comment in the revised manuscript by correcting the values.

Table 4: Comparison with different deep learning architectures on custom drowsiness detection dataset

Technique	Dataset	Model Size	Parameters (Million)	Accuracy (%)	Training Time (M:S)	FPS (CPU)
Proposed	Custom dataset	16 MB	2.2	98.47	30:38	28.07
AlexNet	-	233 MB	60	98.07	34:04	5.33
VGG16	-	528 MB	138	98.75	35:77	2.98
VGG19	-	574 MB	143	99.22	37:35	1.83
ResNet101	-	98 MB	20	99.00	55:13	7.43
MobileNet	-	13 MB	4.2	95.13	34:27	20.23
NesNetMobile		23 MB	5.3	97.05	44:13	14.22

Point #2: In the introduction, I think it is appropriate to cite a very recent report by Bussan et al. on masks that propose that masks should pass strict quality control to ensure safety in addition to all the other benefits they offer: Bussan, D. D., Snaychuk, L., Bartzas, G., & Douvris, C. (2021). Quantification of trace elements in surgical and KN95 face masks widely used during the SARS-COVID-19 pandemic. Science of the Total Environment, 151924.

Response: Most recent papers have been cited in the introduction section of the revised manuscript.

Point #3: 3. I do believe the authors should show a figure where the proposed model fails. It would be more interesting to the readers to see a visual of an incorrectly classified person wearing or not wearing a mask, and a discussion of why this may have occurred. Without a visual of an incorrectly labeled classification, it’s hard to see the reason behind the failure.

Response: The results of the proposed method are reflected in the confusion matrix shown in table 2. Moreover, the proposed model can be challenged in a hazy and foggy environment, where the face detection algorithms cannot capture the faces clearly. We are aiming to use a generative adversarial network to enhance the quality of a given image for better performance evaluation.

Reviewer 3 Report

Summary 

The manuscript entitled “Real-Time Face Mask Detection to Ensure Covid-19 Precautionary

Measures in the Developing Countries” by Farman et al. shows the development of a novel to monitor and determine people with masks and no-mask.

 

 

General comments 

In general, the work is accurate and clearly presented, and the results are of interest to readers of Applied Sciences. The article style is correct, but it should be reviewed in a few points. Thus, I believe that the text needs some technical adjustments to be published. Therefore, I recommend that this manuscript can be published in Applied Sciences after minor revision. 

 

 

 

Specific comments

 

The article's grammar, punctuation, and style are not completely adequate, and the manuscript needs to be proofread if the authors are willing to publish this manuscript in a reputable journal like Applied Sciences.

 

 

Going into details on the specific issues, here some comments are reported:

 

- 1. Introduction. A few notions included here are not very connected to the topic. It s more a collection of general claims spinning around the COVID-19 pandemic (e.g., precautions to fight against COVID-19)

 

- 2. Literature Review. It is quite odd seeing this kind of section here. It seems a sort of Mini-Review into an original article. Please remove it

 

- It is not clear why this method should be useful for ’’developing countries’’ only. Please clarify it.

 

 

 

-  Novel face masks with different features, including structure and properties, are under development [https://doi.org/10.1002/chem.202004875]. Does the proposed method have the possibility of being applied for a possible next-generation face mask?

 

 

 

 

Conclusion

 

The topic of this manuscript falls within the scope of Applied Sciences. I like the concept proposed in this paper, anyway I think the manuscript needs a few improvements. I believe the article is of sufficient quality and novelty to meet the Applied Sciences publication standards after a minor revision.

 

 

 

 

 

 

 

Author Response

Point #1: The article's grammar, punctuation, and style are not completely adequate, and the manuscript needs to be proofread if the authors are willing to publish this manuscript in a reputable journal like Applied Sciences.

Response: The revised manuscript has been thoroughly checked for grammatical and spelling mistakes.

Point #2: Introduction. A few notions included here are not very connected to the topic. It’s more a collection of general claims spinning around the COVID-19 pandemic (e.g., precautions to fight against COVID-19).

Response: We are very thankful to the reviewer for identifying this point which has certainly improved the quality of the overall paper. We removed the unstructured sentence and more general information from the manuscript.

Point #3: Literature Review. It is quite odd seeing this kind of section here. It seems to be a sort of Mini-Review of an original article. Please remove it.

It is not clear why this method should be useful for ’’developing countries’’ only. Please clarify it.

Novel face masks with different features, including structure and properties, are under development [https://doi.org/10.1002/chem.202004875]. Does the proposed method have the possibility of being applied for a possible next-generation face mask?

Response:

Yes, the proposed method has the possibility of being applied for a possible next-generation face mask, we just need a few images from the next-generation face mask for model training, after the training we can easily detect it.