A Garbage Classification Method Based on a Small Convolution Neural Network

Round 1

Reviewer 1 Report

Dear Authors,

thank you for this good study and presentation of your research.  

Author Response

Dear reviewer:

We greatly appreciate your review of this research manuscript! Many thanks for your positive comments. It is our great honours to receive your reply.

Best regards!

Reviewer 2 Report

Reviewer Reports:

I recommend a major amendment at this level.

General comments:

The manuscript entitled “A garbage classification method based on a small convolution neural network” was reviewed. The work carried out in the manuscript is interesting and aimed at using CNN to build a garbage classification network, and the model is derived after training the CNN, so that the model can be used to effectively detect domestic garbage. The manuscript has a lot of information however, there is some correction needed before possible publication. It is better to do not to use the first-person's pronoun. Do not use "we, us, or our" throughout the paper. The authors are suggested to proofread the paper with a native English speaker and restructuring of sentences is required for the entire manuscript. What progress against the most recent state-of-the-art similar studies was made? The innovation and the importance of this work are not clearly highlighted in the abstract, introduction and conclusions. Please work on this and prove to us why this work is valuable. The authors do not validate the obtained results and compare them with the other works in the field. The authors need to present how the results can be validated and verified. Other main remarks that in my opinion needs attention are the following:

 

Detailed comments:

Title:

Ok.

Abstract:

The abstract does work well. However, a good abstract should address these issues: what are you trying to do, why, what you found and what is the significance of your findings? In the abstract, please add an indication of the achievements from your study that are relevant to the journal's scope. The abstract should state briefly the purpose of the research, the principal results and major conclusions. In the abstract, please add an indication of the achievements from your study that are relevant to the journal's scope. Please be concise - maximum 1-2 lines.

 

Introduction:

The literature review should clarify the "contribution" of your study. The authors failed to present the study debates and failed to discuss the debates. In general, the authors should present the specific debate for your study. This should more clearly show the knowledge gaps identified and link them to the paper's goals. A high-quality paper has to provide a proper state-of-the-art analysis after the literature review and only based on the analysis to formulate the paper goals. The lack of proper justification creates the wrong impression that the authors are unaware of the recent developments. The relevant reference may be of interest to the author according to below:

https://link.springer.com/article/10.1007/s13399-022-02871-w

https://www.sciencedirect.com/science/article/pii/S2352554122001103

Please eliminate the use of redundant words. Eg. In this way, Recently, Respectively, therefore, currently, thus, hence, finally, to do this, first, in order, however, moreover, nowadays, today, consequently, in addition, additionally, furthermore. Please revise all similar cases, as removing these term(s) would not significantly affect the meaning of the sentence. This will keep the manuscript as CONCISE as possible. Please check ALL. Avoid beginning or ending a sentence with one or a few words, they are usually redundant. Kindly revise all.

 

Materials and Methods:

Ok

Results and Discussion:

All the findings of the current work need to be compared and discussed with the results of other researchers finding instead of having a general comparison with other researchers' works. The authors should perform a comparison between the forecasting results. In your discussion section, please link your empirical results with a broader and deeper literature review.

Conclusions:

Please make sure your conclusions section underscores the scientific value-added of your paper, and/or the applicability of your findings/results. Highlight the novelty of your study.

References:

Please check the reference section carefully and correct the inconsistency.

Author Response

Dear reviewer:

We appreciate you very much for constructive comments and suggestions on our manuscript. These suggestions allow us to optimize this study more comprehensively.  We have responded to each of your valuable suggestions.

Point 1:

General comments:

The manuscript entitled “A garbage classification method based on a small convolution neural network” was reviewed. The work carried out in the manuscript is interesting and aimed at using CNN to build a garbage classification network, and the model is derived after training the CNN, so that the model can be used to effectively detect domestic garbage. The manuscript has a lot of information however, there is some correction needed before possible publication. It is better to do not to use the first-person's pronoun. Do not use "we, us, or our" throughout the paper. The authors are suggested to proofread the paper with a native English speaker and restructuring of sentences is required for the entire manuscript. What progress against the most recent state-of-the-art similar studies was made? The innovation and the importance of this work are not clearly highlighted in the abstract, introduction and conclusions. Please work on this and prove to us why this work is valuable. The authors do not validate the obtained results and compare them with the other works in the field. The authors need to present how the results can be validated and verified.

Response 1: We are extremely grateful to you for pointing out these problems.

1. Do not use "we, us, or our" throughout the paper. We agree and have replaced the above words with "this research" and "this paper" in the article.
2. We have invited professionals to review the language of the manuscript and have made some language revisions to the manuscript. For example, we modified some sentences in this manuscript to passive voice to better express the meaning of the sentences.
3. We thank you for your suggestions on the innovative issues of manuscripts. Due to the imperfect writing of some parts of our article, the research contribution in the manuscript is not obvious enough. We have made some changes in the abstract and conclusion parts of the manuscript. The main contributions of this study include three aspects. (1) We propose a garbage image preprocessing method. An adaptive image brightening algorithm is developed to average the brightness of the background, and the threshold replacement method is used to reduce the shadow noise. Then the Canny operator is used to assist in cropping the blank background in the image. (2) We design a simplified CNN model based on the MLC-CNN structure, making it simple and efficient. Experimental results show that the model can achieve 96.77% accuracy on the self-built dataset and 93.72% accuracy on the TrashNet dataset, which is higher than the accuracy of MLC-CNN. (3) We also found that the Adamax optimization algorithm based on Adam algorithm has better optimization network effect.

Point 2:

Abstract:

The abstract does work well. However, a good abstract should address these issues: what are you trying to do, why, what you found and what is the significance of your findings? In the abstract, please add an indication of the achievements from your study that are relevant to the journal's scope. The abstract should state briefly the purpose of the research, the principal results and major conclusions. In the abstract, please add an indication of the achievements from your study that are relevant to the journal's scope. Please be concise - maximum 1-2 lines.

Response 2: Thank you for this valuable suggestion. We have carefully revised the content of the abstract. In the abstract of the manuscript, we presented the main solutions and obtained results of this study against some existing problems and compared it with similar studies. We also proposed the use of Adamax optimization algorithm to enhance the optimization effect of the network. At the end, we briefly describe the purpose and social implications of this study.

Point 3:

Introduction:

The literature review should clarify the "contribution" of your study. The authors failed to present the study debates and failed to discuss the debates. In general, the authors should present the specific debate for your study. This should more clearly show the knowledge gaps identified and link them to the paper's goals. A high-quality paper has to provide a proper state-of-the-art analysis after the literature review and only based on the analysis to formulate the paper goals. The lack of proper justification creates the wrong impression that the authors are unaware of the recent developments. The relevant reference may be of interest to the author according to below:

https://link.springer.com/article/10.1007/s13399-022-02871-w

https://www.sciencedirect.com/science/article/pii/S2352554122001103

Please eliminate the use of redundant words. Eg. In this way, Recently, Respectively, therefore, currently, thus, hence, finally, to do this, first, in order, however, moreover, nowadays, today, consequently, in addition, additionally, furthermore. Please revise all similar cases, as removing these term(s) would not significantly affect the meaning of the sentence. This will keep the manuscript as CONCISE as possible. Please check ALL. Avoid beginning or ending a sentence with one or a few words, they are usually redundant. Kindly revise all.

Response 3: Thank you very much for your advice on literature review. This can help us better optimize the article and increase the background of the article.

Through learning and understanding of the two articles you introduced, we agree that they can deepen the background of the research on garbage classification in this article, and also allow readers to have a deeper understanding of the meaning and background of this article. We have set the two references you recommended as references in this paper and have cited them in lines 31 and 38 of the manuscript. We have carried out a detailed analysis and description of the development and analysis of this research in the second part of the manuscript, "Development Status and Analysis", expressing the purpose and advantages of this research work.

We also greatly appreciate your suggestion to eliminate redundant words, which will benefit the normative nature of our articles. We have examined these redundant words throughout the text and revised sentences to eliminate them.

Point 4:

Results and Discussion:

All the findings of the current work need to be compared and discussed with the results of other researchers finding instead of having a general comparison with other researchers' works. The authors should perform a comparison between the forecasting results. In your discussion section, please link your empirical results with a broader and deeper literature review.

Response 4: Thank you very much for your suggestion in the Results and Discussion, it was very inspiring for us. In this paper, we focus on analyzing and comparing the effects of image preprocessing optimization methods and network optimization algorithms on improving the accuracy of garbage classification. Because most relevant researches have enhanced data during image preprocessing, resulting in inconsistent image data and parameters input by convolutional neural network for training, we can only make small-scale comparison with similar experiments. However, through relevant practices, we have verified their good practicability, and modified the comparison results to make them more detailed. (Change on line 357.)

Point 5:

Conclusions:

Please make sure your conclusions section underscores the scientific value-added of your paper, and/or the applicability of your findings/results. Highlight the novelty of your study.

Response 5: Thank you very much for your suggestion on writing the conclusion. We think your suggestion can highlight the novelty of this study in the conclusion description. We describe the main progress and results of this study in the conclusion section of the manuscript, showing the garbage classification effect that the CNN model of this study can bring. At the same time, we emphasize the improvement of the garbage classification effect brought by the image preprocessing method in this paper, and proposed that the Adamax optimization algorithm to improve the network performance better. We also briefly describe the broad applicability of the CNN model and its social implications.

Point 6:

References:

Please check the reference section carefully and correct the inconsistency.

Response 6: Thanks a lot for your advice on checking references. We have checked the references of this manuscript and corrected the format. We modified the text of the journal title and volume to italics, and also thickened the text of year.

We would like to thank the referee again for taking the time to review our manuscript.

Best regards!

Reviewer 3 Report

Speed and time of detection is very important here as we are dealing with millions of garbage items daily, what is the execution time for preprocessing first and for classification later?

 

Why use term luminance in R and Brightness in others G and B

Author Response

Dear reviewer:

We appreciate you very much for constructive comments and suggestions on our manuscript. These suggestions allow us to optimize this study more comprehensively.  We have responded to each of your valuable suggestions.

Point 1: Speed and time of detection is very important here as we are dealing with millions of garbage items daily, what is the execution time for preprocessing first and for classification later?

Response 1: We thank the reviewer for pointing this out. After experimental testing, the time for preprocessing one image is about 1.6 seconds. The CNN model in this paper takes about 0.22 seconds to perform a classification task on an image, and about 2.37 seconds for an array of 1000 image data. We have supplemented the above data results to line 377 of the manuscript.

Point 2: Why use term luminance in R and Brightness in others G and B.

Response 2: We are extremely grateful to you for pointing out this problem. We are very sorry. This is because we did not have a unified vocabulary when editing manuscripts, resulting in this difference. We have replaced ' luminance ' with ' Brightness ' (The revised position is at line 217 of the manuscript).

We would like to thank the referee again for taking the time to review our manuscript.

Best regards!

Reviewer 4 Report

This paper proposes a garbage classification method based on small convolutional neural networks. While the topic of the paper is interesting, some parts of this manuscript require revision. I recommend major revisions as a condition before publication. Some comments on this manuscript and suggestions to the authors are as follows:

1. The language of the manuscript is too poor. It is necessary to invite professionals to review and revise the manuscript.

2. The literature review is insufficient. Research related to garbage sorting methods has been ignored in the current version, resulting in the manuscript's importance and necessity not being clearly described. In addition, authors can appropriately increase the number of references.

3. There are some problems with the bibliography format. For example, "Wang Cong et al. (line 57)" should be modified to "Wang et al.". The same problem occurs on lines 63, 67, 79, etc. Also, the reference on line 201 is problematic.

4. The innovation of this work is not clearly emphasized in the manuscript. In addition, the contribution of this work should be further stated.

5. Try not to use "we" and other similar descriptions in the manuscript, but use "this study", "this study", "this paper", etc.

6. Furthermore, what are the results if the economics and practicality in the waste sorting method proposed in the manuscript are taken into account?

Author Response

Dear reviewer:

We appreciate you very much for constructive comments and suggestions on our manuscript. These suggestions allow us to optimize this study more comprehensively.  We have responded to each of your valuable suggestions.

Point 1: The language of the manuscript is too poor. It is necessary to invite professionals to review and revise the manuscript.

Response 1: We are extremely grateful to you for pointing out this problem. We have invited professionals to review the language of the manuscript and have made some language revisions to the manuscript.

For example, we modified some sentences in this manuscript to passive voice into better express the meaning of the sentences.

Original example: This study set the threshold to 255 for evaluation.

Modified example: The threshold is set to 255 for evaluation (at line 181).

Point 2: The literature review is insufficient. Research related to garbage sorting methods has been ignored in the current version, resulting in the manuscript's importance and necessity not being clearly described. In addition, authors can appropriately increase the number of references.

Response 2: We thank the reviewer for pointing this out. We agree and have updated. We added 7 new references on the basis of the original manuscript, and revised certain contents in the article for citation. The specific modifications are as follows:

For the newly added reference 1, cited at line 31 of the manuscript.

For the newly added reference 4, cited at line 35 of the manuscript.

For the newly added reference 5, cited at line 38 of the manuscript.

For the newly added reference 8, cited at line 72 of the manuscript.

For the newly added reference 9, cited at line 75 of the manuscript.

For the newly added reference 10, cited at line 76 of the manuscript.

For the newly added reference 19, cited at line 143 and 237 of the manuscript.

Point 3: There are some problems with the bibliography format. For example, "Wang Cong et al. (line 57)" should be modified to "Wang et al.". The same problem occurs on lines 63, 67, 79, etc. Also, the reference on line 201 is problematic.

Response 3: Thank you for this valuable feedback.

We’ve changed [Cuiping Shi] to [Shi et al.] (Page 2, line 76).

We’ve changed [Cong Wang et al.] to [Wang et al.] (Page 2, line 83).

We’ve changed [Qianqian Luo et al.] to [Luo et al.] (Page 2, line 88).

We’ve changed [Jinqiang Bai et al] to [Bai et al.] (Page 2, line 93).

We’ve changed [Sidharth R et al] to [Sidharth et al.] (Page 3, line 104).

We’ve changed [Tanya Gupta et al] to [Gupta et al.] (Page 3, line 109).

We’ve changed [Mindy Yang et al.] to [Yang et al.] (Page 3, line 138).

Point 4: The innovation of this work is not clearly emphasized in the manuscript. In addition, the contribution of this work should be further stated.

Response 4: We thank you for your suggestions on the innovative issues of manuscripts. Due to the imperfect writing of some parts of our article, the research contribution in the manuscript is not obvious enough. We have made some changes in the abstract and conclusion parts of the manuscript.

The main contributions of this study are three aspects.

1. We propose a garbage image preprocessing method. An adaptive image brightening algorithm is developed to average the brightness of the background, and the threshold replacement method is used to reduce the shadow noise. Then the Canny operator is used to assist in cropping the blank background in the image.
2. We design a simplified CNN model based on the MLC-CNN structure, making it simple and efficient. After the experiment, the model can achieve 96.77% accuracy on the self-built dataset and 93.72% accuracy on the TrashNet dataset, which is higher than the accuracy of MLC-CNN.
3. We also found that the Adamax optimization algorithm based on Adam algorithm has better optimization network effect.

Point 5: Try not to use "we" and other similar descriptions in the manuscript, but use "this study", "this study", "this paper", etc.

Response 5: We thank the reviewer for pointing this out. We agree and have updated words like "we" and other similar descriptions in the manuscript. We have used "this study" and "this paper" instead.

 

Point 6: Furthermore, what are the results if the economics and practicality in the waste sorting method proposed in the manuscript are taken into account?

Response 6: After experimental testing, the time for preprocessing one image is about 1.6 seconds. The CNN model in this paper takes about 0.22 seconds to perform a classification task on an image, and about 2.37 seconds for an array of 1000 image data. The processing efficiency of the model is relatively good. We have supplemented the above data results to line 377 of the manuscript. The combination of the model and the Raspberry Pi or the combination of the host computer can derive a lot of practicability, such as intelligent garbage classification container, garbage classification big data or intelligent garbage classification robots and so on. The economic cost of the model trained through the network is low, and the computer with the discrete graphics card can accelerate the training through the GPU to output the model and deploy it in certain application scenarios.

We would like to thank the referee again for taking the time to review our manuscript.

Best regards!

Round 2

Reviewer 2 Report

Reviewer Reports:

 

I have reviewed the revised version manuscript entitled” A garbage classification method based on a small convolution neural network”. The work is interesting and it falls within the scope of the journal. The paper has been improved and can be accepted. 

 

Reviewer 4 Report

no further comments.