CNN-Based Inspection Module for Liquid Carton Recycling by the Reverse Vending Machine
Round 1
Reviewer 1 Report
A careful and major revision is required for each section of the paper. Important details of methodology and results are missing. The comments/suggestions are:
1. It is suggested to modify the title to reduce its length. Moreover, the title does not represent the actual study i.e. detection of waste liquid carton.
2. Introduce your own work at the end of intro section. More details can be provided.
3. The work is based on CNN model development for detection of liquid cartons. The abstract and intro section unnecessarily discuss other information in detail.
4. Similarly, length of section 3 can be reduced. Instead of these unnecessary details, other related information can be added as suggested in few of the following comments.
5. What are the number of images for each class? What kind and how many of images are there in both classes? How many images are there after data augmentation?
6. What is TL? It is not commonly used. Complete ‘Transfer learning’ word is better to use. Moreover, write full name at first mention.
7. Table 2 shows performance comparison on the basis of just 3 epochs? Is it enough for a fair comparison and to show comparison results?
8. The details of transfer learning are not given. Authors can add everything in detail.
9. Add images of different samples taken from different cameras?
10. Please clarify that: The authors collected datasets of same image samples from four different cameras and then compare the models developed from each dataset for results given in table 3? The details for results in table 3 are not given? How these are obtained? Which CNN structure is used, etc.? What are the number of epochs?
11. Why the results from different cameras are different? The images taken for each sample from all the cameras are in same conditions? Please clarify. How the minimum loss given in the table is taken (the loss at last epoch)? Add the loss and accuracy curves for models developed from different camera images?
12. What does the following line means: “The validation accuracy was 0.96 at this round of training. The outcomes can be even more accurate by the process logic in Figure 2, because it evaluates the inspection three times. Then, with the 96% accuracy, the overall accuracy can be increased up to 99.99%, because the possibility of three times the false detection in a row is 0.006496%.”
13. Avoid the use of uncommon terminologies throughout the paper like vision-based CNN technology, CNN inspection agent, etc.
14. In abstract: “With this model, the results showed that the accuracy of detection was over 99%, and the time to inspect one item was less than 2 seconds” How 99 % by the CNN model? And the information about time cost is not given in the paper?
15. Provide the implementation details in the paper.
16. The actual work detection of liquid carton is not mentioned properly in abstract.
17. Remove unnecessary details about COVID-19, etc. from both the abstract and intro section.
18. The last para of conclusion section can be improved.
Author Response
Authors would like to thank the reviewers and editor for their time and efforts. We believe the manuscript has been improved substantially. For the comments raised, we have addressed them accordingly in the attachment file.
Thank you, again.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments for author File: Comments.pdf
Author Response
Authors would like to thank the reviewers and editor for their time and efforts. We believe the manuscript has been improved substantially. For the comments raised, we have addressed them accordingly in the attachment file.
Thank you, again.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The revisions and responses are satisfactory; and the paper can be accepted in present form.
