Next Article in Journal
Global Formulation of the Synthetic Storm Technique Oriented to Satellite Link–Budget Design
Next Article in Special Issue
BER Minimization by User Pairing in Downlink NOMA Using Laser Chaos Decision-Maker
Previous Article in Journal
Predictive Maintenance System for Wafer Transport Robot Using K-Means Algorithm and Neural Network Model
 
 
Article
Peer-Review Record

YOLO-Based Object Detection for Separate Collection of Recyclables and Capacity Monitoring of Trash Bins

Electronics 2022, 11(9), 1323; https://doi.org/10.3390/electronics11091323
by Aria Bisma Wahyutama and Mintae Hwang *
Reviewer 1:
Reviewer 2: Anonymous
Electronics 2022, 11(9), 1323; https://doi.org/10.3390/electronics11091323
Submission received: 17 March 2022 / Revised: 11 April 2022 / Accepted: 19 April 2022 / Published: 21 April 2022
(This article belongs to the Special Issue Advances in Intelligence Networking and Computing)

Round 1

Reviewer 1 Report

The authors have proposed YOLO based object detection method to separate recycle trash. The YOLO based object detection detect and classify the objects in the correct category. There are a few issues that are needed to be addressed:

  1. Why the Raspberry Pi has been selected for the implementation of the YOLO object detection?
  2. Comment on the computational expenses to implement the YOLO. Also, does it have resources to expand the YOLO with more convolutional layers?
  3. How the algorithm will work with composite materials such as foam etc.? Has it been tested? Comment on that.
  4. Figure 13. and 14 are not very clear, please modify them.

Author Response

I would like to express my deepest gratitude for your kind review and input to this research paper which I also appreciate. All the answers regarding the review, comment, and input are described in detail in the attached file. 

Thank you for your understanding.

Regards,

Aria Bisma Wahyutama

Author Response File: Author Response.pdf

Reviewer 2 Report

Authors have done a proof of concept smart recycle bin. They have used the YOLO object detection algorithm on raspberry pi to identify 4 types of recyclables and a bunch of arduino uno’s for controlling motors to open a compartment, connecting to WiFi, etc. 

The writing has to improve. Please get it proofread by native English speaker. 

Authors have to highlight the scientific contributions of the paper. This seems like a science project. What do you expect the scientific community to learn from your research? Where are the  evaluations with other similar implementations such as https://ieeexplore.ieee.org/document/8622212 and https://arxiv.org/pdf/2006.05873.pdf.

Clearly state your objectives, scientific findings and conclusions  in the paper.

What is the purpose of GPS location? Who will read the GPS coordinates of bins on their apps and what will they do with that? Are you building an integrated recycle management system? It is not clear from the paper what the objective is here. 

Specific comments:

Line 10-11: “The results will open up the trash bin’s compartment …” need to be reworded. 

Similarly, lines 11-12 don't make sense.

Introduction and related work is too short and should be expanded. Use more authoritative references. The only related work you cover is of reference 8. How about more recent works such https://ieeexplore.ieee.org/document/8622212

Figure 1: is not legible and readable. Seems like you just copied it from Ref 5?

In general all the figures in paper are blurry and not readable.

Show a clear diagram with all the components working E2E.

Figure 3 : What is the purpose of the bottom line with false branches? How are such errors handled? 

Line 111: scientific research should be reproducible. Provide references to Kaggle or Google dataset used in your training.

Line 120: delete “is”

Figure 5: is the trash bin for public use of home use. If it is in a public place, how will it connect to WiFi? If it is for home use, why do you need GPS tracking?

Table 2: YOLO line is repeated.

Line 172-173, as commented previously, for reproducibility, you have to state how you searched and what all data was collected. No. of images/videos, number of objects, etc.

Line 183: Reference 14 is incorrect.

Eq 1: can you provide a generalized equation for different sizes of trash bins?

You haven’t shown the system working E2E. With this type of data it is hard for other researchers to replicate your results. 

Show precision, recall and f1 scores in evaluations.

Use another standard dataset such as trashnet to evaluate your results. Also, you should include test results of real-world recycle item classification and how much is the accuracy compared to the synthetics datasets. 

85% accuracy seems too low. Other older studies such as https://ieeexplore.ieee.org/document/8622212 have shown 95% accuracy in classification with DenseNet121. 

You have to compare the results of your system with other previous works and provide a comparison and conclusions.

All graphs in the results section should be revised as they are not legible.



Author Response

I would like to express my deepest gratitude for your kind review and input as well as my appreciation for your consideration of this research paper.  All of the answers regarding the comments, inputs, and reviews are described in detail in the attached file. Thank you for your understanding.

Best regards,

Aria Bisma Wahyutama

P.S. the manuscript has been proofread by Editage.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have provided responses to the questions and also modified the manuscript accordingly. The manuscript can be accepted for the publication.

Back to TopTop