Promotion of Color Sorting in Industrial Systems Using a Deep Learning Algorithm

Round 1

Reviewer 1 Report

The presented research problem in manuscript “Which are the latest developed algorithms in the field of machine vision that can be applied to reduce the operator’s subjective feeling when defining the criteria for a good product?” was actually not explored in this manuscript, instead, it provides answer to a question: what is performance of the YOLOv3 algorithm in case of classifying rasberries during online inspection? In this regard, the context is now more in the form of a technical report with minor novelty. For this reason, it should be considered to rework the content either to explicitly respond to original research question (requires comparison and analysis of multiple classification methods applied to the monitoring problem, plus reported literature review) or to reformulate the research question.

The research problem definition now include the reduction of subjective human  factor in determination of criteria for monitoring, whereas in Conclusions the results are compared to industrial (automatic) color sorters. This is also a mismatch between the research question and a reference measurement for the study. In the abstract, no results of comparisons to  the references (operator or automatic color sorter) are reported as in the manuscript also.

Also, the objective now remains unattained, namely “And the major objective was to develop a method to reduce the subjective feeling of the human factor when defining the parameters (criteria) of a good product". To reach the objective, it would be still necessary to design a setup, where comparison between performances of manual and algorithm-based classification would be analysed. There are also gap in developing the rasberry monitoring method, since basically only well the established YOLOv3 algorithm is applied to the task.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

General comments

-This manuscript, by authors, studied “

 “Promotion of color sorting in industrial systems using a deep learning algorithm”.

Overall, the topic is of interest to Applied Sciences, readers. However, the following are the specific comments on the article concerns, before publication as a major revision.

Figures and Table numbers must be arranged.  Figure 1, 2, 3, and so on or Fig 1. (a, b, c, d). Similarly, Tables as well.

Specific Comments and Suggestions

-Abstract

-Add more results in detail.

-Significance of your study? Specific findings?

- Add quantitative results.

-Introduction

-Need to summarize and be specific with your concerned study.

Add more details about the relationship between agriculture and original raspberry images.

-Revise the introduction section with summarizing and the significance of the study.

YOLOv3 (Is it an abbreviation?). Define its full form as well as detail for modeling.

-Mention specific objectives of your study

-Reference should be explained with separate details. e.g. too many references cited together? And again and again, is not appropriate (1-4, 5-7,12-15, 16-21,26-30).

This paragraph is to move to the last part of the Introduction section. In addition, mention your objectives specifically.

“A major research question was as follows: Which are the latest developed algorithms in the field of machine vision that can be applied to reduce the operator’s subjective feeling when defining the criteria for a good product? And the demands that a product should meet: to work in real-time, to be suitable for detection, localization, and classification of objects, and to be able to detect small-size objects.”

-Materials and Methods

-Methodology also needs to summarize and be specific. 

-Add more details, which will strengthen your current study.

-Unnecessary details make the manuscript confusing.

- Please add experimental setup (field and lab experiments), pictures, and more figures to make the manuscript more attractive.

-Too many details make a schematic diagram to show your steps in a better way.

-F1 measure  Mean average precision – mAP… Why in bullets? not heading or sub-heading?

-Results and Discussion

-Revise thoroughly with discussions supported by other studies

-A weak discussion about results as mentioned in the introduction section. Please add other studies references to support your current study and its significance.

 

-        Why highlighted within Tables. Use a scientific way with statistical analysis.

-        Figure 3.1 c is not clear.

-        Figure 3.2 need to improve or changed with better resolution.

-Conclusions

-More specific, please?

 

-Significant differences on which basis? 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Authors have revised the manuscript towards a scientific report. However, there are still issues related to experimental scheme and organisation of the manuscript.

The presented research lacks the benchmark, namely the comparison of results with YOLOv3 against some other classification method(s) (and/or optical monitoring equipment). This can be regarded as a standard procedure in scientific reports today.

In the Conclusions-section there are also results presented, instead it should be rewritten to include only conclusions derived from the actual results of your work (for example, there are now some direct conclusions at rows 364-388 that are not in the Conclusions-section).

The discussion section, is now kind of a blended Introduction and Discussion, it should be re-organised in a way that the results of this work are discussed and then these findings are reflected against the literature (if relevant and comparable).

In a recent article published in Agriculture (MDPI), YOLOv5s outperformed YOLOv3 in grape yield detection at various visual conditions: https://www.mdpi.com/2077-0472/12/8/1242    

A claim at rows 81-84 should be referenced.

Acknowledgments and Conflicts of interest are doubled in the end on the manuscript, the original help text should be removed from there.

In section 4.2 a web application is presented. This section is kind of separate study, it should be connected tightly to serve achieving the aim of the research. Also, there is an invalid www-link presented at row 451, namely http://localhost:43311/

If the web application is left to the manuscript, the same detailed information level should be applied to the reporting as in the other sections of the manuscript, and its performance should be evaluated.

The manuscript includes now both active and passive sentences, depending on the politics of the Journal, use the other of the forms, not the both.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors made great efforts to improve. It is accepted in the present form but improvement in discussion will be beneficial before publication.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

After revisions, the manuscript can be now considered to be published in Applied Science, taking into account the following notifications:

- The authors need to further (shortly) motivate in text the exclude of comparison of multiple algorithms applied to same photographs as the YOLOv3 they tested. Also, the discussion and conclusions about the applicability of the tested YOLOv3 are needed to shape according the above shortcoming,

- Some of the Figure-texts are without the "." at the end of the sentence.

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf