Benchmarking Machine Learning Approaches to Evaluate the Cultivar Differentiation of Plum (Prunus domestica L.) Kernels

Round 1

Reviewer 1 Report

The authors of the paper have addressed all the questions logically. So the manuscript can be accepted.

Author Response

The authors are very grateful to the Reviewer for this comment. 

Reviewer 2 Report

Dear Editor

Present form of this paper is appropriate and could be published.

Regards

Author Response

The authors would like to thank the Reviewer for this comment.

Reviewer 3 Report

The article also has some format problems and details. For example, in the second part, the introduction of evaluation indicators and methods can add subheadings to increase the readability of the article, and the paragraph format of the fifth part.

Author Response

The authors are grateful to the Reviewer for a thorough review of the manuscript and valuable comments.

The manuscript has been checked and English changes have been made throughout the manuscript.

The format of the subsections has been corrected. 

For subsection 2.2. Image analysis, the following subheadings have been added:

2.2.1. Image acquisition  

2.2.2. Image processing 

In the case of subsection 2.3. Discriminant analysis, the following subheadings have been added:

2.3.1. Cultivar discrimination of plum kernels

2.3.2. Performance metrics

2.3.3. Machine learning algorithms

Section 5. Conclusions have been corrected including the format change. This section has been divided into 3 paragraphs. Minor corrections to the text have been made.

Additionally, the manuscript has been improved including other details marked in red.

Reviewer 4 Report

I have no further remarks.

Author Response

The authors are grateful to the Reviewer for the manuscript review and this comment.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This paper compares machine learning approaches to assessing plum cultivar differentiation. Some of the claims presented in this paper need to be reconsidered.

1- One of the issues that needs further clarification is the non-destructive nature of the method used. In this research, a plum-related cultivar is identified from the plum kernel. Do you detect plum cultivars without damaging the fruit?

2- What is the need to distinguish plum cultivar from its kernel? Do you not think that it is a laborious and destructive method? In my opinion, it may be possible to distinguish the cultivar easily from the spectral characteristics of the fruit itself or by using the color, shape and texture features.

3- Another issue that needs further explanation is paper innovation. Respected authors should explain what new technology or new method has been used in this research? It seems that the idea of distinguishing a plum cultivar from its kernel may be a new idea, and other items used in this research, such as the research method, are not much different from other published papers in this field.

Reviewer 2 Report

Dear Authors

I go through the manuscript,

1. i think introduction must be improved for topic of machine learning in different plants for example for shape, yield and so on, analysis in materials and methods must be a little in details.

2. in this review i didn't see any appropriate discussion and comparisons with those scientists who have prominent results especially about machine learning.

3. References must be check once again for better format 

Reviewer 3 Report

This paper introduces a method of almond variety identification based on machine learning. Its main content is to collect the images of plum kernels (‘Emper’, ‘Kalipso’ and ‘Polinka’). The MaZda software was oriented toward image analysis involving calculations of the texture features on kernel surface, expecting to convert the image signals to feature parameters. A variety of mathematical methods, such as co-occurrence matrix, run-length matrix, Haar wavelet transform, gradient map, autoregressive model and histogram, were successfully used to extract the texture parameters for well-matched models of cultivar discrimination. It may be useful in the processing industry to avoid cultivar mixing and falsification.

1. Many references about plum are cited in the paper, but there are few references about fruit kernel classification, which is lack of logic. In line 56-58 describes the classification standards of plant varieties. As for the content of this paper, it is necessary to supplement the research status, so as to lead to the research methods of this paper.
2. The paper is not specific and detailed enough in part 2（materials and methods）; The benefits of the machine learning method selected in this paper should be supplemented. Advantages in dealing with the problems to be solved in this paper. At the same time, the explanation of treatment methods and evaluation indicators shall be supplemented.
3. In part 2, the comparison results of the three machine learning algorithms finally selected with other algorithms are not displayed, and the comparison results of color space Lab and color channel B of texture features with other channels are not displayed.
4. Part 3 is a little complicated. You can put the machine learning method and feature parameters finally selected in this paper into Part 2.At the same time, the research status at home and abroad introduced by Part 3 and related research are enlarged to part 1.

Reviewer 4 Report

I reviewed the paper entitled “Benchmarking machine learning approaches to evaluate the cultivar differentiation of plum (Prunus domestica L.) kernels.” The contribution has merit and is acceptably presented because:

(a) The paper is correctly directed to the proposed target audience of Agriculture, and the manuscript's organisation flows well.

(b) The references used in this article are appropriate and up to date.

(c) The authors provide an interpretation of the results following the domain of the case study. Lastly, they discuss how this outcome is valuable and meaningful.

However, the authors should strengthen the paper. In this regard, I have only four remarks to the authors:

1. Add the parameter settings of the classifiers.
2. It would be best to include (or make publicly available) the dataset. The purpose of this suggestion is transparency and reproducibility (in fact, the trend in research is to include code and data). Besides, this action may increase the impact of this study in terms of both applications and citations.
3. The internal, external, and construct threats were not discussed. You should discuss them in the conclusion section. If you have some complications, then, alternatively, you should at least discuss the scope and the limitations of this research.
4. Lastly, the paper should discuss some directions for future research.