Research on a Rice Counting Algorithm Based on an Improved MCNN and a Density Map

Round 1

Reviewer 1 Report

The manuscript presents a new approach to grain counting using deep learning and neural networks. The work falls within the scope of the journal, and the research theme is relevant. However, there are several flaws in the manuscript that prevent a positive recommendation. Thus, considering the high stands  of this prestigious journal, I do not recommend the publication of this manuscript. See below for a brief list of suggestions and criticisms for an eventual resubmission.

1. First of all, a deep revision in the English is needed, particularly in Introduction. A striking typo is present in almost all tables is the mispelled word "oringinal" instead of "original". Besides, the use of the pronoum "ours" in the table column title is not fine. "Current work" for example sounds a better title for instance.
2. Second, references are formatted incorrectly. Most of them do not contain essential information, such as names of the authors, name of the magazine, volume, pages and date. The correct citation is a fundamental aspect of a scientific article, so the failure in the references is a serious failure that prevents the proper evaluation of the work.
3. There are several problems with the formulas used throughout the manuscript. Comprehensive review is needed. For example, in Eq. (6) the indices of the two sums are repeated ("j" in both); it is clearly a mistake and one of them should be "i". Another problem, the "F_c" symbol is not explained in the text, unless I have missed it. All abbreviations and symbols must be properly presented.
4. The accuracy of the algorithms is another issue that needs to be clarified. As the authors do not present a formal definition of accuracy, the values presented above 100% are very confusing. In particular, I was unable to follow the discussion of the results due to this failure. Please clearly indicate the precise accuracy formula and the meaning of having values greater than 100%.
5. The legend of figures and tables can be improved by providing more details about what is being shown. I found at least one typo in the caption in Figure 7, but a major revision is suggested.
6. Finally, I have a criticism regarding the results achieved. As mentioned above, the misuse of the term accuracy prevented me from fully understanding the results. But despite this, although it is possible to verify that the proposed algorithm does in fact lead to better results in the identification of grains, it is not clear how this improvement represents a gain in the calculation of the real value of the mass of a thousand grains, or even the capacity of the algorithm to recognize the class of rice being analyzed. In my view, the methodological approach should be revised to clarify the code's ability to assess the correct weight of a thousand grains weight of different types of grains. No comparison between the results of the analyzed images and the actual value of the weight of a thousand grains is shown. Thus, in my view, the manuscript lacks scientific soundness.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a network to output a density of rice which is used in order to count the rice. This is purported to be an important problem. The network and signal processing ideas are not new and straightforward but it seems to works better than previous attempts. The strong parts of the research are the collection of the image dataset and tweaking the network to get good results.

There are tables with entries like 999.00% accuracy which I do not understand. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

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.