Detection of Varroa destructor Infestation of Honeybees Based on Segmentation and Object Detection Convolutional Neural Networks

Round 1

Reviewer 1 Report

Dear Editor,

I am writing to you as a reviewer of the manuscript titled "Detection of Varroa Destructor Infestation of Honeybees Based on Segmentation and Object Detection Convolutional Neural Networks," which was submitted to Agriculture with the manuscript ID agriengineering-2516442.

The authors proposed a convolutional neural network that integrates machine vision to detect Varroa mite infestation in honeybee colonies. The experimental results showed that the authors' model outperforms other models. Please find the detailed review report for the manuscript attached.

To improve the quality of the publication, I suggest the following corrections:

1. The introduction and literature review are well-prepared. However, a literature review about deep learning in the field of bees and related issues can be strengthened a little to improve this part.

2. A serious question is raised about the data collection and whether the presence of Varroa Destructor Infestation in the sampled bee hives was verified. Or was it artificially introduced, which was not the case according to the authors' description. For this reason, there is a concern about the lack and inadequacy of infected bees. Please clarify this section accordingly.

3. On what basis was the number of 1/200 determined?

4. Can the Varroa image augmentation method provide the necessary standard to create new images to improve the dataset, or not?

5. More information is needed regarding the stage of Varroa infection and its severity.

6. In the materials and methods section, a table that specifies the number of images of healthy and infected bees, as well as the number used in the stages of model training and testing, should be included.

7. Can the authors provide a solution and a suggestion for estimating the number of Varroa in the hives?

8. The Python code prepared should also be attached to the article.

9. Based on sources or experiences, was there a specific time that could produce the best diagnosis result? In this way, the authors could achieve the best result.

10. Table 3 is disorganized and needs to be revised.

11. The results should be presented in terms of the different numbers of images in the training and testing phases to show the capability of the method better.

12. In the results and discussion section, it would be better to refer to the results of other researchers and apply this in the revised article.

13. A section for suggestions and future research based on the limitations of this article should be added.

Based on the above, the article can be accepted after the necessary corrections and re-examination.

Thank you for considering my opinion.

Sincerely,

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is an interesting paper using powerful methods for detection of Varroa infestation load from images of honey bees. As Varroa management is so important in successful beekeeping, the topic is very relevant. However the description of the method in section 2 is quite high level and the paper would be more readable and useful if more details were given. The figure legends would benefit from some more explanation also. The paper generally is quite dense and needs to be made easier to follow. The presentation needs to be improved.

The statement made at line 33 about Varroa mites feeding on hemolymph is now known to be wrong. See the comment and the reference to Ramsey (2019), which should be added.

The in-text referencing needs some attention. “et al.” is often omitted.

In one or two places a definition and reference are needed.

At lines 115/6, the dates have gone wrong.

In section 2.2.1 or elsewhere please explain the nature of the inputs and outputs of the FCN. In general more explanation of the steps would be useful.

Lines 244-245 need some clarification.

For formulae 10-15, please provide the names of these measures.

Table 2 needs a more informative legend.

The Figure 10 legend needs corrected.

The layout around table 3 has gone wrong.

Throughout, biological terms should be in italics.

The references need some tidying.

Please see the annotated document for these and numerous other suggestions. The English needs quite a bit of attention, as do various other details.

Comments for author File: Comments.pdf

Please see the annotated document for some improved wording suggestions.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have attended to most of the required changes. In doing this some further minor edits are now required. Please see the annotated document for these edits and some others noticed while re-reading, which the authors should now attend to.

Comments for author File: Comments.pdf

Please see the marked-up document.

Author Response

Thanks for your valuable comment to improve our work. We have corrected all the questions you raised and marked the sentence in highlight in the revised manuscript. Besides, as the corresponding author, B. X. contributed resources, supervision for this work, therefore three projects led by B. X. are added to the funding part. The detailed information of these three projects are as follows: China Agriculture Research System of MOF and MARA (No. CARS-44), Shandong Modern Agricultural Technology System, China (No. SDAIT-18-06) and the Efficient Ecological Agriculture Innovation Project of the Taishan Industry Leading Talent Program (No. LJNY202003).