Research on Instance Segmentation Algorithm of Greenhouse Sweet Pepper Detection Based on Improved Mask RCNN

Round 1

Reviewer 1 Report

The authors used open source data set for their analysis.

There are some writing issues and fundamental contradictions in the manuscripts that should be resolved before the manuscript gets reconsidered for peer review.

For example, in the end of the abstract, it is stated: Further, the requirements for real-time sweet pepper growth monitoring were met. In conclusion: The main disadvantage of the current instance segmentation algorithm in crop fields is its poor real-time performance, which is not conducive to practical application

few examples of writing issues: what FPS represents in the abstract? Should be defined.

line 103 states,  Hao et al. [28] used.. but the author of ref 28 is Gan et al.. all of the citations provided in the text should be re-evaluated to avoid such mistakes

line 166: Divide it into training and validation sets (Table 1). who should divide it? reader?

The manuscript should be fully proofread before submission for peer review.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1- In introduction, add a little paragraph with related references, about machine and deep learning, the difference between regression and classified methods, then go beyond that, why you used classified deep learning. You can use the following papers for help (https://doi.org/10.1038/s41598-022-16114-5; https://doi.org/10.3390/w14223647; https://doi.org/10.1016/j.compag.2022.107457)

2- Add flowchart at the begining of M&M to summarizes all steps of the work

3- Merge Figure 6 with Figure 7 and add the suitable panel. 

4- Improve the resolution of Figure 11 and maximize the captions to explain what in figure

5- You should discuss the study limitations rather than focusing only on the study strengthes

Author Response

Please see the attachment.

Author Response File: Author Response.pdf