Image and Point Cloud-Based Neural Network Models and Applications in Agricultural Nursery Plant Protection Tasks
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsOverall Comments:
The manuscript, titled “Deep Learning for Image and Point Cloud Data in Agricultural Nursery Informatics: Models and Spraying Applications,” offers a fairly comprehensive and well-organized review of a highly topical topic. The authors have gathered a considerable amount of information on deep learning models applied to image and point cloud processing, with an emphasis on three main tasks (classification, segmentation, and object detection), as well as six key application areas (leaf disure detection, pest identification, weed recognition, target and non-target object detection, seedling information monitoring, and spray drift assessment).
Overall, the manuscript reflects a good understanding of the state of the art and the current technological landscape. However, in its current form, it reads more like a descriptive compilation than a critical analysis. I believe the value of an impact review lies not only in bringing together previous work but also in offering an in-depth, comparative, and thoughtful analysis that provides insight into why certain approaches are more appropriate than others in specific contexts.
In this regard, the manuscript would be improved by a more in-depth discussion of the relative performance of the aforementioned models, a clearer identification of current gaps in the literature, and greater contextualization for nursery settings. A separate discussion section should also be included, as well as original visual material to aid in better interpretation of the text. Regarding language, the writing is quite clear, but could be improved. The text could be more concise and fluid. A professional editorial review would be advisable to optimize the text's presentation.
Major Comments:
1. Lack of comparative analysis between models: Although many models are described, no critical comparison is offered between them. A table summarizing their performance, efficiency, computational requirements, and suitability for nursery environments would be valuable, especially considering constraints such as plant occlusion, variable lighting conditions, and limited hardware. A thorough review should help the reader understand not only "what's out there," but "what models best and why."
2. Limited contextualization to the nursery environment: The manuscript claims to focus on agricultural nursery informatics, but many of the models and applications presented seem extrapolated from general agricultural scenarios. The review would be significantly improved if it were clearly established which nursery-specific challenges have not been sufficiently addressed in the literature (e.g., high plant density, support structures, canopy geometric variability, among others).
3. Lack of original visual material: All figures are taken from previous work. It is suggested to include graphic material developed by the authors themselves, such as flowcharts, conceptual frameworks, or comparative matrices. These types of elements help synthesize the content and provide a unique perspective to the review. It should not be forgotten that "a picture is worth a thousand words" and that graphic material always helps us better interpret the text.
4. A separate "Discussion" section: Although subsections on "future directions" and "research challenges" are included, there is no discussion section that allows for critical reflection on the real barriers to the adoption of these technologies, their practical limitations, or the path toward their effective implementation in real nurseries.
5. Improve the "Conclusion" section: "Conclusion" section could be strengthened. It is suggested to rewrite it by emphasizing the most relevant findings derived from the analysis of models based on both "image" and "point clouds", and highlighting practical implications for the specific field of agricultural nurseries.
Minor Comments:
1. Inconsistent Use of Terms and Abbreviations
It is recommended to standardize terminology throughout the text. For example, expressions such as "plant protection tasks," "operations," and "systems" are used interchangeably without defining whether they refer to the same or different concepts. Likewise, given the large number of acronyms used, it would be very useful to include a table of abbreviations or a glossary.
2. Unexplanatory Figure Legends
The legends describe what the figures show, but do not guide the reader on how to interpret them or why they are important. For example, the legend for Figure 2 correctly describes the methods compared, but does not indicate why one is superior to the other or what conclusions can be drawn from the visual comparison. Adding one or two interpretive sentences to the legends for each figure would improve comprehension.
3. Copyediting and Correcting Minor Errors
A thorough proofreading of the text is recommended to correct typos, minor inconsistencies, and excessively long sentences. This will allow for a more fluid and professional manuscript.
It is recommended that the manuscript be reviewed by an expert or a language editing service.
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsI find this a well written and very interesting and timely review but I think it has too much of a technical edge towards computer science. Moreover I find it more a general overview on agriculture than a specific work on nurseries.
However I am sure that this manuscript deserves publication after a general major revision.
Please find some specific comments below.
Please change the typology of the manuscript, from article to review.
Please check the keywords, they should not repeat the words of the title.
Please improve the abstract. In my opinion it should not only list the main topics but also summarize the contents.
1. Introduction
Please specify at the beginning of the introduction which kind of nurseries you are talking about (ornamental plants, vegetable crops, other...).
In the first part of the introduction (line 18-34) citations are missing. Please provide.
Line 40-43 please add that the use of robots instead of humans for spraying treatments is also important for safety reasons.
Line 44-50 citations are missing. please provide.
Line 52 please add the traditional approach with common manual or tractor treatments.
May you specify some existing examples of real applications of machine learning, deep learning and neural network in nurseries?
In the introduction please add the main robots available on the market, their tasks and their navigation system.
In the introduction the main automatized nursery crops are not reported, please add.
Please specify in the introduction which kind of spraying treatments are usually performed in nurseries like chemical fungicides, chemical herbicides, natural products, bio-stimulants or leaf fertilization.
2. Image-based Neural Networks
Line 89-109 no citations
line 110-127 no citations
For all classification models I think some concrete examples should be provided as it is not easy to understand from the readers of Agronomy such technical information on computer science.
Line 211-230 citations
Please provide examples also for the different segmentation models
The same for object detection models.
Do all these neural network methods can be applied to specific categories of robots? In case ground or aerial? Please specify.
In my opinion even the point cloud models are not easy to understand from the readers of Agronomy. I think some practical task examples could clarify and the differences between neural network and point cloud neural network should be clarified. I think this part is hard to follow if you are not an expert in computer science. In general authors should make an effort in this direction
Concerning the section 4, NN for plant protection, I honestly find the same weaknesses. The manuscript talks about the models but not the applications and in my opinion the example are not exhaustive. Crops are never mentioned and practical scenario are not discussed.
Moreover in my opinion this is a good dissertation but seedling monitoring is just a small part even without concrete application example.
In my opinion the review is more general for agriculture and not specific for nurseries. nurseries can be left but only as an example.
In the light of this, I would suggest to give a more general title to the review.
I find the same issues in future directions and conclusions. I would expect more concrete applications scenarios described. Moreover the article never mention sustainable or organic agriculture. Could these models help to improve the efficiency of the treatments even in this more challenging cropping systems?
Thank you in advance for taking my suggestions into account.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article uses images without clarification of copyright. The article includes figures (Figures 1, 2, 3, 4, etc.) taken from other articles ([116], [117], [120], [123], etc.) without specifying whether permission was obtained for their reproduction. This raises an ethical and legal problem, especially in an academic publication. An explicit statement on the origin and rights of use of these images is required.
There is a lack of depth in the discussion and contributions of the article, which is limited to listing models (CNN, YOLO, PointNet, etc.) and their applications in agriculture (detection of diseases, weeds, etc.), but there is no methodological critique, and there is a lack of comparative tables that could generate a systematic review summarizing the advantages and disadvantages, accuracy, performance of the models under different scenarios, etc. There is also a superficiality in the conclusions, which do not propose concrete solutions or innovative theoretical frameworks.
The article has an unbalanced approach, overemphasizing popular technologies, which creates a bias, as it cannot justify why other less-mentioned models would not be relevant. Problems such as scalability in real environments, implementation costs, interoperability with existing agricultural hardware, etc. are also not mentioned.
The authors cite multiple works of their own, suggesting a possible bias toward self-promotion without contrasting with independent literature.
The article partially fulfills its purpose as a descriptive review but fails to offer critical analysis, originality, or practical guidelines for the agricultural community. Its academic value is compromised by the superficiality of its conclusions and the absence of novel theoretical contributions. A major revision addressing these issues is recommended.
Image permissions must be included. The discussion needs to be deepened by analyzing trade-offs between models and comparative tables. Issues such as dependence on labeled data or performance in adverse weather conditions need to be discussed. Instead of suggesting “improving generalization,” specific techniques should be proposed.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsI thank the authors for their work. The change in the manuscript is truly noticeable, an improvement is evident.
The suggested topics are addressed, and I thank them for doing so.
However, I still have some questions that we could call "minor" but that I would like to see addressed:
1) Why only the IEEE and Elsevier databases? I mean, they may be the "most popular," or the "most prestigious," or the "most recognized" according to Scopus? (Explain Why?). It should be clear in one or two sentences why they were chosen, why those databases were selected. Now, a question for you: Where does MDPI fit in? The manuscript is intended or written for publication in the journal "Agronomy" published by MDPI. I think it is pertinent to include that database. I believe that serious and relevant work can also be found there.
2) NOT all figures are mentioned in the text. Only Figures 1 and 2 are mentioned; the others are missing. Each and every figure must be cited in the appropriate place. And of course, since we're at this point, I ask you to review the captions again to verify that they are self-explanatory, that they guide the reader, and that they answer the questions "What do they show?" and "Why does it matter?"
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThank you for your excellent work in addressing my suggestions.
Author Response
Thank you so much. Your expertise to detail have helped me refine my research.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors have adequately addressed most of the concerns raised in the previous review. Specifically, the issue regarding the unauthorized use of images has been resolved, as the necessary evidence and corresponding references to prove permission for their use have now been provided. Furthermore, the depth of the discussion section has been significantly improved.
Author Response
Thank you so much. Your expertise and attention to detail have not only helped me refine my research but also enhanced my understanding of the subject matter.
