Research on the Yunnan Large-Leaf Tea Tree Disease Detection Model Based on the Improved YOLOv10 Network and UAV Remote Sensing
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsI'm sending comments in the attached PDF. I suggest adding the types of models used to the title, as the bulk of the manuscript is based on that description, not just on the use of UAVs.
Comments for author File: Comments.pdf
Author Response
Thanks very much for your time to review this manuscript. I really appreciate you’re your comments and suggestions. We have considered these comments carefully and tried our best to address every one of them. And the corresponding part in the text has been modified using red font.
- I suggest adding the types of models used to the title, as the bulk of the manuscript is based on that description, not just on the use of UAVs.
Modification instructions: Thank you for the valuable feedback on our manuscript. In response to your suggestion, we have incorporated the type of model used in the title to better reflect the core content of the paper. “Research on the Yunnan Large-leaf Tea Tree Disease Detection Model Based on the Improved YOLOv10 Network and UAV Remote Sensing”
- Abstract: It is suggested to indicate which diseases, whether they are the most important or just a group of diseases.
Modification instructions: Have been changed in accordance with the advice given. “Through testing of common diseases, the research results indicate that, for the improved YOLOv10 network, the Box Loss, Cls Loss, and DFL Loss were reduced by 15.94%, 13.16%, and 8.82% respectively in the One-to-Many Head, and by 14.58%, 17.72%, and 8.89% respectively in the One-to-One Head. Compared to the original YOLOv10 network, Precision, Recall, and F1 increased by 3.4%, 10.05%, and 6.75% respectively.”
- Line 35 Are all types of tea? Line 39 – 41 The field inspection method will always be necessary Line 42-43 Tea is not a staple food, it could not be considered as urgent. Line 62 – 64 Were the studies conducted under controlled conditions (greenhouses)? mention if they were in greenhouses. Line 71 Mention what diseases. It is necessary to describe the relevance of this study to tea producers regarding tea diseases. It focuses only on models developed for other crops and does not provide insight into the problem or the advantages of using technology to benefit agricultural production, especially tea cultivation. It should also specify the type of tea (in other countries, infusions of other plants are called tea).
Modification instructions: Thank you for your detailed suggestions regarding our paper. We have made revisions based on your recommendations and provided detailed responses and improvements for each point raised.
(1) In line 35, the focus of this study is on "Yunnan Large-Leaf tea trees," rather than all types of tea.
(2) In lines 39-41, we have revised the sentence to make the expression more rigorous, emphasizing that field inspection methods remain highly necessary.
(3) In lines 42-43, we have adjusted the phrasing to clarify that the urgent issue to be addressed is how to quickly and accurately obtain disease information in tea plantations, rather than an urgent problem in the entire agricultural sector.
(4) In lines 62-64, regarding your comment on "whether the research was conducted in controlled environments like greenhouses," we would like to clarify that the model studied was tested in an outdoor organic tea plantation to ensure its effectiveness in real tea fields, rather than in a greenhouse environment.
(5) In line 71, regarding your request to "clearly state the specific diseases studied and describe the relevance to tea producers," we have made the appropriate revisions in the manuscript.
(6) In response to your comment, “It focuses only on models developed for other crops and does not provide insight into the problem or the advantages of using technology to benefit agricultural production, especially tea cultivation,” we have revised this section. The updated content reads: "This study aims to provide an efficient, flexible, and highly adaptable technical solution for the intelligent detection of agricultural diseases, thereby promoting the development of intelligent agricultural monitoring systems."
(7) Regarding your suggestion, "It should also specify the type of tea (in other countries, infusions of other plants are called tea)," we have clearly indicated in the paper that the research focuses on "Yunnan Large-Leaf tea," to avoid confusion with teas from other plants.
- Line 86-89 It is recommended to include a location map of the study area. Line 91 "as 8761 °C" is not understood. Line 98 – 99 At what wavelength does the sensor capture images? How fast did they fly? Were images superimposed? Were photogrammetric methods used? What time of day were the photos taken? Line 107 – 108 How did they know what diseases they were? Were they PCR-confirmed? Line 140 – 144 There will always be limitations in images captured by UAVs when the disease density is low. Line 207 -209 Is the spatial resolution the same for all images? Since they were taken at different distances and orientations, does that affect the spatial resolution? Comment.
Modification instructions: Thank you for your detailed suggestions regarding our paper. We have carefully reviewed them and made corresponding revisions and additions based on your recommendations.
(1) In lines 86-89, to help readers better understand the geographical context of the study area, we have added location information for the research region in the paper. The tea plantations in Yunxian and Fengqing counties of Lincang City, Yunnan Province are located at 99°E, 24°N, and the organic tea plantations in Xishuangbanna are located at 100°E, 21°N.
(2) In line 91, thank you for your correction. In the original text, “8761” represents the effective accumulated temperature, which refers to the total temperature above the threshold that has a beneficial effect on plant growth and development, representing the total effective temperature for the entire year. To avoid any confusion for the readers, we have revised this section. Since this parameter is only relevant to the data collection site and is not essential, we have removed it.
(3) In lines 98-99, thank you for your detailed question. We have added more detailed information in the paper regarding the sensor wavelength, flight speed, image overlap, and image collection time, in order to better clarify the specific parameters of the experimental process. The collected image wavelength range is visible light, with the UAV flying at a speed of 2 meters per second, capturing images every 0.5 seconds. To ensure the generalizability of the final model, image data were collected between 7-9 in the morning, 12-14 at noon, and 16-18 in the afternoon.
(4) In lines 107-108, regarding your comment about how the diseases were identified, the diagnosis of the diseases in this study was done through the review of multiple experts, rather than using PCR methods. We invited a team of experts with relevant experience in the field to diagnose the diseases, ensuring the accuracy and reliability of disease identification.
(5) In lines 140-144, regarding your comment that "when disease density is low, UAV-captured images always have certain limitations," we agree that this is a common challenge. The primary issue arises from factors such as the precision of the UAV's imaging and flight altitude, which may make it difficult to obtain sufficiently clear images in low disease density areas. Our improvements to the YOLOv10 network are intended to address these related issues to some extent. The Shape-IoU optimization targets detection errors caused by disease overlap and occlusion due to varied shapes, the Wavelet Transform Conv optimization addresses the model’s weak response to low-frequency information, and the Histogram Transformer optimization addresses the issue of fixed receptive fields that cannot simultaneously consider both local details and global background information. These improvements aim to overcome the accuracy limitations caused by the constraints of UAV-based imaging.
(6) In lines 207-209, regarding your question "Are the image resolutions the same? Will the distance and angle differences during capturing affect the resolution?" Since the same imaging equipment was used, the resolution of all captured raw images is identical. To ensure consistency during the training process, all images were automatically resized to a uniform 640×640 resolution during model training, thus avoiding the impact of different shooting distances and angles on image resolution. Furthermore, to address the image blurring caused by flight, we applied data augmentation to the raw images.
- Line 292 – 297 This is the objective of the study, mentioning it at the end of the introduction. Furthermore, it is already implicit in the previous text. Line 356 -340 Is the performance in detecting different diseases due to the type of disease or the model? Specify this information. Line 405 – 408 Could the effects of the plant in the image (such as shadows) or the conditions of the tea plant be confounding disease detection? For example, couldn't a nutrient deficiency in the plant be mistaken for the effects of a disease? When using visible wavelength images, these differences can't be observed, and they could be classifying a disease as a plant deficiency. The model could be performing well, but the classification of what is being measured could be wrong.
Modification instructions: Thank you for your valuable feedback on our paper. We have made the corresponding revisions based on your suggestions.
(1) In lines 292-297, we have made adjustments in the paper to avoid repetition and to clarify the structure of the introduction.
(2) In lines 356-340, we have explicitly pointed out in the text that the values of the elements in the matrix are influenced both by the type of disease and by the results of model optimization and improvement.
(3) In lines 405-408, thank you very much for your insightful question. The imaging effects of plants and the physiological condition of the plants themselves can indeed interfere with disease detection. To minimize such interference, we further employed data augmentation techniques during the construction of the dataset. Through multi-angle and multi-timepoint image collection, we aimed to reduce the impact of these factors on model classification. Our test results, with the confusion matrix not yielding 1 and the mAP value not being 100%, are precisely due to the reasons you pointed out. We greatly appreciate your professional insights, and we will conduct further in-depth studies on this issue in future experiments and try to address the interference caused by imaging effects and the physiological condition of the plants.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsA partly annotated version is provided. The manuscript fails to clarify how the very good performance in visualizing symptoms if linked to the disease(s), no clarification is provided about the false positive and false negative results i.e. validation for the verification of disease presence by laboratory detection of the pathogen for example.
Comments for author File: Comments.pdf
Some English wording must be clarified as listed in the partly annotated version provided.
Author Response
Thanks very much for your time to review this manuscript. I really appreciate you’re your comments and suggestions. We have considered these comments carefully and tried our best to address every one of them. And the corresponding part in the text has been modified using red font.
- Remove “intelligent”.
Modification instructions: Have been changed in accordance with the advice given.
- Change all instances of “garden” to “cultivation”.
Modification instructions: Have been changed in accordance with the advice given.
- Change all instances of “gardens” to “cultivations”.
Modification instructions: Have been changed in accordance with the advice given.
- Change all instances of “in order to” to “to”.
Modification instructions: Have been changed in accordance with the advice given.
- Add “crops”.
Modification instructions: Have been changed in accordance with the advice given.
- please verify this number.
Modification instructions: Thank you for your valuable feedback on our paper. Regarding the unclear meaning of the value “8761” mentioned in line 91, in the original text, “8761” represents the effective accumulated temperature, which refers to the total temperature above the threshold that has a beneficial effect on plant growth and development, representing the total effective temperature for the entire year. To avoid any confusion for the readers, we have revised this section. Since this parameter is only relevant to the data collection site and is not essential, we have removed it.
- Remove extra spaces.
Modification instructions: Have been changed in accordance with the advice given.
- There is a wording error in Figure 1.
Modification instructions: Have been changed in accordance with the advice given.
- Change all instances of “original” to “diseased”.
Modification instructions: Have been changed in accordance with the advice given.
- There is no explanation provided for “Params” and “Arguments”.
Modification instructions: Have been changed in accordance with the advice given. Params represents the number of parameters used in the network module. Module column represents the module of the layer. Arguments is the specific parameter configuration of the module, which describes the detailed information of the input and output dimensions of the module.
- What are the characteristics of the validation set in five-fold cross-validation?
Modification instructions: In five-fold cross-validation, the dataset is randomly divided into five roughly equal parts. In each iteration, four parts are used as the training set, while the remaining one part is used as the validation set. Each part will take turns serving as the validation set, ensuring that every part of the data is used as the validation set once.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe content of the manuscript is part of the works on the possibility of monitoring the health of agricultural crops using the analysis of images obtained from an aerial drone. In this case, it is tea cultivation. The authors' research focus is on improving the technique of image analysis and automatic diagnosis.
The authors report the analyses performed in a precise manner. The structure of the manuscript is correct. I also see some shortcomings. I believe that the authors should express their opinion regarding the repeatability of the obtained results. This concerns repeatability in the temporal and spatial sense. The question also arises whether the applied methodology will be correct for other types of tea.
In line 91, it is not known what the number 8761 represents (probably not the temperature...).
In figures 3 and 6 the letters are too small. Although the reader can enlarge the image on the monitor, the letters become blurry.
Should line 321 "dfl" be in lowercase since the words in the expansion of this acronym start with an uppercase letter?
In my opinion, the caption for Figure 8 is too succinct.
The acronym YOLOv10 is explained up until line 74 (and appears already in the abstract).
Author Response
Thanks very much for your time to review this manuscript. I really appreciate you’re your comments and suggestions. We have considered these comments carefully and tried our best to address every one of them. And the corresponding part in the text has been modified using red font.
- The authors report the analyses performed in a precise manner. The structure of the manuscript is correct. I also see some shortcomings. I believe that the authors should express their opinion regarding the repeatability of the obtained results. This concerns repeatability in the temporal and spatial sense. The question also arises whether the applied methodology will be correct for other types of tea.
Modification instructions: Thank you for your valuable feedback and thorough review of our paper. Regarding your concerns about the 'reproducibility of results, particularly in terms of temporal and spatial dimensions' and 'whether the applied methods are suitable for other types of tea,' we have added detailed parameters of our data collection in the text, including the geographic coordinates of multiple time periods and collection sites, as well as the UAV's detailed settings during data collection. Furthermore, we have clarified that the subject of this study is the Yunnan large-leaf tea variety. Considering that the growth characteristics and disease manifestations of other tea types may differ from those of Yunnan large-leaf tea, our method has a certain level of generalizability. It can provide a reference for disease detection in other types of tea, but its application may require dataset adjustments and model retraining based on the specific characteristics of each tea variety.
- In line 91, it is not known what the number 8761 represents (probably not the temperature...).
Modification instructions: Thank you for your valuable feedback on our paper. Regarding the unclear meaning of the value “8761” mentioned in line 91, in the original text, “8761” represents the effective accumulated temperature, which refers to the total temperature above the threshold that has a beneficial effect on plant growth and development, representing the total effective temperature for the entire year. To avoid any confusion for the readers, we have revised this section. Since this parameter is only relevant to the data collection site and is not essential, we have removed it.
- In figures 3 and 6 the letters are too small. Although the reader can enlarge the image on the monitor, the letters become blurry.
Modification instructions: Have been changed in accordance with the advice given. We have made adjustments to some areas where the font size was too small, ensuring that the text in the images is clearer and easier to read. Since the original images have a high resolution, directly inserting them into the article would cause the file to open slowly. However, if needed, we can provide the original high-resolution images for your further review.
- Should line 321 "dfl" be in lowercase since the words in the expansion of this acronym start with an uppercase letter?
Modification instructions: Have been changed in accordance with the advice given. Thank you for your valuable feedback on our paper. We have thoroughly reviewed the entire manuscript and have standardized the term “DFL” to uppercase to maintain consistency and accuracy.
- In my opinion, the caption for Figure 8 is too succinct.
Modification instructions: Have been changed in accordance with the advice given. “Figure 8. Comparison of Precision, Recall, and F1 before and after the YOLOv10 network improvement.”
- The acronym YOLOv10 is explained up until line 74 (and appears already in the abstract).
Modification instructions: Have been changed in accordance with the advice given.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe revisions done are substantially acceptable.