Deep-Learning-Based Rice Phenological Stage Recognition
Round 1
Reviewer 1 Report
The authors explored deep-learning techniques to map the phenological stages of rice crops on a localized scale. For this study, weather station-equipped RGB cameras were used to collect images over the complete growth cycle of the rice in the sample plots and saved them in a database. For image analysis, an object detection approach was used to clean the dataset and subsequently served as inputs in the deep learning analysis. For deep learning, a ResNet-50 backbone network was used to extract spatial feature information from the data archive over the six phenological stages (seedling, tillering, booting jointing, heading flowering, grain filling and maturity) the rice crop.
This study is a welcomed development and provides valuable insight into accurately detecting rice crops over their phenological stages.
The paper is well-outlined, clearly states the problems, and effectively outlines the methodology and results. As a technical note, I recommend this paper for publication with minimal correction.
Some minor comments:
Figure 3: Can the authors edit the text in this figure, please? It's not legible in its current format – I suggest increasing the text sizes. A size similar to the previous figure (Figure 2) is preferable.
The authors explored deep-learning techniques to map the phenological stages of rice crops on a localized scale. For this study, weather station-equipped RGB cameras were used to collect images over the complete growth cycle of the rice in the sample plots and saved them in a database. For image analysis, an object detection approach was used to clean the dataset and subsequently served as inputs in the deep learning analysis. For deep learning, a ResNet-50 backbone network was used to extract spatial feature information from the data archive over the six phenological stages (seedling, tillering, booting jointing, heading flowering, grain filling and maturity) the rice crop.
This study is a welcomed development and provides valuable insight into accurately detecting rice crops over their phenological stages.
The paper is well-outlined, clearly states the problems, and effectively outlines the methodology and results. As a technical note, I recommend this paper for publication with minimal correction.
Some minor comments:
Figure 3: Can the authors edit the text in this figure, please? It's not legible in its current format – I suggest increasing the text sizes. A size similar to the previous figure (Figure 2) is preferable.
Author Response
Response to Reviewer 1 Comments
Point 1: Figure 3: Can the authors edit the text in this figure, please? It's not legible in its current format – I suggest increasing the text sizes. A size similar to the previous figure (Figure 2) is preferable.
Response 1: I found the problem you mentioned and I have replaced the image in the original article, please see Figure 3 for the specific changes.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
A good research take up to do deep learning-based rice phenological stage recognition. Authors have presented their research nicely. However, there are few things which needs to be consider are:
1. Has author considered different dates or growth stage images for model training
2. Spectroradiometer based reflectance may be used to collect training and testing samples.
3. Sentences should be rewritten between line no 111 to 123, as objective is not clear to the general audience.
4. There is no discussion about hyperparameters, authors should discuss it first before using.
5. More citation required.
6. Provide codes whatever you have used in public domain, so that the scientific community get benefited based on your research.
Quality of english is OK. Minor change required. Author should check the grammar throughout submitted document.
Author Response
Response to Reviewer 2 Comments
Point 1: Has author considered different dates or growth stage images for model training.
Response 1: Thank you for your question. In this study, I recorded and analyzed the data of each fertility cycle of rice. The data of this phenological stage recognition model is relatively complete and can satisfy the accurate recognition of each phenological stage of rice.
Point 2: Spectroradiometer based reflectance may be used to collect training and testing samples.
Response 2: Thank you for your suggestion, I have not used the spectral data in this experiment yet, and I may use this data in the subsequent experiments. In the near future, I plan to combine the spectral data with visible light data to extract relevant features and then use machine learning to conduct deeper experiments.
Point 3: Sentences should be rewritten between line no 111 to 123, as objective is not clear to the general audience.
Response 3: The corrections you gave are very accurate, and I have corrected the places you mentioned. I have added changes to the previous content, and the updated content is "Based on the above analysis, we found that most of the existing solutions are based on the monitoring of the weather by drones and satellite remote sensing data. However, these solutions are unable to achieve high accuracy and continuous monitoring over long periods of time. Therefore, we need an identification method with higher accuracy and continuous observation of rice phenological periods to solve this problem. In this study, rice is chosen as the research object and a deep learning-based method is proposed to recognise rice RGB images for weather determination. This study focuses on the detection of rice germination using a deep learning approach in conjunction with the collection of relevant data from the rice growth cycle to design an experimental study."specific. The text content you can also find through the last paragraph of the first section of the manuscript, "Introduction".
Point 4: There is no discussion about hyperparameters, authors should discuss it first before using.
Response 4: Thanks to your kind suggestion, I added explanations before conducting each experiment, e.g. 5.2.2, and the additions are shown below "We utilized the ability of target detection to detect multiple classes of objects in an image to identify rice seedlings at the seedling stage in a large field. We selected three detection models, Yolov5, Yolov6 and Yolov7, before training the recognition models.During the training process, we ensured that the network structures of the three models were consistent with the training parameters. ".
Point 5: More citation required.
Response 5: Thanks to your suggestion, I have rewritten the "Discussion" section of Section IV of the article and added references to related content, which are now included in the revised manuscript. I will show some of these references below:
- Sarwinda, D., et al., Deep Learning in Image Classification using Residual Network (ResNet) Variants for Detection of Colorectal Cancer. Procedia computer science, 2021. 179: p. 423-431.
- Zhang, C., F. Kang, and Y. Wang An Improved Apple Object Detection Method Based on Lightweight YOLOv4 in Complex Backgrounds. Remote Sensing, 2022. 14, DOI: 10.3390/rs14174150.
- Yun, Z., et al., Precision detection of crop diseases based on improved YOLOv5 model . Frontiers in Plant Science, 2023. 13.
- Han, J., et al., Real-time detection of rice phenology through convolutional neural network using handheld camera images. Precision Agriculture, 2020. 22(1): p. 154-178.
- Sheng, R.T., et al. Rice Growth Stage Classification via RF-Based Machine Learning and Image Processing. Agriculture, 2022. 12, DOI: 10.3390/agriculture12122137.
- Cheng, M., et al., Application of deep learning in sheep behaviors recognition and influence analysis of training data characteristics on the recognition effect. Computers and Electronics in Agriculture, 2022. 198: p. 107010.
Point 6: Provide codes whatever you have used in public domain, so that the scientific community get benefited based on your research.
Response 6: Thank you for your suggestion, but I am very sorry. The code involved in this experiment is currently under patent application, if you really need it I can disclose the code after the patent application is finished.
Author Response File: Author Response.pdf
Reviewer 3 Report
Reviewers Comments
A major revision is being suggested for the manuscript id: remotesensing-2371279, titled, “Deep learning-based rice phenological stage recognition”. The following are specific comments; the author must revise the manuscript and prepare a rebuttal to the comments for further review.
1. The authors must follow the standard format for the submission of the article to the journal. The authors can explore it on https://www.mdpi.com/journal/remotesensing/instructions#preparation
2. The authors need to revise the introduction, Author must check the contents in lines no 113-117, “The experimental…….. crop gardens”, what is the need and use of these sentences over here? Kindly remove it. Also, the authors have collected and presented the contents nicely, but there are a few mistakes in the placement of these contents, which means what should be placed where. Hence, the authors can refer following articles for better understanding:
https://doi.org/10.1016/j.sciaf.2023.e01577
3. Current section 3: Experimental environment and configuration, section 4: Model evaluation indicators, and Section 5: the contents up to sub-section 5.1 in this paper (it is the methodology component), must be properly named and must be kept under Section 2: Materials and methods, as different subsections.
4. Line 214: is there any difference between eq. (2) and (3). Kindly correct it and also add the rest of the performance parameters deployed for the analysis of such developed models.
5. Current section 5 must be changed to Section 3: Results and discussion, then contents of subsection 5.2 onwards should be kept in it.
6. Results are poorly written, and no critical analysis (root cause) and associate discussions are presented. Authors must also figure out the reasons through the discussion for the claimed improvement in the performance.
7. The contents of current section 6: the first and second paragraphs, need to be moved to the Results and discussion section and the rest 2 paragraphs must be rewritten in 150-200 words and kept under new section 4: Conclusions.
8. Authors must also ensure the flow of the writeup must be proper and should not be impacted due to shifting of the contents from one section to another.
I wish authors a great success.
Author Response
Response to Reviewer 3 Comments
Point 1: The authors must follow the standard format for the submission of the article to the journal. The authors can explore it on https://www.mdpi.com/journal/remotesensing/instructions#preparation
Response 1: Thank you for the constructive suggestions. I have re-referenced the standard format for journal submissions and corrected the overall structure of the article, please check the latest manuscript for updates.
Point 2: The authors need to revise the introduction, Author must check the contents in lines no 113-117, “The experimental…….. crop gardens”, what is the need and use of these sentences over here? Kindly remove it. Also, the authors have collected and presented the contents nicely, but there are a few mistakes in the placement of these contents, which means what should be placed where. Hence, the authors can refer following articles for better understanding:
https://doi.org/10.1016/j.sciaf.2023.e01577
Response 2: The changes you gave are very accurate, the previous paragraph was not placed there properly. I referred to the article you provided and made changes to the places you mentioned. The last paragraph of the introduction I changed to "Based on the above analysis, we found that most of the existing solutions are based on the monitoring of the weather by drones and satellite remote sensing data. However, these solutions are unable to achieve high accuracy and continuous monitoring over long periods of time. Therefore, we need an identification method with higher accuracy and continuous observation of rice phenological periods to solve this problem. In this study, rice is chosen as the research object and a deep learning-based method is proposed to recognise rice RGB images for weather determination. This study focuses on the detection of rice germination using a deep learning approach in conjunction with the collection of relevant data from the rice growth cycle to design an experimental study.".
Point 3: Current section 3: Experimental environment and configuration, section 4: Model evaluation indicators, and Section 5: the contents up to sub-section 5.1 in this paper (it is the methodology component), must be properly named and must be kept under Section 2: Materials and methods, as different subsections.
Response 3: Thank you for your constructive suggestions. I have reorganized the chapter distribution and arrangement of this article, and put the experimental settings, parameter configuration and model evaluation indicators in section 2.3 and 2.4 of the article. I have named section 3 “Results”, section 4 “Discussion”, and section 5 “Conclusion”.
Point 4: Line 214: is there any difference between eq. (2) and (3). Kindly correct it and also add the rest of the performance parameters deployed for the analysis of such developed models.
Response 4: Thank you very much for raising the issue I have here, and I have made changes to this section. Equation (3) is corrected to Accuracyl = (TP+TN)/(TP+TN+FP+FN).In addition, I have added relevant parameters and explanations for each experiment involved, please check the latest manuscript for details, thanks.
Point 5: Current section 5 must be changed to Section 3: Results and discussion, then contents of subsection 5.2 onwards should be kept in it.
Response 5: Thank you for your advice, which is valuable in improving the quality of the manuscript.I repositioned the content of Section 5 to go in the new Section 3, and in addition I kept the content of the original subsection 5.2 in it.
Point 6: Results are poorly written, and no critical analysis (root cause) and associate discussions are presented. Authors must also figure out the reasons through the discussion for the claimed improvement in the performance.
Response 6: Thanks to your kind suggestion, I have rewritten the results section to describe the content as a whole in three parts, the first part is the identification of the phenological period, the second part is the data cleaning, and the third part is the different data and identification accuracy. The process and results of the three experiments can be viewed in the latest manuscript 3.1, 3.2 and 3.3.
Point 7: The contents of current section 6: the first and second paragraphs, need to be moved to the Results and discussion section and the rest 2 paragraphs must be rewritten in 150-200 words and kept under new section 4: Conclusions.
Response 7: Thank you for your suggestions. First, we have revised the original section 6 by moving the first and second paragraphs to the "Results" section in section 3. Secondly, we have also added section 4, "Discussion", which discusses the data collection method, the results of the experimental data, the causes of the results, and the comments on the four levels of the experiment, and added several references. Finally, we have rewritten the conclusion section of the original paper to describe the concluding content of the paper and to suggest ways to improve the design of the subsequent experiments.
Point 8: Authors must also ensure the flow of the writeup must be proper and should not be impacted due to shifting of the contents from one section to another.
Response 8: Thank you for your comments and suggestions on this article. The previous article was indeed purely in some problems, and I have made a comprehensive discussion and adjustment of the logical relationships and related contents of the article. Please refer to the latest submitted manuscript for the revision of the content of the manuscript.
We sincerely hope that this revised manuscript has addressed all your comments and suggestions. We appreciated for reviewers’ warm work earnestly,and hope that the correction will meet with approval.Once again,thank you very much for your comments and suggestions.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Reviewers Comments
A minor revision is being suggested for the manuscript id: remotesensing-2371279_Revised v1, titled, “Deep learning-based rice phenological stage recognition”. The following are specific comments; the author must revise the manuscript and prepare a rebuttal to the comments for further review.
1. The authors are advised to avoid the use of first and third person in their research writing. Check the whole manuscript and update.
2. The authors should avoid unnecessary capitalization in their writing. Check the whole manuscript and revise.
3. The research gap needs to identify properly, and objectives must be in coordination with the presented work.
4. Line 107, “Based on the above analysis,”, is there any analysis presented in introduction section? kindly reframe the sentence appropriately and check the whole manuscript for such dangling sentences.
5. The current conclusion is almost in flow with the discussion part. However, it is expected that author must conclude the major findings of the research with quantitative support (Numbers). Hence, the authors must rewrite the conclusions (150-200 words max).
I wish authors a great success.
Author Response
Response to Reviewer 3 Comments
Point 1: The authors are advised to avoid the use of first and third person in their research writing. Check the whole manuscript and update.
Response 1: Thank you for your kind suggestions, I have revised most of the first and third person sentences in the article. For example, I changed "In this study, we gathered image data of rice growth in the entire growth cycle. The collected data was divided into 11 stages, based on the experimental design and data acquisition methods discussed in Section 2.1. To evaluate the performance of our model, we randomly selected 2539 samples as the test set, with a ratio of 9:1, while the remaining 22846 images were divided into training and validation sets, also with a ratio of 9:1." to "In this study, image data of rice growth throughout the entire growth cycle was gathered. The collected data was divided into 11 stages, based on the experimental design and data acquisition methods discussed in Section 2.1. To evaluate the model's performance, 2539 samples were randomly selected as the test set, maintaining a ratio of 9:1. The remaining 22846 images were divided into training and validation sets using the same 9:1 ratio.." Please see the latest manuscript submission for other changes.
Point 2: The authors should avoid unnecessary capitalization in their writing. Check the whole manuscript and revise.
Response 2: Thank you for your constructive comments, I checked the article for capitalization issues. I changed "S-T", "T-B", "B-H", "H-G", "G-M" to "seedling to tillering", "tillering to booting jointing", "booting jointing to heading flowering", "heading flowering to grain filling" and "grain filling to maturity".
Point 3: The research gap needs to identify properly, and objectives must be in coordination with the presented work.
Response 3: Thank you for your suggestions. I have rephrased the description of the research objectives in the manuscript based on your feedback, hoping that this revision meets the journal's requirements. For specific details of the modifications, please refer to the latest version of the manuscript.
Point 4: Line 107, “Based on the above analysis,”, is there any analysis presented in introduction section? kindly reframe the sentence appropriately and check the whole manuscript for such dangling sentences.
Response 4: Thank you very much for asking my question here and I have revised this part. I changed the content of line 107 of the introduction "Based on the above analysis, we found that most of the existing solutions are based on the monitoring of the weather by drones and satellite remote sensing data. However, these solutions are unable to achieve high accuracy and continuous monitoring over long periods of time. Therefore, we need an identification method with higher accuracy and continuous observation of rice phenological periods to solve this problem. In this study, rice is chosen as the research object and a deep learning-based method is proposed to recognise rice RGB images for weather determination. This study focuses on the detection of rice germination using a deep learning approach in conjunction with the collection of relevant data from the rice growth cycle to design an experimental study." to "Given the challenges of accurately identifying mixed dense rice plants within complex field environments, it is crucial to develop a recognition method that offers higher accuracy and enables continuous observation of rice phenological periods. This study aims to address this issue by focusing on rice plants in a large field setting and proposing a deep learning-based approach for recognizing RGB images of rice to determine its phenological stages. Specifically, the study emphasizes the detection of rice germination using deep learning techniques and incorporates data collection on the rice growth cycle to design an experimental study."
Point 5: The current conclusion is almost in flow with the discussion part. However, it is expected that author must conclude the major findings of the research with quantitative support (Numbers). Hence, the authors must rewrite the conclusions (150-200 words max).
Response 5: Thank you for your suggestion, it will greatly improve the quality of the manuscript. I have rewritten the conclusion section to read "By focusing on the identification of mixed dense rice plants in complex field environments, comprehensive field data was collected and processed throughout the growth and development stages of rice. This study introduced an enhancement to the traditional phenological method by integrating the ResNet-50 algorithm with the Yolov5 algorithm, resulting in a model identification accuracy of 87.33% and a The experiments convincingly demonstrate the effectiveness of this method in accurately identifying crop The experiments convincingly demonstrate the effectiveness of this method in accurately identifying crop phenology, highlighting the potential for the developed rice phenology identification model to be sustainable and practical for long-term crop monitoring."
We sincerely hope that this revised draft can meet all of your comments and suggestions. We appreciate your hard work and enthusiastic contributions, as they have allowed us to improve and refine the article. We genuinely hope that the revised manuscript will receive broad recognition. Once again, we sincerely thank you for providing valuable feedback and suggestions.
Author Response File: 
Author Response.docx
