Round 1

Reviewer 1 Report

 Review Comments

    The presented work explained a RiceDRA-Net 10 deep residual network model and use it to identify four different of rice leaf diseases. They name the rice leaf disease test set with a complex background the CBG-Dataset, and construct a single background rice leaf disease test set, the SBG-Dataset, based on the original dataset. The Res-Attention module uses 3 × 3 convolutional kernels and denser connections than other attention mechanisms to reduce information loss. The experimental results show that RiceDRA-Net achieves a recognition accuracy of 99.71% on the SBG-Dataset test set and possesses a recognition accuracy of 97.86% on the CBG-Dataset test set. However, the following major corrections can be considered by the authors to further improve the quality of the manuscript.

 I have some major corrections and suggestions below:-

1. Organization of the paper can be added at the end of introductions.

2. The computational complexity of the algorithm must be discussed. Also, compare the proposed method in terms of computational complexity?

3. Literature survey is missing and need to be modified based on current state of art methods. Some more paper based on current study in identification of Rice Leaf Diseases must be added.

4. Comparative analysis with respect to various performance metrics is missing for various data sets? The comparison can be a bit unfair if different data is not used for comparative analysis.

5. Has the Author implemented the architecture from scratch and identified the novel condition in deep networks.

6. Visualized Results with respect to various categories of diseases detection must be discussed and presented.

7. Precision vs. recall curves of the proposed algorithms with respect to data sets must be included.

8. Comparative analysis of various performance parameters with respect to other data sets and ground truth data sets must be discussed.

9. Limitations and future work of the proposed work can be added and discussed.

10. In all results tables’ utilized models/methods like in table 6, 7 and 8 etc. must be cited with proper and specific citations.

11. How much data should be considered for training and testing for architecture implementation? Details of training and testing data sets must be tabulated.

12. Comparative analysis with respect to inference/fps and real-time time analysis is missing?

13. Various abbreviations also must be included.

14. Resolution and clarity of figure 9, 10, 23, 14, 15 are very poor.

15. How the loss function has been chosen.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Although the submitted manuscript is not too novel, it is very well written. Namely, it is in good English and the description of the related and own work is good. I think the manuscript can be accepted after a minor revision. Issues to be addressed:

1) THe authors did not report on the applied programming languages, libraries, and hardware configuration.

2) In Table 2, meaning of variable a is unclear.

3) Captions of Fig. 3-8 could be more detailed. Moreover, captions should be self contained.

4) Authors use a lot of abbreviations. This is why, list of abbreviations would be useful at the end of manuscript. MDPI template offers this.

5) Study about typical error cases would be welcomed.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

 

 RiceDRA-Net: Precise Identification of Rice Leaf Diseases in  Complex Backgrounds Using Res-Attention Mechanism

Line 10:

 The researchers develop the RiceDRA-Net 10 deep residual network model and use it to identify four different of rice leaf diseases.

Abstract doesn’t show that, novel dataset is created or not?

Line 23:

 demonstrating its strong robustness and stability.cases.

 

Kindly proofread the entire paper for such errors carefully.

Then reference 1.

Wang, Y.; Wang, H.; Peng, Z.J.E.S.w.A. Rice diseases detection and classification using attention based neural network and 482 bayesian optimization. 2021, 178, 114770.Author 1, A.; Author 2, B. Title of the chapter. In Book Title, 2nd ed.; Editor 1, A., Editor 483 2, B., Eds.; Publisher: Publisher Location, Country, 2007; Volume 3, pp. 154–196. [CrossRef]

 

 

Many Significant studies are not covered in the Introduction and related works like:

Deep feature based rice leaf disease identification using support vector machine-2020

Rice Leaf Disease Classification Using Deep Learning and Target for Mobile Devices

Weed density estimation in soya bean crop using deep convolutional neural networks in smart agriculture

Recognition of leaf disease using hybrid convolutional neural network by applying feature reduction

 

Kindly rewrite the contributions, 2^nd contribution is not significant can be removed.

 I see limited novelty in paper, no mention of epocs?

 

 

 

 

 

 

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

All my comments and concerns has been added successfully.Now paper is acceptable for publication. I accept it in current form.

Reviewer 3 Report

Paper is modified as per comments.

Now readabilty and understanding for work is improved.