A Lightweight Pine Wilt Disease Detection Method Based on Vision Transformer-Enhanced YOLO

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In current methods for tree disease detection, convolutional neural networks (CNNs) are commonly utilized for network architecture, leveraging their strong performance in feature extraction. However, the sensory field of CNNs is constrained by kernel size and network depth, limiting their capacity to effectively model long-term dependencies.

Authors proposed a lightweight pine wilt disease detection method based on ViT enhanced YOLO(Light-ViTeYOLO). This novel lightweight module is used to construct an EfficientViT feature extraction network for global receptive field and multi-scale learning. The algorithm effectively reduces the number of parameters and giga floating-point operations per second (GFLOPs) of the detection model while enhancing overall detection performance.

Does the paper contribute to the body of knowledge?

Yes, the authors conducted research and developed algorithm and software module that can increase the speed and performance of detection of tree disease.

Is the paper technically sound?

Yes. The information in the article is clearly structured. Figures and graphs have detailed titles.

Is the subject matter presented in a comprehensive manner?

Yes. The proposed methodology is sufficient. The authors also confirmed the effectiveness of their module experimentally.

 

Questions and comments:

-        The average annual loss from PWN in China was 7.17 billion yuan from 1998 to 2017 [Xu, Q.; Zhang, X.; Li, J.; Ren, J.; Ren, L.; Luo, Y. Pine Wilt Disease in Northeast and Northwest China: A Comprehensive Risk Review. Forests 2023, 14, 174. https://doi.org/10.3390/f14020174]. In this regard, is it possible to roughly estimate the economic effect of using your method?

-        How can your method be combined with space-based forest monitoring?

-        Besides PWN, what other pine diseases can your method be used for?

-        How exactly can you increase the effectiveness of your method?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript describes methods for detecting pine wilt disease based on images taken by aerial vehicles (UAV). This is a convolution neural network (CNN) YOLO type method which is modified to be “lightweight” here such that it is simpler and faster with fewer parameters than conventional CNN methods.

 

The application of this method is interesting and the results show that it is effective. However, there are a number of big issues related to the novelty of the method and referencing. See the below comments

 

Comment 1) The referencing is the text is very poor.

a) In some cases just mentioning a single author name without a number in brackets (e.g. [1]).

b) In many cases facts or information is introduced without any reference.

c) Sometimes the authors state “literature” without any reference

e.g. “Literature used Faster R-CNN and YOLOv3…..” without a reference

There are 36 references in the reference list but I think only around 4 have been referenced correctly in the text.

 

Comment 2) Following comment 1 it is difficult to assess the novelty of this work because the referencing is poor/missing. However, I am not convinced there is any significant novelty in the methods. For example there are multiple published articles I have found which discuss and apply lightweight style YOLO methods. The novelty of this work against state-of-the-art existing methods must be clarified. Some recent works are given below. If the novelty is that this is the first lightweight methods to be applied specifically to pine wilt disease this should be clarified.

https://doi.org/10.3390/s23146423

 

 https://doi.org/10.1016/j.compag.2023.108562

https://doi.org/10.3390/s23156811

 https://doi.org/10.1016/j.compag.2023.107905

https://doi.org/10.3390/rs15204974

 

In particular the study of Zhang et al. (https://doi.org/10.3390/rs15204974) uses a lightweight YOLO method and also uses an attention mechanism and CARAFE similar to this study

comment 3) The abstract gives values “3.89” and “7.4”. It should be stated what these numbers refer

Comment 4) The new method is only compared conventional CNN type methods in tables 4.2 and 4.3. It would be good to compare against other existing lightweight approaches.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript has been improved.

 

However, there are still a couple of minor issues

 

issue 1) In a couple of places the authors still refer to "Literature" when referring to references [8] and [20]. I think it is more appropriate to refer to the authors names e.g. Wu et al.

 

Issue 2) Many of the equations, tables and figures are not aligned correctly. I suspect this can be rectified by using the correct styles in the template

 

Issue 3) a number of the references in the reference list do not mention all the authors. I think in the reference list all authors should be mentioned (without using et al.)

Author Response

Please see the attachment.

Author Response File: Author Response.pdf