LWSDNet: A Lightweight Wheat Scab Detection Network Based on UAV Remote Sensing Images
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis study proposes an unmanned aerial vehicle based automatic lightweight detection method for wheat scab, which can effectively improve the intelligence level of wheat scab detection. However, there are still the following shortcomings and areas that need to be improved in this study:
1. Throughout the entire paper, try not to use first person expressions, such as "we, I"
2. Please provide a complete description of the abbreviations that first appear in the main text, such as CSP, FPN, and NMS on line 87. Please refer to the entire text for more information
3. Please adjust the size of the image in line 183 so that the image and title appear on the same page; The same issue also applies to the image on line 397
4. Please explain the reason for choosing YOLOv5 as the benchmark network and why we did not choose a more advanced YOLO series network, such as YOLOv8
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article proposes a UAV based wheat scab detection system that improves
accuracy of existing systems. There is no doubt promise for the techniques
discussed exists. The article has so many demonstrative statements stating
amazing superiority but often lacks any statistical backing to the statement.
Further does the training process, obviously requiring user input to build
the training, create a typical anomaly that if you train with lots of
information will the results be of positive virtue. It is in fact impossible
for the results not to turn out outstanding. Can this oe training be applied
to other data sets? Else training is required for every example. Can
these results be replicated? Will a better camera produce better results or
can less data essentially duplicate the results. While an interesting example
of use of technology to analyze the article is weak in truly proving this
is economically worth doing and if it will consistently work.
Line 11-13 strange English for entire sentence and define "small"
Line 13-14 reword as cropping and image enhancement already exist
Line 27-28 3.2M is an abbreviation that needs defining
Line 41-43 need references
Line 44 define "highly precise"
Line 75 - be consistent in significant figures in % I think .01 is not valid
Line 85 - "achieved certain research results" is strange English
Line 87 etc. - you use a lot of abbreviations without defining them first
Line 163 - accuracy compared to what? By itself that term has no meaning
Line 169 - what is a "certain flying altitude"? Use numbers.
Line 170 - GPS or GNSS? Is IMU included?
Line 176 remove "Besides"
Line 192 why did you use 0.15? "Article" should be "test information".
Line 224 "pixel values" is strange term, do you mean "intensity"?
Line 227-228 take out after , as not needed
Line 248-249 Labeling software - is this commercial or built by you
Line 258-260 implies analysis was done on the UAV - this is not true I think
Line 260-261 what is unique about UAV images?
Line 269 "sacb" is scab
Line 280 remove "average"
Line 284 etc. "of network" should be "of the network" in all cases
Figure 7 & 8 Hadamard not discussed in text near figures
Line 426-429 lots of numbers but no discussion of why those numbers are used
Line 445-447 singular/plural issues
Line 459-460 what parameters?
Line 465 define "minimal" no numbers
Line 490 - why is 1% impressive?
Line 512 1% is promising?
Line 515 prove 2.7% is significant
Line 520-526 vs. 79.7% - show why significant or is it within 3 std. dev?
Line 549 why is 6.7% significant?
Line 621-623 as with major portions of the article - terms such as unequivocally,
superior performance, etc. are utilized redundantly and it is impossible
to statistically verify these statements from data presented.
There is a lot of English clean-up required that indicates consistent issues derived from AI analysis of the English
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe reviewed paper presents the concept of detecting wheat scabs using UAV imaging and a new proposed method. The article presents the problem comprehensively, refers to the literature, and presents the idea of the algorithm as well as its implementation and results. However, the reviewer has a few comments:
1. It is necessary to correct the caption under Fig. 1 is now on the next page.
2. What about images taken from different heights? How will the method cope with having images from different distances to crop? Has this issue been studied?
3. Is each image subjected to a change in contrast in order to visualize the disease or only those prepared for the training dataset? Was the algorithm tested for images without changing the contrast?
4. The paper describes a methodology using the Yolov5 model and compares it in the analysis of the results, among others. to v3 and v4. The Yolov9 model is currently available. According to the reviewer, the presented results, although good, use a relatively old model. Was this intentional? Have tests been carried out on newer models? Their results would probably be more interesting.
5. The paper does not refer to the results presented by other authors dealing with this topic and there is no comparison of these results with the method developed by the authors - key points should be added.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsGreat improvements the authors have listened and therefore created a much better article suitable for publication. The only issue I have is there is no proof with only limited data how it will work in general, or is this statistically valid improvements as not enough testing exists.
Comments on the Quality of English LanguageWhile improved I feel an accomplished English review should be performed before publication.
