BPN-YOLO: A Novel Method for Wood Defect Detection Based on YOLOv7

Round 1

Reviewer 1 Report (New Reviewer)

Comments and Suggestions for Authors

This paper covers only one application of defect detection in the wood industry - the result cannot be generalized. Explain in more detail for which type of material and for which industrial problem the image data set used is representative.

Line 44: Placement of the citation may lead to misunderstanding: Vibration-based methods are common in the wood industry but not suitable to localize defects but they allow to determine an overall-value for elasticity/stiffness for a certain piece of wood. Therefore, please shift "[4,5]" after "Ultrasound-based ..."

Line 60: do you mean "dead knots"?

Line 62: "just captures an image of the wood" makes the reader think that it is as easy as taking a snotshot by a mobile phone. In fact, many more has to be considered for image quality suitable for further processing: Wood surface (rough or planed) and reflection properties, lighting and background correction, speed of the wood pieces (and integration time of the camera, image format and resolution, other factors like dust and vibration etc.

Line 90: Please refer to the definition of mAP following in chapter 3.2

Line 100: explain "Pytorch", or do you mean Python?

Line 133: should read "described"

Line 143: what do you mean by "quartzity"?

Table 1:  in the wood industry the terrms "pith" (instead of Marrow) and "resin pocket" (instead of resin) are common. Show at least one typical image for each type of defect! Why did you not include images without defects? In Ref.[28] 10 types of defects are mentioned - why did you choose only 7?

Line 365 and Table 2: to be honest - the improvements in mAP fpr the BPN-YOLO method are only marginal.

Figure 6: inserted texts in the images are too small and therefore unreadable.

Line 440: Can you give a typical value  for the computing time per image?

Figure 7: The original image shows a good example for local surface roughness which would also count as a defect in practice. Apparently the picture was taken under oblique light, which explains the local fluctuations in brightness. In any case, they are not typical of wood defects.

Line 488: "ming et al." is a reference?

 

Author Response

Dear Editors and Reviewers:

Thanks for your letter and for reviewer’s comments concern our manuscript entitled “BPN-YOLO：A novel method for wood defect detection based on YOLOv7” (forests-3046076). Those comments are valuable and helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied all comments carefully and have made correction which we hope meet with approval. Revised portion are marked with underline in the paper.

Thank you again for your comments.

Best regards.

Yours sincerely,

Rijun WANG.

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

Title: BPN-YOLO：A novel method for wood defect detection based on YOLOv7

Author: Wang et al.

 

General

This article describes various ANN methods for identifying defects in wood. The article is nicely and clearly written and has potential value.

 

Specific

Use the passive instead of the active.

Figure 6: Enlarge the letters that describe the nature of the defects. Describe the meaning of the numbers in the figures.

Add the speeds of the methods used and the time for method training.

L141 – What type of wood (figures) was used for the training? Figure 6 shows the figures for softwood. What about hardwood?

Author Response

Dear Editors and Reviewers:

Thanks for your letter and for reviewer’s comments concern our manuscript entitled “BPN-YOLO：A novel method for wood defect detection based on YOLOv7” (forests-3046076). Those comments are valuable and helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied all comments carefully and have made correction which we hope meet with approval. Revised portion are marked with underline in the paper.

Thank you again for your comments.

Best regards.

Yours sincerely,

Rijun WANG.

Author Response File: Author Response.pdf

Reviewer 3 Report (Previous Reviewer 3)

Comments and Suggestions for Authors

Repeated reviewer's comments on the paper titled “BPN-YOLO： A novel method for wood defect detection based on YOLOv7”

The paper is significantly improved after the first review. The major reviewers comments were addressed. But at the same moment one more comment should be addressed: 

All the primary contributions outlined at the end of the Chapter 1. “Introduction” should be answered at the Chapter 5. “Conclusions”.

 

Author Response

Dear Editors and Reviewers:

Thanks for your letter and for reviewer’s comments concern our manuscript entitled “BPN-YOLO：A novel method for wood defect detection based on YOLOv7” (forests-3046076). Those comments are valuable and helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied all comments carefully and have made correction which we hope meet with approval. Revised portion are marked with underline in the paper.

Thank you again for your comments.

Best regards.

Yours sincerely,

Rijun WANG.

Author Response File: Author Response.pdf

Reviewer 4 Report (New Reviewer)

Comments and Suggestions for Authors

Detection of wood defects is a key, but insufficiently studied link in the processing and production of wood, on which the quality and reliability of wood products depends. To improve the accuracy of wood defect detection, the authors of the manuscript propose a new wood defect detection method called BPN-YOLO. For example, to alleviate the problems of large memory, high computational costs, and poor detection accuracy of small targets, researchers have proposed the use of sparse queries. The BPN-YOLO model is capable of more accurately and quickly identifying and localizing wood defects compared to other models.

The manuscript is easy to read, and the new theoretical results and their analysis are of interest to many potential readers.

 The research topic corresponds to the Forests.

The research methods used correspond to the topic of the manuscript.

The manuscript contains new results related to modern technologies for identifying wood defects.

The findings are consistent with the evidence and arguments presented.

 No additional controls are required for this study.

 Таbles 1-5. Suitable.

 Figures 1-7. Suitable, but see Remark 3.

 Remarks:

1.  Lines 225, 514. …mAP?

2. Lines 487-488. Please, check “... which is more widely applied than ming et al.”

3.      3. Figure 1, 2, 3. The logic of the flowcharts is good. However, black characters on a dark (such as brown) background make it difficult to read. The background should be lighter.

Author Response

Dear Editors and Reviewers:

Thanks for your letter and for reviewer’s comments concern our manuscript entitled “BPN-YOLO：A novel method for wood defect detection based on YOLOv7” (forests-3046076). Those comments are valuable and helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied all comments carefully and have made correction which we hope meet with approval. Revised portion are marked with underline in the paper.

Thank you again for your comments.

Best regards.

Yours sincerely,

Rijun WANG.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report (New Reviewer)

Comments and Suggestions for Authors

Gentle advice: Please reduce the number of self-citations

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. Please check sentences 174. typing error. 

2. Table 2. please explain how author obtains percentage of images with the defects in the datasets. 

3. Please mention wood species used in this experiment, age?. Size of the wood specimens? density of the specimens. Minimum thickness and maximum thickness can used for the detection? can be applied to all wood species? 

4. Sentence 351. The mAP values is improved by 2.8%, 2,8% and 4.7%. is this improvements values are significant?

5. If possible please include the figures of the real equipments used for each instruments used in this study. 

Author Response

Please see the attachment. Thanks.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

your article is not a scientific work, rather a technical note, I am not sure, if this matter belongs to this journal. For example, it is just about identification of the object, absolutely far from any forestry and applied in the end to timber. Although it looks like science, it is just technical development, I do not see anything scientific in the meaning of new knowledge that the paper shows. No relation to timber grading is presented, so basically it should be in some journal connected to CNN. Publications are mainly some open access stuff, not many serious journals.

Comments on the Quality of English Language

It is probably improved already using some of the AI tools, I do not see many problems.

Author Response

Please see the attachment. Thanks.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Reviewer's comments on the paper titled “BPN-YOLO： A novel method for wood defect detection based on YOLOv7”

The current paper is an original study, which presents detailed and fresh information regarding using a novel method for wood defect detection basing on the developed new method named BFN-YOLO. The method  replaces the ordinary convolution of the ELAN module of the YOLOv7 backbone network with the partial convolution, and proposes the P-ELAN module to improve the performance of wood  defects detection, such as knots and cracks. Proposed BPN-YOLO model was evaluated using an optimized wood surface defects  dataset by the experiments. The results demonstrate the effectiveness of proposed method, showing 7.4% improvement in mean average precision (mAP) compared to the original algorithm. It was stated, that the proposed model can satisfy the need for accurate detection  of wood surface defect. The following comments should be addressed to made more clear content of the paper: 

 

1) List of keywords should be  corrected. Combinations of words defect detection,

wood defects and YOLOv7 should be replaced by the others so as just are involved

in the title of the paper.

 

2) The suggested a single-stage target detection algorithm  should be explained in more details in the chapter 2. “Methodology”.

 

3) Chapter 3 "Experiment and Results" should be supplied by the more detailed explanations of the experiment was carried out. The photos, illustrating the  experiment realization, should be added.

 

4) Descriptions of the experiments, carried out to compare the different

proposed network models,  should be provided in more details. So, sub-chapter 3.4.

"Comparison with Other  Methods and Experiments"  should be completed by the

dimensions and spices of the  wooden specimens, shown  on the Figure 6. Time in

seconds, which is necessary to treat the considered specimens also should be mentioned.

 

5) It should be also mentioned, does any limitations relating to the dimensions of the

specimens and defects exist. If yes, its should be explained in details.

 

6) It should be explained in more details in the chapter 4."Conclusions", what was developed in the current study - method, model or algorithm.

 

 

Author Response

Please see the attachment. Thanks.

Author Response File: Author Response.pdf