DP-YOLO: Effective Improvement Based on YOLO Detector
Round 1
Reviewer 1 Report
The article presents a new approach to the complex problem of object detection. Because the experimental details and evaluation criteria are clearly stated, I believe this study could serve as a useful reference for other research in the field. However;
* The introduction can be rearranged to make the article more comprehensive. More information can be given about originality-innovation, dataset diversity, computational cost, application areas and cost.
* It is not clear which aspects of existing methods are inadequate and how the proposed method addresses these shortcomings. An evaluation or analysis table comparing the proposed method with existing methods should be created. This could be critical for understanding the effectiveness of the proposed method. The evaluation table could be placed under the related works section.
* The image resolutions are too low to be understood; they should be increased.
* The references are incorrect.
* An assessment should be made as to whether this approach will lead to overfitting.
* More information can be provided about the limitations of the proposed method.
Author Response
See Attachment
Author Response File: Author Response.docx
Reviewer 2 Report
Title: DP-YOLO: Effective improvement based on YOLO detector
Authors: Qijin Wang, Chao Wang, Yu Qian, Yating Hu, Ying Xue, Hongqiang Wang
Improving the detection accuracy of traditional YOLOv5 is a crucial problem especially in the field of urine sediment analysis. The authors propose the more accurate DP-YOLO method verified on both the COCO2017 dataset and the specific Urised11 dataset with a small computational cost. Of particular note, the average precision of DP-YOLO is higher by 3.2 AP in comparison with the average precision of YOLOv5s. In my opinion, this paper can be recommended for publication with a few clarifications.
The following points should be addressed:
1. Why is the model check performed on the COCO2017 dataset and not on the most recent 2020 COCO object detection dataset?
2. The paper considers only the particles of organized urine sediment (Table 3). However, there is still unorganized urine sediment contains crystals of salts. How the proposed method identifies these objects?
3. Page 4, line 168; page 5, lines 223-224, 226; page 7, lines 248-249;page 11, lines 382, 393-394; page 13, lines 424, 432; page 14, lines 447, 455-456, 459, 465: “Error! Reference source not found.”
Author Response
Please see attachment
Author Response File: Author Response.docx
Reviewer 3 Report
This article introduces DP-Yolowhich, an improved real-time object detection model that enhances Yolo v5 by integrating deformable convolutions. It optimizes the positive sample selection during label assignment, resulting in a more scientifically grounded process.It presents an improvement over YoloV5 on the COCO benchmark while maintaining parameters and computational efficiency and demonstrating substantial advancement in object detection performance. However, specific observations must be addressed before further consideration of this article's publication.
.
1. The writing of the paper can be improved.
2. Figure 2 seems a bit distorted; please redraw or resize it for better visibility.
3. The caption of Figure 2 seems to be very lengthy; summarizing it and presenting it in fewer lines is suggested.
4. The caption of Figure 4 must be adjusted on the same page as of the image; please caption it in fewer lines.
5. Table 3 and its caption to be presented on the same page.
6. Heading 4.2 on line 361 should be adjusted on the next page.
7. Error! Reference source not found at lines 382 and 393; reference giving an error. It needs correction wherever referred.
8. Improve the caption at line 398. our proposed should be changed to Our proposed…
9. The overall formatting and figure discussion must be improved.
10. Error! The reference source is not found on line 406, there seems error regarding the reference of the research paper.
11. Impact of PSA on Label Assignment To be adjusted on next Page.
12. Different fonts or sizes at line 438 and 439.
13. Errors at line 455 and 459, which is the serious issue of this paper.
14. Some of the values as listed in Table 8 are in bold letters. If emphasis is required on it, they should be discussed in the text.
15. Discussion in the experiment cannot be understood much due to many reference error of other research papers.
16. For parameter settings, it is best to maintain a table and list all of them in it with values suitable for your research. Also, discussion is required on them.
17. Discuss all the figures well in the text.
18. Discuss the results well. A bit of detail is suggested.
Moderate updates in language are required. Better to go through the paper and remove any typos or grammatical errors before resubmitting.
Author Response
Please see attachment
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors made the changes indicated in the peer review. It can be published as is. However, the resolution of the images is still low. This problem needs to be fixed.
Author Response
Thanks. The images in the PDF converted from the website are low resolution, but clear enough in the manuscript (word). We have corrected it and provide a clearer version as a zip file.
Author Response File: Author Response.docx
Reviewer 3 Report
Author have addressed all the issues, its accepted now from my side.
Author Response
Thanks for your time.