PVNet: A Used Vehicle Pedestrian Detection Tracking and Counting Method

Round 1

Reviewer 1 Report

Line 84: Double T has appeared please check and correct it.

The author needs to add more recent references because the related work has little references. The author should add at least 5 more recent references.

Comparison between the authors methodology and the one presented in the literature should be given in form of a Table.

What is the limitation of the work and a future recommendation?

The quality of English is ok except some minor typing errors which i mentioned above for correction.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1.      Can you provide concrete statistics or examples to support the claim of "excellent performance" of the model?

2.      How does the proposed model address the safety concerns related to the increase in the number of cars and autonomous driving technology?

3.      Could you provide more context on the prevalence and impact of road accidents in the context of this discussion?

4.      Are there any ethical considerations or potential privacy issues associated with implementing this pedestrian-vehicle detection technology?

5.      Can you explain in more detail why optimizing the darknet53 backbone and using PVFPN for multi-scale feature fusion were necessary for improving the model?

6.      What kind of dataset was used for training and testing the model, and how representative is it of real-world scenarios?

7.      Were there any limitations or challenges encountered during the development of the model, and if so, what were they?

8.      Could you outline potential future directions or areas of improvement for this pedestrian-vehicle detection model?

 

9.      Can you provide examples of how this technology can be practically implemented in real-world traffic scenarios to enhance road safety?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper proposed a pedestrian and vehicle detection model based on deep learning. By optimizing the darknet53 trunk feature extraction network, PVFPN multi-scale feature fusion method is adopted, and byte tracking method is adopted to count and track targets, so that the whole model is suitable for pedestrian and vehicle detection and tracking in traffic scenarios. This is a meaningful work, but there are some issues need noting.

1. There are writing errors in the text, such as a Chinese period in section 4.2 and some redundant Spaces in the whole paper. It is suggested to check the whole text.

2. At the beginning of the article, the authors said that the darknet53 backbone feature extraction network was adopted in this paper, but there was no description of the relevant content.

3. Why did the authors only consider the comparison with the YOLO series of algorithms? There are many excellent object detection algorithms based on other methods, but the authors did not consider them.

4. In the comparison between your method and the results of YOLOv8s shown in Fig. 9, it is difficult to see any outstanding results of your detection results, so it is recommended to choose results which are easier to observe the difference.

5. At the end of the article, there is no description of the future. I am curious about the application prospect of your method, improvement measures and future work direction, and I suggest you add these contents.

6. Detection is widely used and also has great uses in the field of air pollution control. It is recommended that the authors cite the following articles to expand the description of the application field and make the article structure more complete: “Deep dual-channel neural network for image-based smoke detection”, “Vision-based monitoring of flare soot”.

 

need to be polished.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf