Meta-YOLOv8: Meta-Learning-Enhanced YOLOv8 for Precise Traffic Light Color Detection in ADAS

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Corrections are given in the PDF document.

Comments for author File: Comments.pdf

Author Response

Dear editor and Dear reviewer,

Thank you for the opportunity to revise our manuscript.

Below, we have addressed each comment and outlined the corresponding changes made in the revised manuscript

Best regards

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript scope is dealing with Meta-YOLOv8: Meta-Learning-Enhanced YOLOv8 for Precise Traffic Light Color Detection in ADAS. However, some comments may improve the quality of the manuscript such as the following:

1. In the Introduction section, the authors mainly list the content of existing studies and lack a comprehensive summary of their limitations. Please revise it to highlight the contribution of this manuscript.

2. The paper involves the derivation of the Yolo algorithm and the application of autonomous vehicles (AVs), and some important relevant references that are missing. See for example the following works: Phase diagram in multi-phase heterogeneous traffic flow model integrating the perceptual range difference under human-driven and connected vehicles environment. Chaos, Solitons & Fractals, 20224, 182: 114791; On detection method of foreign object intrusion into subway track through in-vehicle monocular camera. Journal of Shenzhen Institute of Information Technology, 2021, 19(5): 71-76.

3. What makes the proposed method suitable for this unique task? What new development to the proposed method have the authors added (compared to the existing approaches)? These points should be clarified and explained.

4. Abbreviations and simple sections should be added.

5. A few typographical and grammatical errors exist and the authors should check carefully and correct them.

 

 

Author Response

Dear editor and Dear reviewer,

Thank you for the opportunity to revise our manuscript.

Below, we have addressed each comment and outlined the corresponding changes made in the revised manuscript

Best regards

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper used Meta-YOLOv8 for traffic light detection in advanced driver assistance systems. A comparative assessment proved Meta-YOLOv8 is better than the traditional models. However, the authors should address the following comments. 

1. This paper focuses on the preparation and processing of data sets. It seems lack of innovations. Authors are recommended to analyze the principle of Meta-YOLOV8 and discuss why it can achieve better performances in traffic light detection.

 2. There are only 315 images in the data set. And only 20 images are allocated for adaptability training and testing. It is better to add more images to the data set. Furthermore, although the data set contains some images concerning the bad weather conditions such as fog and rain, authors are expected to consider the other complex traffic scenarios in the case of faint light or glare light.

Author Response

Dear editor and Dear reviewer,

Thank you for the opportunity to revise our manuscript.

Below, we have addressed each comment and outlined the corresponding changes made in the revised manuscript

Best regards

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This paper introduces Meta-YOLOv8 traffic light recognition model to improve the performance of ADAS. The paper's workload is substantial, and the experimental data is sufficient. However, the overall description is somewhat verbose. It is recommended to revise and then accept the paper. Here are the detailed comments:

 

1. There are repetitions in the problem description and YoloV8 improvements in the Introduction section. The redundant parts should be removed.

2. The authors use a deep learning model to recognize tiny-objects, but there are no models specifically targeting tiny-object recognition mentioned in the Related Work section, such as Tiny-Yolo. It is suggested to add such references.

3. The text inside the block diagrams in Figure 2 and Figure 3 is too small. The font size should be increased.

4. The technical indicators such as mAP, mIoU, and F1 score in Section 4.4 are common knowledge and do not need to be elaborately introduced. Please shorten this part.

5. In Figure 14, only the trend of Epochs is shown. The actual iteration numbers should be indicated. Additionally, the text on the other two axes and the subplot titles is too small, which will affect the readability of the printed version.

6. More result images should be provided for the demo experiments in Figures 16, 17, and 18, with at least four images per row. A single image is not sufficiently convincing.

Author Response

Dear editor and Dear reviewer,

Thank you for the opportunity to revise our manuscript.

Below, we have addressed each comment and outlined the corresponding changes made in the revised manuscript

Best regards

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

In Section 4.3.1,authors introduced a labeling method. It is recommended to describe the method more clearly instead of stating the performances from Line 422 to 430. Although the method brought some good performances, it is only a small improvement. Authors are recommended not to use the word “novel”.

In Section 5, the processing examples in fog and rain were presented. The performances in low light traffic scenarios are not obvious since Figs.14(a) and (b) are very clear. Authors are expected to consider the worse weather conditions.

Author Response

Dear editor and Dear reviewer,

Thank you for the opportunity to revise our manuscript.

Below, we have addressed each comment and outlined the corresponding changes made in the revised manuscript

Best regards

Author Response File: Author Response.docx