Smart City Community Watch—Camera-Based Community Watch for Traffic and Illegal Dumping

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper focuses on effectively identifying illegal dumping using video surveillance in smart cities. Here are some comments and suggestions for improvements before resubmission:

· Incomplete parameter information in  “Table 1. Literature Survey of illegal dumping detection methods”.

· Captions should be rephrased to highlight what is discussed in the Figures/Tables.

· Dataset description is not clear in “Table 4. List of datasets”.

· Terms should be used consistently throughout the document. For instance, in Section 3.2, the term “images” has been replaced with “photographs”.

· References are not properly utilized. Please ensure that all references are correctly cited and integrated into the text as required.

· “YOLOv3” is mentioned on “page no. 8”, but its relevance is unclear since YOLOv5 is used in the proposed scheme. Please ensure that the references to object detection models are consistent with the actual methods used in the research project.

· The steps mentioned in “Figure 4. Video Pre-processing Steps”, should be discussed in detail to provide a clearer understanding of the process.

· Content from “Section 3.2” is repeated in “Section 3.3”. Please review and remove the redundant information to ensure clarity and conciseness.

· What was the rationale behind choosing a Euclidean distance of 75+? Have you experimented with other distance values?

· There is no discussion on the model evaluation metrics. Please include a detailed evaluation of the model, including metrics, and performance analysis.

· What is the original model? Is it a model proposed by someone else? If so, please provide a reference.

· The results are insufficient to effectively evaluate the proposed technique. To provide a comprehensive evaluation, additional experiments or analyses should be conducted.

 

· The comparison with previous studies lacks depth. In what ways does your technique differ from others?

Author Response

Please find attached document for the responses to the comments. 

Thank you.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The study on detecting illegal dumping of garbage through cognitive AI of image recognition is very interesting. The authors should have given examples of the shortcomings of the relevant research techniques before introducing their research in the first paragraph. The reviewer has previously seen smart city tracking techniques that combine garbage truck and dumping data in Brazil. Difficult and challenging points of the prior technology should have been emphasized.

 

Image processing related techniques of OpenCV and YoLo are very popular and used by everyone. However, as far as the reviewer knows, the latest version of YoLo has been upgraded to 10, have the authors considered utilizing the latest version?R-CNN algorithmic models are very common, and DeepSORT is the first time the reviewer has heard of it. Can the authors be more specific about the application of DeepSORT in this study?

 

There are many pre-trained models such as CoCoSSD, what do the authors think about their detection for non-specific targets? The reviewer thinks that Fine-Tuning of Models should be performed for different purposes.The authors collect the dataset in Chapter 3 section and utilize video for data preprocessing and detection, what is the reason for not directly detecting in the form of real-time video transmission? And what are the difficulties and challenges?

 

License plate and car model recognition, as well as person and object recognition are certainly important. However, how do the authors determine the intent of people's behavior? Is a moving time window needed to gradually infer and validate people's behavior? Also, how are cameras set up in display applications? How do the ethical aspects of the data recorded by the camera respond to human security and privacy protection? Litter detection is also part of the author's proposal, in the concept of the smart city domain, can the author visualize on a map which areas are often littered and which areas are relatively tidy?

 

The reviewer is aware that much of the authors' model building is done on the GPU. However, what is the evaluation of the cost of data and time spent on modeling for real-time detection? The authors' Figure 20 expresses clearly that current research progress is stuck in detecting littering behavior in the parsing of specific videos. This in itself is indeed a big step forward. The reviewer believes that there will be some scientific references in the future for detection processing of video streams with big data.

 

That is all. Thanks.

Author Response

Please find attached document for the responses to the comments. 

Thank you.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for your response and efforts. However, there are still some comments and suggestions for improvements that need to be addressed before resubmission:

· Incomplete parameter information in “Table 1. Literature Survey of illegal dumping detection methods” still stands.

·   In image resolution (e.g. 256*256), * should be replaced with x.

·  Figures are not discussed appropriately, e.g. Figure 11. Vehicle License Plate Model Evaluation. Kindly explain the analysis of the graphs or evaluation metrics in detail.

·  References are not properly utilized. Please ensure that all references are correctly cited and integrated into the text as required. For example, I couldn’t find reference 5 cited in the text. Also, Most of the references are links to web pages. They should be substituted with reputable conference papers or journal articles. The sequence of reference should also be considered.

·  The Figures/Tables should come right after the discussion in the paragraphs and they should be close to the content where they are discussed.

 

·   Some Equations are screenshots, they should be written appropriately.

Author Response

Please find attached document for the responses to the comments. 

Thank you.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

In addition to the 97% detection accuracy, what is the direction of future work after detection? What language is used to develop the mobile application? Does it have any limitations? The reviewer suggested that the author must collect real-time data in future work, even if the detection is asynchronous, at least the data is closer to local actual applications. Weather, lighting, and viewing angle are all factors that affect image clarity and directly affect the detection results. How can the author improve these? At the same time, the reviewer is also very concerned about the replicability of the technology. Is this technology used in Asian countries including China? How can the author ensure that each image captured is the best viewing angle for detection? Has the application been launched on the market? What realistic and unavoidable challenges will it face in the future?

 

That's all. Thanks.

Author Response

Please find attached document for the responses to the comments. 

Thank you.

Author Response File: Author Response.docx