OSO-YOLOv5: Automatic Extraction Method of Store Signboards in Street View Images Based on Multi-Dimensional Analysis

Round 1

Reviewer 1 Report

Exciting research that provides a new approach for retrieving store signboards. Results showed that by using OSO-YOLOv5 network data acquisition is more reliable 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

 

The manuscript presents an innovative approach of information extractions using the morphological structure of store signboards. While the method appears sound and innovative to me, I do not understand the general purpose of this study. In how far does this form of information extraction contribute to geoiformation sciences or any related field, such as urban plannning? Moreover, the results and discussion sections do not refer to any previous study. In this way, the outcome of this study remains without any connection to the (international) state-of-the-art debate. Therefore, I do not see an extension of research by the current version of this manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this article, the authors propose the OSO-YOLOv5 network, which integrates location attention and topology reconstruction under the constraint of rectangular features. They added the coordinate information before the convolutional layer in the backbone structure. Also, they improved the C3 module in the backbone structure. Finally, they proposed an improved spatial pyramid pooling model to reduce the computational load and improve the nonlinear learning ability of the model.

The topic of the work is interesting and creates new directions of research. The submission is well-organized and has the proper structure.

The authors give enough details regarding the new approach. The research is well designed, and a clear objective is set. The review of the state-of-the-art is sufficient. The number of references is satisfactory and up-to-date. Also, the contribution of the paper is highlighted, as well.

I think this article has good potential, but before being considered ready for publication, some aspects need to be clarified and improved.

1)In the Comparison Method Selection section, methods are noted. In order to substantiate the reliability and efficiency of your proposed methods, you should compare them with published work related to the subject under consideration. Specifically, you should provide a comparative analysis with previous studies based on features, datasets, or algorithms.

2)Please describe the limitations and the potential issues of this study. 

3)The quality of the English language is not at a satisfactory level. Extensive scrutiny is required as there are many issues.

4)Equation (1) is out of margin

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

I truly enjoyed reading your paper. This study developed an improved methodology to extract information from store sign boards in street view images. The results clearly show significant model performance improvement (precision rate increase by at least 5%). In overall, I think the study sufficiently shows a good quality. I have some suggestions. 

1. 3.3.1 Dataset Production: How did you collect street view dataset? Please provide a detailed information of CTW and Baidu. Did you use some already established datasets or publicly accessible APIs? 

2. 3.1.4 Comparison Method Selection: You chose 5 models against yours. Why did you choose the five models?

3. Figure 13. I don’t think the trend graph is not good for comparing values across models. Consider  some other graph types. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors re-submitted their manuscript. However, I still cannot see a deep progress in the new version regarding the discussion and results sections. The discussion and conclusion section are hardly connected to any state-of-the-art findings. Results without a clear connection to previous studies remain solitary works and do not justify a publication in an established and known international journal. Therefore, I suggest a rejection of the manuscript.