Leveraging Deep Convolutional Neural Network for Point Symbol Recognition in Scanned Topographic Maps
Round 1
Reviewer 1 Report
Dear Authors,
Congratulations, it is a very interesting and an esceptionally written article. Your research design is easy to follow and your results are convincing.
A few minor comments (I think they can be corrected during proofreading if the other reviewerts do not ask for more revision):
Line 47: STMs
Line 61: And the anchors are not even the same on an international level.
Line 88: they instead of the (?)
Line 94: to instead of with
Table 1: The meaning of symbols should be also in the table - some of them may be not evident for international readers.
Figure 9, 10: Font size in the figure is bigger than the font size of the paper, it is strange.
Figure 23: The resolution is quite bad.
Thank you again!
Author Response
Dear anonymous reviewer,
Thank you very much for your valuable comments on our manuscript. Based on constructive suggestions, we made a thorough revision of the manuscript (indicated in blue), mainly to the existing formal problems. Finally, we believe that your comments will help improve the paper for potential future readers.
Thank you again!
Author Response File: Author Response.pdf
Reviewer 2 Report
Dear authors,
I found this study about the symbol recognition on topographic maps very interesting. I practically haven't any objections about the design of the study.
But only a few comments:
- I wonder how this method will work on different map series with completely different map key?
- What was the age of the map sheets? When we are working with old military mappings from 18th and 19th century in the Central Europe, it is very often complicated to read even for humans.
From the formal point of view:
- It is little bit strange to use acronym of figure caption in text e.g. row 29 - "Fig. 1" and full caption below figure e.g. "Figure 1.".
- Why the order of figures 9 and 8 is swapped in the text?
Author Response
Dear anonymous reviewer,
Thank you very much for your valuable comments on our manuscript. Based on constructive suggestions, we made a thorough revision of the manuscript (indicated in blue). We believe your comments will help improve the paper for potential future readers. In response to your question, I have indicated the source of the map used in the article on lines 104-107 of the article. The model is able to identify topographic maps of the same type; For different types of topographic maps, the model in this article can retain the main part of the network for transfer learning. Knowledge distillation can also be used to train subnetworks to obtain models on new data, so as to achieve better results.
Thank you again!
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors describe a DCNN method for finding point symbols in scanned topographic maps in this paper. The paper is well-written and scientifically sound and it can be published in its present form. It is important to state whether maps originate from a China Mapping Agency and whether the method can be applied to other topographic maps as well.
Author Response
Dear anonymous reviewer,
Thank you very much for your valuable comments on our manuscript. Based on constructive suggestions, we made a thorough revision of the manuscript (indicated in blue). We believe your comments will help improve the paper for potential future readers. In response to your question, I have indicated the source of the map used in the article on lines 104-107 of the article. The model is able to identify topographic maps of the same type; For different types of topographic maps, the model in this article can retain the main part of the network for transfer learning. Knowledge distillation can also be used to train subnetworks to obtain models on new data, so as to achieve better results.
Thank you again!
Author Response File: Author Response.pdf