Review Reports

Round 1

Reviewer 1 Report

 

Review of the manuscript (paper) titled Generation Method for Shaded Relief based on Conditional Generative Adversarial Nets

 

In this paper a new method for generating relief shading based on cGAN and manual relief shading – DEM slice pairs was proposed. It was applicationed a deep learning technology to propose a generation method for shaded relief based on conditional generative adversarial nets. The paper contains the compares the network shading with analytical relief shading, manual relief shading, and relief shading generated by other networks. Test results indicated that the network successfully simulated the manual shading style of Swiss cartographers, greatly enhancing the artistic effect of the shaded relief. Therefore, the paper is a valuable contribution to the consideration cGAN from within the field of deep learning, taking manual shaded relief and the corresponding DEM as the dataset to propose a shaded relief generation method based on cGAN. The paper is well written.

 

Below are some suggestions for improving your article.

•   All references listed in References were cited in the text. However, listing references 5, 6 and 33 would be good to equalize (arXiv 2014. arXiv preprint arXiv:1406.2661; arXiv, 2014, arXiv:1411.1784; Available online: arXiv:1412.6980 [cs] 2017.).

•   Most of the cited papers are available online, so it would be good to list DOI.

•    In line 554 should be written Imhof, E. Cartographic Relief Presentation; ESRI Press: Redlands, 2007.

•   In line 268 should be written Patterson, T. Designing the Equal Earth Physical Map. Available online: https://youtu.be/UYQ6vhxc9Dw (accessed on 8 April 2022).

·       In line 582 and 583 should be written April.

·      Check that it can be written Fig. instead of Figure below the figures and in the text.

 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

In order to improve the quality of shaded relief, cGAN from the field of deep learning was introduced, taking manual shaded relief and corresponding DEM as the dataset to propose a method of generating shaded relief based on cGAN.

The test results showed that the network successfully simulated the style of manual shading by Swiss cartographers, greatly enhancing the artistic effect of the shaded relief. The trained network highlighted the main features of the terrain, while reducing the broken details of the terrain, achieving a relatively good generalization of the terrain. At the same time, the network was able to simulate illumination adjustments completed according to local terrain features in manual shading, thus enhancing the expressiveness of the terrain.

There are also certain limitations that the proposed method has, but overall the work can allow cartographers to more easily obtain high-quality shaded relief to make even more beautiful maps.

Author Response

We are grateful for your comments and suggestion.

Reviewer 3 Report

Dear authors,

This is an interesting piece of research dealing with deep learning of hill shading based on manually shaded samples. In general, the paper is well structured and illustrated both regarding methods and results, so that I do not have any major objections.

My only recommendation would by to introduce manual shading in more detail, since many (especially younger) readers may not be familiar with this procedure.  You frequently mention that manual shading better adjusts to the local terrain, e.g. “making the main terrain more obvious while also allowing some smaller but important terrains to be identified”. This description may appear abstract to the average GIS-cartographer who might be familiar with analytical hill shading and unable to understand its limitation. An additional figure to illustrate these characteristics or at least some reference to one of the figures used later in the text might be helpful.

On several occasions, you refer to a lack of expressiveness of non-manually shaded relieve. This sound to be a rather subjective assertion to me! What exactly to you mean with “expressiveness”. Are there any empirical studies supporting your statement? If not, you are probably speculating.

A final consideration: Each analytical is comparable since it results from comparable rules. However, each human hill shading expert applies an induvial technique with personal criteria. Can this detail affect your method? Will a network trained in a “Swiss hill-shading slang” equally understand an e.g.  “US hill-shading slang”? Did I make myself clear?

Author Response

Please see the attachment.

Author Response File: Author Response.docx