Pixel-Based Approach for Generating Original and Imitating Evolutionary Art

Round 1

Reviewer 1 Report

The article presents an interesting approach of generating artworks using genetic, pixel-based, self-developed algorithms. The text has both scientific soundness as well as an appropriate layout. The text is presented clearly and informative. It could be challenging to find shortcomings here. The only requirement could be that the text is checked again in terms of English language before publishing (some minor errors can be spotted like ex. blind spaces, missing commas, lack of spaces before citation brackets etc.). Congratulations!

Author Response

Thank you very much for your review work.

As you mentioned, we corrected some errors of the text, including the blind spaces, missing commas, missing spaces (24 places revised), and one typo (Revised in section 3.2 paragraph 1).

Thank you again for your patient review.

Reviewer 2 Report

The paper introduces an evolutionary art generation technique. I suggested major revision. You should revise the following points:

1) A more detailed introduction is needed to show what is the advantages and main goals of the research, what are the possible applications of this work. 

2) For clarity, add spaces before and after the equations because a reading the text is a bit messy in this way.

3) You mentioned several existing paper which generate artworks. You should compare the proposed work with some state-of-the-art methods, especially focusing on Generative Adversarial Networks, and deep learning based style transfer techniques.

4) You should set up some metrics for quantitative evaluation of the proposed method and for comparison with other methods.

Author Response

Thank you very much for your review work.

 

1. We clarified the motivation, advantages, and main goals of our proposed method (Revised in section 1 paragraph 3). Several possible applications of our work were pointed out (Revised in section 5 paragraph 1).

2. Spaces have been added before and after the equations as requested. (11 places revised)

3. We added a description about the state-of-the-art methods, including the ArtGAN and the CNN based method. The characteristics, advantages and disadvantages were listed and compared with our proposed method (Revised in section 2 paragraph 4).

4. We agree that metrics are important for evaluating evolutionary methods. In the previous researches, we proposed some improved evolutionary algorithms, and evaluated their statistical data with metrics such as fitness, standard deviation, and convergence. The goal of designing an evolutionary algorithm is clear and easy to describe mathematically.

However, the evaluation of evolutionary art generation method is slightly different. The main goals of evolutionary art generation methods are not to provide more accurate fitness, faster convergence speed, but to provide a better aesthetic feeling or generate more human-like artworks. As we mentioned in section 5 paragraph 1, setting up metrics for quantitative evaluation of evolutionary art is an aesthetic selection problem that has not been completely solved for decades. On the one hand, our approach functionally outperforms other approaches, because our method provides an originally creative, self-organized, fully-automated way to generate evolutionary artworks. On the other hand, we believe that instead of setting some biased metrics, the results of the questionnaire are more convincing. We hope this clarifies.

 

Thank you again for your patient review.

Reviewer 3 Report

In this work, the authors propose a pixel-based method to automatically generate evolutionary arts. The method can generate different artworks, such as original artworks and imitating artworks, with different artistic styles and high visual complexity. The authors provide a very thorough review of the literature in their work introduction. After describing the proposed approach, they provide a list of some evolution results and the results of a user study where they have mixed man-made artworks and their evolutionary ones. Finally, the authors give some conclusions and share some lines of future work.

Although I'm not an expert in this field, I find the work very interesting and, from my point of view, it means a valuable contribution to the community. I only have two concerns:

1. The first concern is about the layout quality of Figure 2. I guess you can enhance its quality without putting much effort.

2. The second concern is about sharing the source code of your evolutionary artwork generator project. I find it would highly increase the strength of your presented document if you include a link to the source code of your project, if possible, maybe to a GitHub repository.

 

Author Response

Thank you very much for your review work.

 

1. The layout of Figure 2 has been rebuilt, the shapes have been colored, and a note of different shapes has been added. (Revised in section 3)

 

2. We agree that the source code sharing is important for the community development. The evolutionary art generation project, which is written with Java as a windows program, had been patented. The patent owner is the first author and the college to which the author belongs. To share the source code, we have to submit an application to the college office. Since we are currently in the middle of summer holiday, the code sharing license will be obtained after October.

 

Thank you again for your patient review.

Round 2

Reviewer 2 Report

Thank You for the revision.

For future work, I suggest setting up some test applications where participants can score the generated artworks and they can compare the proposed method with other algorithms.