A Study on DNN-Based Practical Model for Predicting Spot Color

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors focused on reproducing specific spot colors in the package printing field. The use of standardized spot colors such as the Pantone Matching System is important for the industry. Nevertheless, the existing popular CMYK-based conventional printing methods are expensive, as the repetitive process of manual color mixing and test printing by colorists is needed. To address this issue, the authors proposed a deep learning-based prediction model that minimizes the perceived color difference. The authors employed the Deep Neural Networ, architecture to effectively process structured data for model design. Its contribution is predict the color of spot ink without actual mixing.

 

The authors have cited mainly the previous works of mathamatical approach for addressing this issue. The authors need to talk about the deep learning applications in the printing field, more specifically, about the current state of arts for the prediction of deep learning methods in the spot color applications.  

 

The empirical data of the paper is from the colour-matching 285 techniques of a printing company. The Table 7 showed the illustration of the prediction result comparison on random samples for a few colors.  The experimental results are promising, but may not be convincing enough. The dataset is from a one printing company. Will the proposed method like to work in the working environments in other companies? How good is the obtained result when compared with other methods? More comprehensive evaluations are needed for journal publication.

Comments on the Quality of English Language

Quite good, but the authors should also pay attention to regular spacing between the words like:

"of Industrial Technology(KITECH)"

 

 

Author Response

Thank you for your detailed review of this manuscript. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The objective of this research is to forecast the resulting color of spot ink without the need for physical mixing. Through the implementation of a deep learning model, with CIEDE2000 serving as the loss function, authors have developed a predictive system that aims to minimize the perceived color discrepancy. Authors anticipate that the implementation of their practical model will decrease the variability in color mixing outcomes attributed to individual workers, ultimately leading to improved efficiency in the packaging printing process.

This is an interesting work; however, I have some comments:

- Abstract needs to be improved to understand the summary of the work done highlighting the principal contributions.

- Research Gaps identified from the Literature is to be written clearly. Please include more recent references in the introduction.

- Results are clearly presented and described. I commend the authors for their nice work in making their results clear. However, the obtained results should be compared with some previous works. 

- line 273: Authors claim that certain predictions displayed noticeable inconsistencies, which might be attributed to the model's training process that primarily depends on ink mixing ratios rather than specific color details. Consequently, if some data patterns are not sufficiently captured in comparison to others, the model may struggle to learn and account for them effectively. How authors can resolve this problem ?

- the calculation of  ∆E (A10) should be more explained. 

- The Conclusion should be added by integrating the limitations and the perspective. 

Concluding, the paper has potential to be appreciated by the readers and the above comment are formulated such that to enhance its impact.

 

Comments on the Quality of English Language  

English is generally good, but needs to be polished further. The manuscript should be formatted better and some spelling and grammar should be checked carefully.

Author Response

Thank you for your detailed review of this manuscript. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper presents a deep learning-based prediction model to minimize the perceived color difference.

 

Some parts of the paper need improvement.

1. The representation of color is related to the material and inherent color of an object. How is this reflected in the paper?

2. The paper uses a single dataset, so it can't verify the model's generalization capability.

3. Given the small size of the dataset used, how can it be proven that the model hasn't overfitted?

Author Response

Thank you for your detailed review of this manuscript. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

In the abstract, CMYK is mentioned. Though it is a well-known model, I think it needs the complete form here as well as in the body of the paper where it is first mentioned (it is present in line 30).

In line 24, PMS is first introduced. It should have been referenced. So, please cite the paper which introduced PMS or which has a detailed description of PMS.

In section 2.2, the authors mentioned that "....entries with reflectance values below 0 or above 100, as well as those with zero basic ink usage, were removed....". While it is understandable that it has been done to obtain reliable results, it needs a bit of clarification. A one or two-sentence clarification is enough here.

I am not sure whether Table 5 was referenced inside the text.

Isn't there any state-of-the-art technique to compare the results? If yes, please provide a comparison with some methods.

Author Response

Thank you for your detailed review of this manuscript. Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The revised version answers my question.