Research on the Intelligent Modeling Design of a Truck Front Face Driven by User Imagery
Round 1
Reviewer 1 Report
Authors proposed a "Research on intelligent modelling design of truck front face driven by user imagery". The topic is very interesting, but I do have some comments:
1- What is the computational complexity of the proposed method?
2- Authors should provide a flowchart or an algorithm for the proposed method.
3- The paper lacks the mathematical analysis. Not a single equation was provided to strengthen the paper.
4- Authors should elaborate more on the methods they used to pre-process the images. For example, in Fig. 4, authors should explain the method used for removal of disturbing factors and examine other well-known methods in literature.
The English quality is fair enough.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this study, the authors employ deep learning techniques to generate images of truck front faces and subsequently utilize EEG experiments to assess the alignment between the generated designs and the expected images. Overall, the manuscript provides comprehensive details, and the proposed technique holds promising applications. Nevertheless, I have two questions for the authors:
- Firstly, it is crucial for the authors to clarify whether English is the native language of the subjects (graduate students) invited to participate in the experiments. Since the semantic space of emotional imagery is constructed using English words, there is an inevitable bias if English is not the participants' native language. This bias could lead to subjects focusing more on simpler words they are familiar with, affecting the validity of the results.
- Secondly, I’d like to know the rationale behind selecting truck front face images as the case study. It might not be a familiar design task for most readers of this work. The authors should explicitly outline the advantages of using truck front face images as an example and demonstrate how this similar technique can be applied in other contexts or industries.
- Lastly, I noticed that the authors use abbreviations without providing definitions upon first use. For instance, "EEG" is not defined in the abstract. It is essential to define any abbreviations the first time they appear in the text to ensure clarity and understanding for the readers.
English language is fine, but I encourage further polishing.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper proposes a novel method for fast generation of truck front face imagery modelling solutions. It also conducts quantitative experiment of user emotional image matching based on EEG technology.
I feel that there are three topics mixed in this paper. I would rather like to see one unified topic and contribution. Nonetheless, the paper provides the good contribution in terms of truck face modelling.
Fine
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The topic of the article is interesting but it is not properly presented. The article does not present a strong structure and important choices in the methodology are not well explained or supported by references. The method does not seem reproducible and can be improved by focusing more on the link between design/semantic/emotions instead of the image generation.
Major comments:
-Introduction
No linear explanation. Try to link and motivate the use of the quotation.
-Methods
Create a schematic for explain the methodology
Unclarity, non liner explanation, multiple interpretation (e.g. line 214 "generate design solutions that match the imagery", which design solutions?, which imagery?)
Why using a human classification?
The method should not report results and data calculations.
The selection of the words is not reproducible using a focus group. Why not using automatic tools to select the most representative words (i.e. SEO)?
Heatmaps instead of tables
How do also the colors influence the emotional imagery?
-Results
There is a mix of methods, experiment explanations and results.
-Discussion
This part should elaborate and comment the results obtained and compare them to what was expected and other works.
-Minor improvements:
line 158: explain GAN abbreviation
line 173: Self-made? Later is described as a manual collection of images
line 208: reference to methods
line 263: which factors?
line 268: the method should not report the software language
line 283: the method should report the testing experience, but only the innovative proposal
line 291: the how is not explained. is it a manual operation?
line 314: 1-2 is not clear (same for line 329).
line 438: age range of the group?
line 542: why reporting the DCGAN method if it is not suitable for this case?
line 567: those are not methods but functions. Use references and explain the theory behind
line 614: how the artifacts are detected?
line 774: repetition
line 805: What are the emotional needs of the users? How this method enhances market competitiveness?
line 829: "home-made" is not appropriate
Figure 18 is cropped on the left
The quality of the grammar must be improved, there are online services which can help with reviewing the language of the document.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report
A very good piece of work indeed.
To enhance the work with more details about the methodologies used by the authors, I would suggest the following:
- adding some more background knowledge about the Kansei Engg. in the Introduction section. I am sure many readers would understand the subject of the paper better when knowing more about this science.
- giving an algorithmic description of the research methods or the computer flow charts of the programs used in the work. I am sure this will be appreciated by the computer researchers who would like to use your results to build similar or modified designs like yours.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 6 Report
In this study, the semantics of 2D images, based on the P300 wawe increase of spike amplitude, was objectively proved. However, only the results of 2D images are examined in this research, which is a matrix of planar images. In reality, one only encounters a matrix of 3D images, and each object in space looks much more diverse, with strokes of falling light and shadows. It can be seen that only when colors appeared in the pictures, the amplitude of P300 increased significantly.
Maybe this study should focus on color (Color pattern dataset (Figure 11)), with more diverse color gamuts and composite components of the front view, without both monochrome (Grayscale pattern dataset (Figure 11)) and linear (Line-draft pattern dataset (Figure 11)) models as insignificant?
Are researches planning to examine color 3D modeling in the future? I wonder how the results of the research will change then.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors addressed most of my comments from the previous revision. However, in my previous review report, I asked the authors to discuss the computational complexity of the proposed approach. The authors discussed the computational complexity in their reply but did not include in the revised manuscript. I suggest to shorten the response and include it in the revised manuscript.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Only the first two comments have been implemented and I recognize the optimal improvement.
For the rest, the paper still presents a weak structure with part of the results in the methodology (which should be a reproducible process) and methods in the results which are not explained before-hands.
Please review the full document with the suggested improvements.
English improved but can be better refined.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
