A Machine Vision Development Framework for Product Appearance Quality Inspection
Round 1
Reviewer 1 Report
I would suggest to move Appendix to the end of the paper, after list of references.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
There are some minor corrections to be made in the article.
Please correct the size some of the letters, the number of pictures- including their quality and legibility.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper proposes a tool for automatic defect analysis. While the proposal is interesting, it can be vastly improved. Here are some suggestions.
1. Section 1, pages 1/2. Authors enlist some libraries and tools for image analysis. However, this list is largely incomplete and requires an extension. Please cite some other examples, such as "Machine-learning based vulnerability analysis of existing buildings", Ruggieri et al., 2021., and "Image analysis tools for evaluation of microscopic views of immunohistochemically stained specimen in medical research–a review", Prasad et al., 2012.
2. Section 2.1, page 3. Please add some reference to this generic architecture.
3. Section 2.4 title: small refuse (capital letter).
4. Section 3.1, page 8: probably a typo in row 244.
5. Section 3.4, page 9. Please provide further details on the detection algorithms that can be implemented. As an example, can NN-based algorithms be implemented (e.g., YOLO, R-CNN)? There is a reference in YOLOv4 in the experimental section, but it's not clear if it can be embedded within the proposed interface structure.
6. Section 4, page 13, rows 377-381: this description is too generic. Please provide further details.
7. Section 4, page 13, rows 383: please provide further details on the deep learning model used (e.g., which architecture, which type of training, and which dataset).
8. Section 4, page 14, rows 401-403: please further describe the machine used, i.e., how much RAM, which type of CPU, and if a GPU can be processed to improve speed.
9. Section 4, page 14, rows 405-406, please provide at least one experimental result, as it would be interesting to evaluate numerically the performance of the whole system for reproducibility.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
This paper describes a software development framework for applications that are primarily vision based in industrial control and quality assurance applications. The paper describes in some detail the various modules that such an environment should contain, and gives an example a specific example of that environment. However it is difficult to see how the material in this paper is interesting either to the software developer that is familiar with modern software development technologies, or the quality assurance engineer who is interested in developing an application that solves their immediate needs, or even the image processing expert since the discussion of the relative merits and how to incorporate the various specific algorithms were not part of this study.
The literature review did mention some related commercial offerings that seem to offer solutions very close to what is proposed in this paper. Some of the critique given in the review seemed a little unfair, given that the HALCON does allow one to extend the system end the Cognex does allow generic industrial cameras. In addition, in the literature review some of the references are not perhaps the canonical references that we would expect for some of the statements made.
Overall the paper reads like an instruction manual for the specific application at hand, where considerable length is given to the actual logging in and security for this particular application, and this level of detail is not really very generic nor important for the average reader. Some of the detailing of the actual software design, especially in the tables in the appendix) is also not particularly valuable to the reader at the level of detail that is given.
In summary, this paper would be improved if the authors gave far more advice as to how one can build a suitably flexible environment that tailors for a wide variety of needs and shows examples of such an application as opposed to a single example with quite a specific need. Furthermore the framework needs to be extremely extendable and more attention given to trying the various image processing algorithms and adjusting the workflow. Some of these points are hinted at in the paper, but not really discussed in sufficient detail that makes a valuable contribution.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The authors successfully revised the paper, so it can be accepted in its current form.
