CNN-Based Classification of Façade Motifs in Market-Developed Housing: A Computational Approach to Tel Aviv’s 1980s–1990s Urban Fabric
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis article addresses the long-standing "limitations of case studies" in urban architectural history research by employing computer vision techniques to analyze the non-iconic façade features of Tel Aviv's market-oriented apartment buildings. It provides valuable insights into methodological innovation and practical application; however, improvements are still needed in writing content and methodological transparency.
1. The abstract is currently somewhat verbose, especially the introduction on the research background and the necessity of digital humanities, which occupies excessive space. The author is advised to streamline the section by avoiding repetition of time frames and location names, thus improving information density.
2. The introduction contains overly detailed descriptions of method execution, leading to unnecessary repetition. For example, lines 69–75 discuss failed early experiments, dataset reconstruction, and model selection, which would be more appropriate in the "Methods" section. The author should condense these into a brief overview.
3. The subsection "Positioning This Study" in the introduction lacks clear orientation. It is recommended to add a numbered heading such as 1.1 Positioning This Study and to explicitly define the research question at the end of the introduction.
4. The author is advised to include a research roadmap in Section 2.1, as visualization enhances readability. Similarly, in Section 2.2, when referring to "specific façade elements" identified by the research team, the inclusion of illustrations would help readers understand the uniqueness of the research samples.
5. The current definition of features in Section 2.2 is too vague. It is recommended to clarify the criteria for each "semantic theme" (e.g., staggered balconies) with a table, additional descriptors, and an explanation of how experts handled ambiguous cases during annotation.
6. Likewise, in Section 2.3.2, the author needs to explain in detail the “rigorous screening methods applied to the broader corpus.” The selection criteria for images must also be clearly specified.
7. In Section 2.3.3, manual annotation by experts serves as a crucial basis for the study. However, the current description is overly general. The author should elaborate on the specific visual principles followed by experts during annotation and include inter-rater reliability testing, reporting Cohen's Kappa value.
8. The reduction of VTLV's 24,500 images to only 877 represents a very small sample. The author is advised to apply K-fold cross-validation to provide more persuasive average performance metrics and explicitly acknowledge the limitations imposed by the small sample size in identifying specific geometric features.
9. Section 4.5 currently employs too many sub-sections (4.5.1 to 4.5.4), which renders the conclusion fragmented and cumbersome. It is recommended that the author remove tertiary headings and consolidate future research directions into logically coherent paragraphs.
Author Response
Please see attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript is titled: Deep Learning of Urban Design Landscapes: Computer Vision Analysis of Market-Developed Apartment Houses in Tel Aviv, 1980s-1990s.
The manuscript proposes a digital humanities approach that integrates architectural expertise with computer vision for the empirical study of the architectural history of the built fabric of everyday urban housing in the city of Tel Aviv . Focusing on Tel Aviv’s market-based apartment houses of the 1980s-1990s, regarded in the literature as privatized non-architecture, the authors identified several façade motifs that form the architectural vocabulary of extensive development, among them the rise above four floors, staggered balconies, the use of color, and other motifs.
The topic addressed is original and relevant. It fits within the scope of the journal. The manuscript is well-structured. The objectives should be more clearly defined. The methodology used seems appropriate for the work.
The work is an interesting case study, whose main interest should lie in its replicability using other models.
Therefore, I believe the study should focus on the methodology employed and its future replicability.
The article's title should be clarified. It should specify what type of "Computer Vision Analysis of Market-Development" is being conducted so that the title reflects the article's content and does not generate false expectations.
There should be less repetition of concepts in the abstract and introduction.
It would be interesting to delve deeper into the methodological development. To this end, I find the lack of graphs or flowcharts that would aid in understanding the methodology lacking. The manuscript's abstract itself mentions an iterative workflow, the description of which could be accompanied by graphics.
It would be helpful if the authors clearly indicated whether there is previous work done with this methodology in the area of architecture or other areas.
The conclusions should be reformulated, placing greater emphasis on the findings of this study, the replicability of the model in cities, and the possibility of using other motifs. Personally, I believe the conclusions do not live up to the expectations generated by the manuscript's title.
Author Response
Please see attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe author has completed all comments.
Author Response
Dear reviewer,
We thank the reviewer for their positive assessment and recommendation to accept our manuscript. We deeply appreciate the time and effort invested in reviewing our work throughout the review process.
Sincerely,
The Authors
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have improved the article by addressing and incorporating the reviewers' suggestions. I appreciate the effort they have made. Therefore, in my opinion, the article can be published.
As a minor suggestion, I think the flow chart illustrating the phases or processes of the methodology used could be improved. I think it could be made more visual and easier to read, and therefore easier for the reader to understand.
Author Response
Dear reviewer,
We thank the reviewer for their positive assessment and for approving our manuscript for publication. We appreciate the time and effort invested in reviewing our work.
Reviewer 2's Comment:
"As a minor suggestion, I think the flow chart illustrating the phases or processes of the methodology used could be improved. I think it could be made more visual and easier to read, and therefore easier for the reader to understand."
Our Response:
We thank the reviewer for this valuable suggestion. We have revised the methodology flowchart to enhance its visual clarity and readability. The updated figure now features:
- Improved visual hierarchy and layout
- Clearer differentiation between phases and processes
- Enhanced readability through better spacing and organization
The revised flowchart (provided as an SVG file) better facilitates reader understanding of our methodological approach.
We believe these improvements address the reviewer's suggestion and enhance the overall quality of the manuscript.
Sincerely,
The Authors
