Reproduce “Computer vision-based analysis of buildings and built environments: A systematic review of current approaches” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Małgorzata B. Starzyńska, Robin Roussel, Sam Jacoby, Ali Asadipour
A systematic review by Starzyńska et al. analyzed 88 papers on computer vision applications in architectural and urban analysis, classifying research areas and evaluating methodological trends, data sources, and significant issues in reproducibility and comparability. The work provides a foundational assessment for future transdisciplinary research between computer science and architectural studies.
Key insights
- Computer vision applications in architectural analysis largely fall into four categories: image classification of building style, building detail detection, qualitative environmental analysis, and building condition/value estimation.
- Deep Convolutional Neural Networks (CNNs) are the predominant computational approach across most architectural applications and scales, broadening the scope of analysis for large, unstructured datasets.
- Current research is hampered by inadequate data curation (e.g., lack of architecturally-specific and context-rich datasets, geographic bias) and low reproducibility due to a lack of shared code, models, and custom datasets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.