Summary of Scalability in Building Component Data Annotation: Enhancing Facade Material Classification with Synthetic Data, by Josie Harrison et al.
Scalability in Building Component Data Annotation: Enhancing Facade Material Classification with Synthetic Data
by Josie Harrison, Alexander Hollberg, Yinan Yu
First submitted to arxiv on: 12 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed method utilizes computer vision models trained on Google Street View images to create material cadastres, which can revolutionize the way we think about material reuse. By fine-tuning a Swin Transformer model on a synthetic dataset generated with DALL-E, this study demonstrates a reasonable alternative to manual annotation, which is time-consuming and often plagued by class imbalance. The results show that the synthetic dataset performance is comparable to that of a manually annotated dataset, paving the way for more efficient development of material cadastres. This research has far-reaching implications for architects, providing valuable insights into opportunities for material reuse, ultimately contributing to the reduction of demolition waste. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Computer vision models can help create maps of materials used in buildings. Right now, these models are only as good as the pictures they’re trained on, which often need human annotation. This takes a lot of time and effort, and it’s hard to get accurate labels for every picture. To solve this problem, researchers fine-tuned a special kind of AI model called a Swin Transformer on fake data created using DALL-E. They compared the performance of this synthetic dataset to one that was manually annotated. While human annotation is still the gold standard, the results show that the synthetic dataset can be a good alternative. This research will make it easier for architects to create maps of materials in buildings, which can help reduce waste and promote sustainability. |
Keywords
» Artificial intelligence » Fine tuning » Transformer
