Scalability in Building Component Data Annotation: Enhancing Facade Material Classification with Synthetic Data

arXiv:2404.08557v1 Announce Type: cross
Abstract: Computer vision models trained on Google Street View images can create material cadastres. However, current approaches need manually annotated datasets that are difficult to obtain and often have class imbalance. To address these challenges, this paper fine-tuned a Swin Transformer model on a synthetic dataset generated with DALL-E and compared the performance to a similar manually annotated dataset. Although manual annotation remains the gold standard, the synthetic dataset performance demonstrates a reasonable alternative. The findings …

abstract annotation arxiv building challenges class classification computer computer vision create cs.cv cs.lg current data data annotation datasets google however images material paper scalability street swin swin transformer synthetic synthetic data transformer transformer model type view vision vision models