VLM-BCD: Unsupervised Building Change Detection

Yiyun Zhang, Zijian Wang

Dec 2023 · ACM International Conference on Multimedia in Asia

Abstract

Building Change Detection (BCD) is one of the most important parts of remote sensing analysis. However, most of the existing BCD approaches require a large amount of pixel-level annotation, which limits their applicability due to intensive labour costs. To alleviate this issue, we propose a vision-language model-based framework, VLM-BCD, which performs BCD tasks without requiring any labels. Specifically, the proposed framework consists of two stages: (1) Bi-temporal building localisation by leveraging open-vocabulary DETR. (2) Unchanged mask suppressing by the Change Resolver module to detect the building change in bi-temporal satellite images. An application with an interactive dashboard is implemented to maximise the usability of the developed framework.

BibTeX

@inproceedings{zhang2024vlmbcd,
  author = {Zhang, Yiyun and Wang, Zijian},
  title = {VLM-BCD: Unsupervised Building Change Detection},
  year = {2024},
  url = {https://doi.org/10.1145/3595916.3626357},
  doi = {10.1145/3595916.3626357},
  booktitle = {Proceedings of the 5th ACM International Conference on Multimedia in Asia},
  articleno = {109},
  numpages = {3},
}