Zero-shot building footprint extraction and regularization based on Segment Anything model with Mesh Model

View Full Paper

Ask AI

Bookmark

The method achieves 78.31% AP50 on the Vaihingen dataset, demonstrating effectiveness in extracting building footprints.
It utilizes a three-step process involving coarse prompt generation, fine mask generation, and footprint regularization.
Kinetic Partition and Markov random field are key components for effective contour regularization of building footprints.
The approach addresses the limitations of traditional manual methods and deep learning techniques requiring labeled data.

Authors

Discussion

No takes yet. Share an insight, caveat, or question.

This framework improves building footprint extraction accuracy in urban scenarios, suggesting enhanced methods for urban planning.

Zhong et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark