Floor_SAM

This paper proposes a LiDAR point cloud indoor floor plan reconstruction method based on multi-dimensional information union, combining the point cloud density map with the zero-shot segmentation capability of SAM to achieve efficient and robust indoor floor plan reconstruction. ！！！Our code will be fully released after the paper is accepted. The following is a partial display of our code effect：

1、Semantic Segmentation of Rooms

https://github.com/user-attachments/assets/aef0819b-abd4-4aa0-bb44-7b7b672e29b0

2、Point cloud boundary point extraction

https://github.com/user-attachments/assets/350de6dd-4b68-49df-95be-e8444dc04e1c

3、Floor plan drawing

https://github.com/user-attachments/assets/15da709d-cac7-442f-9a51-6abef44a147c

4、Our floor plan reconstruction results on Giblayout and ISPRS datasets：

Giblayout：

<img width="905" height="604" alt="experience1" src="https://github.com/user-attachments/assets/4822b72d-cb79-45b6-86fb-b50d3b3c9aa1" />

ISPRS：

<img width="1326" height="465" alt="experience4" src="https://github.com/user-attachments/assets/69458dfe-e9c6-405d-a80b-a4f664b7952f" />