Awesome

RSI-MMSegmentation

Official code for ``GAMUS: A Geometry-aware Multi-modal Semantic Segmentation Benchmark for Remote Sensing Data''

The new version of the GAMUS dataset can be downloaded from the following XShadow/GAMUS.

The old version of data containing DFC 2019: link

Update

Remove the cities (OMA and JAX) from the DFC 2019 dataset to ensure the label quality
Add pytorch dataloader
Simplify the codebase
Re-run all experiments

Dataset classes:

0-others (backgoround)
1-ground
2-low vegetation
3-buildings
4-water
5-road
6-tree

Network Architectures

<div align="center"> <img src="resources/cnn_fuse.png" width = "620" height = "620" alt="GAMUS" align=center /> </div> <div align="center"> <img src="resources/trans_fuse.png" width = "990" height = "415" alt="RSMSS" align=center /> </div>

References

@article{xiong2023gamus,
  title={GAMUS: A Geometry-aware Multi-modal Semantic Segmentation Benchmark for Remote Sensing Data},
  author={Xiong, Zhitong and Chen, Sining and Wang, Yi and Mou, Lichao and Zhu, Xiao Xiang},
  journal={arXiv preprint arXiv:2305.14914},
  year={2023}
}