README.md

OTS

The offical implementaion of "Open-set Hierarchical Semantic Segmentation for 3D Scene".

Install

conda env create -f enviroment.yaml
pip install git+https://github.com/NVlabs/nvdiffrast
conda activate tree_seg

cd semantic_sam/ops
bash ./make.sh

cd tree_segmentation/extension/_C
python setup.py build_ext --inplace

Download pretrained weights

Download sam_vit_h_4b8939.pth from Segment Anything in floder weights

Download swinl_only_sam_many2many.pth and swint_only_sam_many2many.pth from Sematic-SAM in floder weights

Acknowledgement

• Segment Anything
• Sematic-SAM

Citation

@InProceedings{icme24_ots,
  title = 	 {Open-set Hierarchical Semantic Segmentation for 3D Scene},
  author =       {Wan, Diwen and Tang, Jiaxiang and Wang, Jingbo and Chen, Xiaokang and Gan, Lingyun and Zeng, Gang},
  booktitle = 	 {IEEE Conference on Multimedia Expo 2024},
  year = 	 {2024},
}