---

Preparing data

We use datasets including ScanNet, ScanRefer, ReferIt3D, and the Multi3DRefer.

Install

Install MinkowskiEngine:

# install MinkowskiEngine for Mask3D detector
git clone --recursive "https://github.com/NVIDIA/MinkowskiEngine" # clone the repo to third_party
cd MinkowskiEngine
git checkout 02fc608bea4c0549b0a7b00ca1bf15dee4a0b228
python setup.py install --force_cuda --blas=openblas
pip install hydra-core loguru albumentations open3d

Install PointNet++

pip install "git+https://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib"

Run Training

bash scripts/train.sh

Citation

If you find this code useful for your research, please cite the following paper.

@misc{huang2026pointlinguistmodelsegment,
      title={Point Linguist Model: Segment Any Object via Bridged Large 3D-Language Model}, 
      author={Zhuoxu Huang and Mingqi Gao and Jungong Han},
      year={2026},
      eprint={2509.07825},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2509.07825}, 
}