The repository contains official Pytorch implementations for LRPNet.
For Jittor user, https://github.com/li-xl/LRPNet is a jittor version.
The paper is in Here.
pip install numpy torch tensorboardX open3d
pip install git+https://github.com/mit-han-lab/[email protected]
Download the scannet and prepare it.
PYTHONPATH=./:$PYTHONPATH python3 tools/prepare_scannet.py --meta_path=data/meta --in_path=<scannetpath>
--out_path=data/scannet
We release our trained models and training logs in "work_dirs".
Like VMNet, we repeat val/test for 8 times.
To evaluate the model, run:
bash run.sh 0 configs/scannet_largenet_f10_scale_val.py --task=val
# for test
bash run.sh 0 configs/scannet_largenet_f10_scale_trainval_test.py --task=test
If you find our repo useful for your research, please consider citing our paper:
@article{li2023long,
title={Long Range Pooling for 3D Large-Scale Scene Understanding},
author={Li, Xiang-Li and Guo, Meng-Hao and Mu, Tai-Jiang and Martin, Ralph R and Hu, Shi-Min},
journal={arXiv preprint arXiv:2301.06962},
year={2023}
}
This repo is under the Apache-2.0 license. For commercial use, please contact the authors.