1、下载PointNext源码
PointNext:下载给github上代码,其中openpoint文件夹里面是空的,下载并移动到该文件夹下。
guochengqian/PointNeXt: [NeurIPS'22] PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies (github.com)https://github.com/guochengqian/PointNeXtguochengqian/openpoints at baeca5e319aa2e756d179e494469eb7f5ffd29f0 (github.com)
https://github.com/guochengqian/openpoints/tree/baeca5e319aa2e756d179e494469eb7f5ffd29f0
2、配置环境
1、安装cuda和cudnn,并配置环境
(超详细)CUDA与cuDNN安装教程-CSDN博客https://blog.csdn.net/qq_45904458/article/details/132144817
2、安装vs2019并配置环境变量
conda create -n openpoints python=3.8
conda activate openpoints
# install PyTorch
pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113
#建议直接下载torch_scatter-2.0.9-cp38-cp38-win_amd64.whl安装,省去编译安装缓慢。
pip install torch-scatter -f https://data.pyg.org/whl/torch-1.11.0%2Bcu113.html
#requirements.txt中的deepspeed==0.3.16
pip install -r requirements.txt
# install cpp extensions, the pointnet++ library
cd openpoints/cpp/pointnet2_batch
python setup.py install
cd ../
# Blow are functions that optional. Necessary only if interested in reconstruction tasks such as completion
cd chamfer_dist
python setup.py install --user
cd ../emd
python setup.py install --user
# grid_subsampling library. necessary only if interested in S3DIS_sphere
cd subsampling
python setup.py build_ext --inplace
cd ..
# # point transformer library. Necessary only if interested in Point Transformer and Stratified Transformer
pointops/setup.py
'''
注释以下代码:
# (opt,) = get_config_vars('OPT')
# os.environ['OPT'] = " ".join(
# flag for flag in opt.split() if flag != '-Wstrict-prototypes'
# )
'''
cd pointops/
python setup.py install
cd ..