SASSD复现

创建环境

conda create -n SASSD python=3.7

进入环境

conda activate SASSD

先配置清华源

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --set show_channel_urls yes

安装pytorch

conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=10.2 

安装相关环境依赖包

pip install opencv-python -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install shapely -i https://pypi.tuna.tsinghua.edu.cn/simple

安装mayavi，要注意先安装其他依赖包！！！

pip install vtk==8.1.2
pip install pyside2 -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install mayavi -i https://pypi.tuna.tsinghua.edu.cn/simple

安装cmake

参考Ubuntu安装cmake

安装spconv1.2.1

从github上下载1.2.1版本

GitHub - traveller59/spconv: Spatial Sparse Convolution Library

或者用 命令下载，注意查看是否是1.2.1版本，不是的话可以手动下载

git clone https://github.com/traveller59/spconv.git --recursive

下载完成后，从终端进入spconv执行命令

 python setup.py bdist_wheel
 cd ./dist && pip install *

mmdet/ops中，编译 C++/CUDA模块

 cd mmdet/ops/points_op
 python setup.py build_ext --inplace
 cd mmdet/ops/iou3d
 python setup.py build_ext --inplace
 cd mmdet/ops/pointnet2
 python setup.py build_ext --inplace

因为我们使用的环境版本比作者的高，可能会出现报错

setup.py build_ext fail RuntimeError: Error compiling objects for extension

需要修改文件配置

SA-SSD/mmdet/ops/points_op/src/points_op.cpp
#define CHECK_CONTIGUOUS(x) AT_CHECK(x.is_contiguous(), #x, " must be contiguous ")
to
#define CHECK_CONTIGUOUS(x) AT_ASSERT(x.is_contiguous())

SA-SSD/mmdet/ops/pointnet2/src/interpolate.cpp
cudaStream_t stream = THCState_getCurrentStream(state);
to
cudaStream_t stream = at::cuda::getCurrentCUDAStream();

SA-SSD/mmdet/ops/iou3d/src/iou3d.cpp
#define CHECK_CUDA(x) AT_CHECK(x.type().is_cuda(), #x, " must be a CUDAtensor ")
#define CHECK_CONTIGUOUS(x) AT_CHECK(x.is_contiguous(), #x, " must be contiguous ")
to
#define CHECK_CUDA(x) AT_ASSERT(x.type().is_cuda())
#define CHECK_CONTIGUOUS(x) AT_ASSERT(x.is_contiguous())

to前面是源码，后面是修改后的代码！

未完待续.....