在Ubuntu24.04中配置开源直线特征提取软件DeepLSD
本文提供在Ubuntu24.04中配置开源直线特征提取软件DeepLSD的基础环境配置、列出需要修改的文件内容,以及报错解决方案集锦。
基础的编译安装环境
- python3.8.12
- CUDA12
- gcc/g++ 9.5(系统自带的g++-13版本太新,会产生额外编译错误)
切换系统默认的gcc/g++版本
Ubuntu24中可以使用update-alternatives命令指定系统当前的gcc/g++版本,命令如下:
sudo apt-get install gcc-9 g++-9
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-9 9
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-9 9
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-13 13
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-13 13
sudo update-alternatives --config gcc
sudo update-alternatives --config g++
上述命令首先安装gcc-9 g+±9,然后利用--install
参数将系统中的两个版本的gcc/g++各自标记为"9"/“13”,最后利用--config
选择gcc/g++ 9为默认版本。
安装glog、gflags
DeepLSD全功能中用到了glog和gflags两个库,使用如下命令安装:
sudo apt-get install libgoogle-glog-dev libgflags-dev
编译ceres
ceres solver除了依赖glog和gflags之外,还依赖ATLAS(BLAS和LAPACK)、Eigen和SuiteSparse
因此为编译ceres,还需要运行以下命令安装这些依赖:
sudo apt-get install libatlas-base-dev
sudo apt-get install libeigen3-dev
sudo apt-get install libsuitesparse-dev
后续安装命令:
wget -O ceres-solver-2.2.0.tar.gz http://ceres-solver.org/ceres-solver-2.2.0.tar.gz
tar zxf ceres-solver-2.2.0.tar.gz
mkdir ceres-bin
cd ceres-bin
cmake ../ceres-solver-2.2.0
make -j3
make test
sudo make install
编译OpenCV-Contrib
依赖OpenCV的线提取模块,因此需要编译OpenCV和Contrib。
否则将报错:
DeepLSD/third_party/pytlbd/src/LineBandDescriptor.cpp:9:10: fatal error: opencv2/line_descriptor.hpp: No such file or directory
9 | #include <opencv2/line_descriptor.hpp>
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
gmake[2]: *** [CMakeFiles/tlbd.dir/build.make:79: CMakeFiles/tlbd.dir/src/LineBandDescriptor.cpp.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:185: CMakeFiles/tlbd.dir/all] Error 2
gmake: *** [Makefile:146: all] Error 2
以opencv4.11.0版本为例,编译命令为:
wget -O https://github.com/opencv/opencv/archive/refs/tags/4.11.0.zip
wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/4.11.0.zip
unzip opencv_contrib.zip
mkdir opencv_build
cd opencv_build/
conda activate DeepLSD
cmake -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.11.0/modules ../opencv-4.11.0
make
sudo make install
源码修改
解决报错:error: ‘unordered_map’ in namespace ‘std’ does not name a template type
修改以下两个文件:
- flann_neighborhood_graph.h
- neighborhood_graph.h
在这两个文件的开头包含<unordered_map>
解决报错:error: ‘clamp’ is not a member of ‘std’
在VSCode中搜索包含std::clamp
的文件。在所有调用std::clamp
的文件开头加上包含语句#include <algorithm>
。需要修改的文件包括(可能也不限于):
DeepLSD/third_party/progressive-x/examples/cpp_example.cpp
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/GCRANSAC.h
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/estimators/essential_estimator.h
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/estimators/estimator.h
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/estimators/fundamental_estimator.h
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/estimators/perspective_n_point_estimator.h
DeepLSD/third_party/progressive-x/graph-cut-ransac/src/pygcransac/include/estimators/solver_p3p.h
除此之外,还要修改.cmake文件的配置使用c++17标准编译,详见下节。
更改cmake配置
将以下两个文件:
DeepLSD/third_party/progressive-x/lib/pybind11/tools/pybind11Tools.cmake
DeepLSD/third_party/progressive-x/graph-cut-ransac/lib/pybind11/tools/pybind11Tools.cmake
关于设置PYBIND11的c++语言标准部分改为:
function(select_cxx_standard)
if(NOT MSVC AND NOT PYBIND11_CPP_STANDARD)
check_cxx_compiler_flag("-std=c++17" HAS_CPP17_FLAG)
check_cxx_compiler_flag("-std=c++11" HAS_CPP11_FLAG)
if (HAS_CPP17_FLAG)
set(PYBIND11_CPP_STANDARD -std=c++17)
elseif (HAS_CPP11_FLAG)
set(PYBIND11_CPP_STANDARD -std=c++11)
else()
message(FATAL_ERROR "Unsupported compiler -- pybind11 requires C++11 support!")
endif()
set(PYBIND11_CPP_STANDARD ${PYBIND11_CPP_STANDARD} CACHE STRING
"C++ standard flag, e.g. -std=c++11 or -std=c++14. Defaults to latest available." FORCE)
endif()
删除CMakeCache.txt,重新运行bash install.sh
,即可。
CMake最低版本更改
在VSCode中打开DeepLSD文件夹,搜索如下的表达式(正则):
将cmake_minimum_required(VERSION 2.*)或cmake_minimum_required(VERSION 2.8.12)
改为
cmake_minimum_required(VERSION 3.11)
将3.x版本都改为3.11:
搜索:cmake_minimum_required(VERSION 3.[0-5])
替换为:cmake_minimum_required(VERSION 3.11)
报错解决集锦
忽略 /opt/anaconda3 目录及其子目录
Anaconda3中存在一些库如glog/gflags甚至gdal等,可能在cmake配置过程中被当作最高优先级的链接目标,而这是不对的。因此,如果出现链接错误“GLIBCXX_3.4.30”类似的错误:
libopencv_core : undefined reference to std::condition_variable::wait(std::unique_lockstd::mutex&)@GLIBCXX_3.4.30’
std::condition_variable::wait(std::unique_lockstd::mutex&)@GLIBCXX_3.4.30’
则考虑在报错的项目的CMakeLists.txt中添加下面的语句以忽略Anaconda的库目录:
list(APPEND CMAKE_IGNORE_PATH "/opt/anaconda3")
fatal error: arlsmat.h: 没有那个文件或目录 #include <arlsmat.h>
sudo apt install libarpack*
OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.
一般由于没有安装CUDA导致。运行以下命令安装CUDA:
wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-keyring_1.0-1_all.deb
sudo dpkg -i cuda-keyring_1.0-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda
python38/bin/…/lib/libstdc++.so.6: version `GLIBCXX_3.4.30’ not found
编译完成后在运行时可能会产生此错误。DeepLSD编译链接时所使用的是系统的libstdc++.so.6,而python运行时加载了Anaconda自带的libstdc++.so.6,其版本比系统的libstdc++.so.6旧,所以找不到符号。
首先,运行下面命令确定系统的libstdc++.so.6具有GLIBCXX_3.4.30版本:
strings /usr/lib/x86_64-linux-gnu/libstdc++.so.6 | grep GLIBCXX
如果命令行的输出有此版本,则删除python38/bin/…/lib/libstdc++.so.6,并创建符号链接:
rm ~/miniconda3/envs/python38/lib/libstdc++.so.6
ln -sf /usr/lib/x86_64-linux-gnu/libstdc++.so.6 ~/miniconda3/envs/python38/lib/libstdc++.so.6