安装MMdetection_蜉蝣之翼❉的博客

cat /proc/version
gcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)
CUDA 版本

nvcc --version
Cuda compilation tools, release 10.2, V10.2.89
CUDA_HOME=/usr/local/cuda
CUDNN

下载linux版本，解压
~/software$ tar -zxvf cudnn-10.2-linux-x64-v8.0.4.30.tgz
所有的文件会解压到~/software/cuda
以下都在~/software目录下操作
添加路径，其中 /home/admin17/software/cuda/lib64 是我下载cudnn解压的路径
gedit  ~/.bashrc
export LD_LIBRARY_PATH=/home/admin17/software/cuda/lib64:$LD_LIBRARY_PATH
source  ~/.bashrc
cd ~/software
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
有一些高版本的版本信息在cudnn_version.h 中，需要将cudnn_version.h 复制过去 
sudo cp cuda/include/cudnn_version.h /usr/local/cuda/include
 添加读写权限： 
sudo chmod a+r /usr/local/cuda/include/cudnn.h
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*
sudo chmod a+r /usr/local/cuda/include/cudnn_version.h
低版本查看版本的命令：
 cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
 高版本查看版本的命令：
 cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2
 输出 
UDNN_MAJOR -A 2
#define CUDNN_MAJOR 8
#define CUDNN_MINOR 0
#define CUDNN_PATCHLEVEL 4
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#endif /* CUDNN_VERSION_H */
说明版本 8.0.4 
conda
 
conda -V
 conda 4.5.11 
Anaconda
 
You need install anaconda at first, then run below 
conda create -n platform python=3.7.0
conda activate platform
PyTorch
 
pip install torch==1.6.0
pip install torchvision==0.7.0
MMdetection
 
git clone https://github.com/open-mmlab/mmdetection.git
cd mmdetection
git checkout v2.8.0
pip install mmcv-full
pip install matplotlib==3.1.0
pip install -r requirements/build.txt
python setup.py develop
可能的错误：
 gnutls_handshake() failed: The TLS connection was non-properly terminated.
 网不好 重新 git clone https://github.com/open-mmlab/mmdetection.git 
Pyinstaller
 
pip install pyinstaller
test_endecode
 
解压 /home/admin17/code/platform/test_endecode.zip
 conda activate platform
 cd ~/code/platform/test_endecode/
 mkdir build
 cmake … 
报错
 Could not find a package configuration file provided by “pybind11” with any
 of the following names: 
pybind11Config.cmake
pybind11-config.cmake
Add the installation prefix of “pybind11” to CMAKE_PREFIX_PATH or set
 “pybind11_DIR” to a directory containing one of the above files. If
 “pybind11” provides a separate development package or SDK, be sure it has
 been installed. 
查找Pyinstaller
 pip show Pillow
 Location: /home/admin17/anaconda3/envs/platform/lib/python3.7/site-packages 
在Location 搜索pybind11Config.cmake 找到目录
 /home/admin17/anaconda3/envs/platform/lib/python3.7/site-packages/pybind11/share/cmake/pybind11 
在/home/admin17/code/platform/test_endecode/CMakeLists.txt中添加
 set(pybind11_DIR /home/admin17/anaconda3/envs/platform/lib/python3.7/site-packages/pybind11/share/cmake/pybind11) 
重新编译
 conda activate platform
 cd ~/code/platform/test_endecode/
 mkdir build
 cmake …
 make
 成功 
gedit ~/.bashrc
 在.bashrc中添加 
export LD_LIBRARY_PATH=/home/admin17/code/platform/test_endecode/lib:$LD_LIBRARY_PATH
采用命令source ~/.bashrc激活 
pytest-runner
 
pip install pytest-runner
 pip install mmcv-full==1.2.6 
sudo dpkg -i codemeter_7.0.3925.501_amd64.deb 不一定
 /home/admin17/code/platform/utils/utils.py", line 11, in 
 import endecode
 ModuleNotFoundError: No module named 'endecode’
 我注释了endecode相关代码
                    查看版本ubuntucat /proc/versiongcc version 7.5.0 (Ubuntu 7.5.0-3ubuntu1~18.04)CUDA 版本nvcc --versionCuda compilation tools, release 10.2, V10.2.89CUDA_HOME=/usr/local/cudaCUDNN下载linux版本，解压~/software$ tar -zxvf cudnn-10.2-linux-x64-v8.0.4.30.tgz所有的文件会
-- Python checking for Python version 3.5 or greater
-- Python checking for Python version 3.5 or greater - found
-- Python checking for Mako templates 0.4.2 or greater
-- Python checking for Mako templates 0.4.2 or greater - "import mako
	windows10
	无cuda
用c++调用pytorch模型官网上面有详细教程,也有很多博客,可以参考以下链接:pytorch怎么使用c++调用部署模型？_豆芽菜-CSDN博客_pytorch用c++
以及:Pytorch的C++前端(libtorch)在Windows下的预测 - 知乎
在操作的过程中,主要遇到三个比较大的问题:
问题一:cmake编译时以及找不到torch和opencv相关的cmake文件(以下代码是复制的stac
				Could not find a package configuration file provided by "pybind11" with any of the following names:
pybind11Config.cmake
pybind11-config.cmake
not use pip to install pybind
conda install pybind11
				今天catkin build 整个工程，编译catkin_tools_prebuild时遇到失败的情况，如下：
The catkin CMake module was not found, but it is required to build a linked
 workspace. To resolve this, please do one of the following, and ...
CMake Error at CMakeLists.txt:7 (find_package): By not providing "FindOpencv.cmake" in CMAKE_MODULE_PATH this project has asked CMake to find a package configuration file provided by "Opencv", but CMake did not find one.
Could not fin
在ROS编译过程中经常会遇到找不到ROS包的情况，如下所示
CMake Error at /opt/ros/kinetic/share/catkin/cmake/catkinConfig.cmake:83 (find_package):
  Could not find a package configuration file provided by...
    Could not find a package configuration file provided by
      "gazebo_ros_control" with any of the following names:
        gazebo_ros_controlConfig.cmake
        gaz...
				cmake时出现
-- Detected version of GNU GCC: 54 (504)
-- Found ZLIB: /usr/lib/i386-linux-gnu/libz.so (found suitable version "1.2.8", minimum required is "1.2.3") 
-- Could NOT find TIFF (missing:  TIFF_
				1. 首先，你需要安装Anaconda或者Miniconda，这是一个Python的环境管理器，可以方便地安装和管理Python的依赖包。
2. 安装CUDA和cuDNN，这是深度学习框架需要的GPU加速库，可以大大提高训练速度。
3. 安装PyTorch，这是一个深度学习框架，mmdetection是基于PyTorch实现的。
4. 克隆mmdetection的代码库，可以使用git命令：git clone https://github.com/open-mmlab/mmdetection.git
5. 安装mmdetection的依赖包，可以使用conda命令：conda install pytorch torchvision -c pytorch
6. 编译mmdetection的C++扩展，可以使用以下命令：
cd mmdetection
pip install -r requirements/build.txt
pip install "git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI"
pip install -v -e .
7. 运行mmdetection的示例代码，可以使用以下命令：
cd mmdetection
python tools/train.py configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py
以上就是在Ubuntu上安装mmdetection的步骤，希望对你有帮助。