链接 OpenCV 和 TensorRT 到 conda 环境

最新推荐文章于 2024-05-14 15:41:33 发布

涟绮

最新推荐文章于 2024-05-14 15:41:33 发布

阅读量183

点赞数 1

分类专栏：杂七杂八的笔记文章标签： opencv conda python

本文链接：https://blog.csdn.net/liuliu123456790/article/details/131004744

版权

杂七杂八的笔记专栏收录该内容

5 篇文章 0 订阅

订阅专栏

1. 重新安装和编译 OpenCV

Jetson Xavier NX, Jetpack 4.6.1, OpenCV 4.1.1, TensorRT 8.2.1.8

1.1 卸载现有的 OpenCV

sudo sudo apt-get purge *libopencv*
sudo apt autoremove

1.2 安装依赖项

sudo apt-get update
sudo apt-get install -y build-essential cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev
sudo apt-get install -y python2.7-dev python3.6-dev python-dev python-numpy python3-numpy
sudo apt-get install -y libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libdc1394-22-dev
sudo apt-get install -y libv4l-dev v4l-utils qv4l2 v4l2ucp
sudo apt-get install -y wget

1.3 下载并解压 opencv-4.1.1 和 opencv_contrib-4.1.1

wget https://github.com/opencv/opencv/archive/4.1.1.zip -O opencv-4.1.1.zip && unzip opencv-4.1.1.zip
wget https://github.com/opencv/opencv_contrib/archive/4.1.1.zip -O opencv_contrib-4.1.1.zip && unzip opencv_contrib-4.1.1.zip

1.4 下载 boostdesc_bgm

访问 boostdesc_bgm_eg.zip，下载boostdesc_bgm_eg.zip并上传到opencv_contrib-4.1.1文件夹

unzip boostdesc_bgm_eg.zip -d modules/xfeatures2d/src/ && rm boostdesc_bgm_eg.zip

1.5 编译 OpenCV 4.1.1

1.5.1 创建并进入opencv-4.1.1/build文件夹

mkdir -p opencv-4.1.1/build && cd opencv-4.1.1/build

当前的文件夹结构应为

home
└─jetson
  └─wheels
    ├─opencv-4.1.1
    │  └─build    (当前的位置)
    └─opencv_contrib-4.1.1
       └─modules

1.5.2 编译 OpenCV 到 conda 环境

需要配置链接 OpenCV 的 conda 环境的路径为/home/jetson/miniconda/envs/Yolov5, 请自行替换对应环境的路径

cmake \
-DBUILD_opencv_python3=ON \
-DBUILD_opencv_python2=OFF \
-DPYTHON3_EXECUTABLE=/home/jetson/miniconda3/envs/Yolov5/bin/python3.6m \
-DPYTHON_INCLUDE_DIR=/home/jetson/miniconda3/envs/Yolov5/include/python3.6m \
-DPYTHON_LIBRARY=/home/jetson/miniconda3/envs/Yolov5/lib/libpython3.6m.so \
-DPYTHON_NUMPY_INCLUDE_DIRS=/home/jetson/miniconda3/envs/Yolov5/lib/python3.6/site-packages/numpy/core/include \
-DPYTHON_PACKAGES_PATH=/home/jetson/miniconda3/envs/Yolov5/lib/python3.6/site-packages \
-DPYTHON_DEFAULT_EXECUTABLE=/home/jetson/miniconda3/envs/Yolov5/bin/python3.6m \
-DCMAKE_INSTALL_PREFIX=/usr/local \
-DBUILD_EXAMPLES=OFF \
-DOPENCV_GENERATE_PKGCONFIG=ON \
-DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda \
-DCMAKE_BUILD_TYPE=RELEASE \
-DOPENCV_ENABLE_NONFREE=1 \
-DWITH_FFMPEG=1 \
-DCUDA_ARCH_BIN=7.2 \
-DCUDA_ARCH_PTX=7.2 \
-DWITH_CUDA=1 \
-DENABLE_FAST_MATH=1 \
-DCUDA_FAST_MATH=1 \
-DWITH_CUBLAS=1 \
-DBUILD_opencv_xfeatures2d=OFF \
-DOPENCV_EXTRA_MODULES_PATH=../../opencv_contrib-4.1.1/modules ..

1.5.3 编译

make -j6
sudo make install

如果在编译时出现错误，可以在问题解决后直接使用make -j6命令继续编译
OpenCV编译问题笔记
编译后的文件在opencv-4.1.1/build/lib下

3. 链接 OpenCV 和 TensorRT 到 conda 环境

3.1 opencv

完成 opencv4.1.1 的编译后, 要链接到 conda 环境的文件是cv2.cpython-36m-aarch64-linux-gnu.so, 路径是opencv-4.1.1/build/lib/python3/

ln -s /home/jetson/wheels/opencv/opencv-4.1.1/build/lib/python3/cv2.cpython-36m-aarch64-linux-gnu.so /home/jetson/miniconda3/envs/Yolov5/lib/python3.6

当前所处的文件夹是/home/jetson/wheels/opencv-4.1.1/build/
需要链接的 conda 环境的路径是/home/jetson/miniconda3/envs/Yolov5/
请根据需要自行替换路径

3.2 TensorRT

sudo ln -s /usr/lib/python3.6/dist-packages/tensorrt* /home/jetson/miniconda3/envs/Yolov5/lib/python3.6/site-packages/

Jetpack 预装的 TensorRT的路径是/usr/lib/python3.6/dist-packages/tensorrt*
需要链接的 conda 环境的路径是/home/jetson/miniconda3/envs/Yolov5/
请根据需要自行替换路径

OpenCV编译相关问题

1. liblibprotobuf.io

报错信息

[ 26%] Linking CXX static library ../lib/liblibprotobuf.a 
[ 26%] Built target libprotobuf Makefile:162: recipe for target 'all' failed 
make: *** [all] Error 2

解决方法

因为 eigen 库默认安装在了/usr/include/eigen3/Eigen路径下, 需使用下面命令映射到/usr/include路径下
```
sudo apt-get install libeigen3-dev
sudo ln -s /usr/include/eigen3/Eigen /usr/include/Eigen
```

2. opencv_perf_ximgproc

报错信息

[ 85%] Linking CXX executable ../../bin/opencv_perf_ximgproc
[ 85%] Built target opencv_perf_ximgproc
Makefile:162: recipe for target 'all' failed
make: *** [all] Error 2