Download and install TensorRT-7.2.2.3
Download tensorrt-7.2.2.3 on the official website Ubuntu-18.04. x86_ 64-gnu. cuda-11.1. cudnn8. 0.tar. gz
tar xzvf TensorRT-7.2.2.3.Ubuntu-18.04.x86_64-gnu.cuda-11.1.cudnn8.0.tar.gz
Configure environment variables after decompression:
sudo vim ~/.bashrc
After entering, add environment variables at the bottom:
export TRT_PATH=/home/lindsay/TensorRT-7.2.2.3 export PATH=$PATH:$TRT_PATH/bin export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$TRT_PATH/lib export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$TRT_PATH/targets/x86_64-linux-gnu/lib
Enter the python folder under the decompression path and install with pip:
cd TensorRT-7.2.2.3/python pip3 install tensorrt-7.2.2.3-cp37-none-linux_x86_64.whl
Install uff to support tensorflow
cd TensorRT-7.2.2.3/uff pip3 install uff-0.6.9-py2.py3-none-any.whl
Install graphsurgeon to support custom structures
cd TensorRT-7.2.2.3/graphsurgeon pip3 install graphsurgeon-0.4.5-py2.py3-none-any.whl
Install onnx_graphsurgeon to support onnx
cd TensorRT-7.2.2.3/onnx_grahsurgeon pip3 install onnx_graphsurgeon-0.2.6-py2.py3-none-any.whl
OpenCV download and installation
Download address: OpenCV official website, select the latest version 4.4.0 (if the download speed is too slow, copy the link address and use Xunlei)
https://opencv.org/releases/
Copy the downloaded compressed file to the virtual machine
Compilation and installation
Installing cmake
OpenCV needs to be compiled using cmake
sudo apt-get install cmake
Installation dependency
sudo apt-get install build-essential pkg-config libgtk2.0-dev libavcodec-dev libavformat-dev libjpeg-dev libswscale-dev libtiff5-dev
The following problems occurred:
The following packages have unmet dependencies:
libavcodec-dev : Depends: libavutil-dev (= 7:3.4.2-2) but it is not going to be installed
Depends: libswresample-dev (= 7:3.4.2-2) but it is not going to be installed
libavformat-dev : Depends: libavformat57 (= 7:3.4.2-2) but it is not going to be installed
Depends: libavutil-dev (= 7:3.4.2-2) but it is not going to be installed
Depends: libswresample-dev (= 7:3.4.2-2) but it is not going to be installed
libgtk2.0-dev : Depends: libglib2.0-dev (>= 2.27.3) but it is not going to be installed
Depends: libgdk-pixbuf2.0-dev (>= 2.21.0) but it is not going to be installed
Depends: libpango1.0-dev (>= 1.20) but it is not going to be installed
Depends: libatk1.0-dev (>= 1.29.2) but it is not going to be installed
Depends: libcairo2-dev (>= 1.6.4-6.1) but it is not going to be installed
Depends: libx11-dev (>= 2:1.0.0-6) but it is not going to be installed
Depends: libxext-dev (>= 1:1.0.1-2) but it is not going to be installed
Depends: libxinerama-dev (>= 1:1.0.1-4.1) but it is not going to be installed
Depends: libxi-dev (>= 1:1.0.1-4) but it is not going to be installed
Depends: libxrandr-dev (>= 2:1.2.99) but it is not going to be installed
Depends: libxcursor-dev but it is not going to be installed
Depends: libxfixes-dev (>= 1:3.0.0-3) but it is not going to be installed
Depends: libxcomposite-dev (>= 1:0.2.0-3) but it is not going to be installed
Depends: libxdamage-dev (>= 1:1.0.1-3) but it is not going to be installed
Recommends: python (>= 2.4) but it is not going to be installed
libjpeg-dev : Depends: libjpeg8-dev but it is not going to be installed
libswscale-dev : Depends: libavutil-dev (= 7:3.4.2-2) but it is not going to be installed
Depends: libswscale4 (= 7:3.4.2-2) but 7:3.4.8-0ubuntu0.2 is to be installed
libtiff5-dev : Depends: libtiff5 (= 4.0.9-5) but 4.0.9-5ubuntu0.4 is to be installed
E: Unable to correct problems, you have held broken packages.
Using aptitude
Aptitude, like apt get, is an extremely powerful package management tool in Debian and its derivative systems. Unlike apt get, aptitude is better at dealing with dependency issues. For example, when aptitude deletes a package, it will delete the package it depends on at the same time. In this way, there will be no useless packages left in the system, and the whole system will be cleaner.
sudo aptitude install install build-essential pkg-config libgtk2.0-dev libavcodec-dev libavformat-dev libjpeg-dev libswscale-dev libtiff5-dev
After operation, the non installed scheme is not accepted, and the degraded scheme is accepted.
decompression
unzip opencv-4.4.0
Enter the file directory, create the build directory and enter
cd opencv-4.4.0/ mkdir build cd build
Generating makefile files using cmake
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D WITH_GTK=ON -D OPENCV_GENERATE_PKGCONFIG=YES ..
CMAKE_BUILD_TYPE=RELEASE: indicates the compiled release version
CMAKE_INSTALL_PREFIX: indicates the installation path of generating dynamic library, which can be customized
WITH_GTK=ON: this configuration is to prevent GTK configuration failure: libgtk2 is installed 0-dev dependency, or an error is reported and not installed
OPENCV_GENERATE_PKGCONFIG=YES: indicates that the pkgconfig file of OpenCV is generated automatically, otherwise it needs to be generated manually.
compile
make -j8
-j8 means that multiple system kernels are used for compilation to improve the compilation speed. If you don't know the number of your own system kernels, you can use make -j$(nproc)
If there are errors in the compilation, you can try not to compile with multiple kernels. Although it takes longer to compile, you can avoid some strange errors
install
sudo make install
Note: if you need to cmake again, please empty the files in the build directory first, and then cmake again to avoid errors
Environment configuration
Add OpenCV's library to the system path
Method 1: configure LD so. Conf file
sudo vim /etc/ld.so.conf
Add a line include /usr/loacal/lib in the file. This path is the dynamic library installation path filled in during cmake compilation, plus / lib
Configure LD so. Conf file
Method 2: manually generate OpenCV Conf file
sudo vim /etc/ld.so.conf.d/opencv.conf
It is a new empty file, and the path is added directly. Similarly, this path is the dynamic library installation path filled in during cmake compilation plus / lib
/usr/local/lib
After the above two methods are configured, execute the following commands to make the configured path effective
sudo ldconfig
Configure system bash
Because the pkgconfig file that automatically generates OpenCV is selected during cmake, you can see the file in / usr/local/lib/pkgconfig path
opencv4.pc
Make sure the file exists and execute the following command
sudo vim /etc/bash.bashrc
Add at the end of the text
PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig export PKG_CONFIG_PATH
As follows:
bash.bashrc
Save and exit, and then execute the following command to make the configuration effective
source /etc/bash.bashrc
So far, Linux \ Ubuntu 18 The installation and configuration of OpenCV in 04 environment have been completed. You can use the following command to check whether the installation and configuration are successful
pkg-config --modversion opencv4 pkg-config --cflags opencv4 pkg-config --libs opencv4
Using TensorRT
Download yolov5 and tensorrtx
git clone https://github.com/wang-xinyu/tensorrtx.git git clone https://github.com/ultralytics/yolov5.git
Copy tensorrt/yolov5 to yolov5: sudo cp -r tenorrtx/yolov5 yolov5
Generate WTS corresponding to pt: Python Gen_ wts. py
Put wts into the build peer directory
Modify cmakelists txt:
# tensorrt #include_directories(/usr/include/x86_64-linux-gnu/) #link_directories(/usr/lib/x86_64-linux-gnu/) include_directories(/home/lindsay/TensorRT-7.2.2.3/include) link_directories(/home/lindsay/TensorRT-7.2.2.3/lib)
Modify yololayer h:
static constexpr int CLASS_NUM = 80;#Modify according to your own category
Start compiling and testing:
mkdir build && cd build cmake .. make -j6 sudo ./yolov5 -s ../yolov5s.wts ../yolov5s.engine s# Generation engine sudo ./yolov5 -d ../yolov5s.engine ../samples#Test c++ python3 yolov5_trt.py #Test python