看到github上的一个开源项目:https://github.com/BaofengZan/DBNet-TensorRT
就用来练手TensorRT。
首先下载tensorRT7库,这个在前面的笔记(一)中已经有过说明,此处说明我下载的是:
TensorRT-7.0.0.11.Windows10.x86_64.cuda-10.0.cudnn7.6
相应的,我本地也安装了cuda10.0和cudnn7.6.5。
将上面Github上项目下载到本地。
git clone https://github.com/BaofengZan/DBNet-TensorRT.git
然后在DBNet-TensorRT文件夹下新建build文件夹,在build文件夹下启动powershell,执行:
cmake-gui ..
通过CMAKE-GUI配置项目。
原项目CmakeLists.txt内容如下(自己使用时需要根据本地的安装路径设置 opencv路径和tensorrt路径):
cmake_minimum_required(VERSION 2.6)
project(dbnet)
add_definitions(-std=c++11)
option(CUDA_USE_STATIC_CUDA_RUNTIME OFF)
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_BUILD_TYPE Debug)
#设置cuda信息
find_package(CUDA REQUIRED)
message(STATUS " libraries: ${CUDA_LIBRARIES}")
message(STATUS " include path: ${CUDA_INCLUDE_DIRS}")
include_directories(${CUDA_INCLUDE_DIRS})
set(CUDA_NVCC_PLAGS ${CUDA_NVCC_PLAGS};-std=c++11; -g; -G;-gencode; arch=compute_75;code=sm_75)
enable_language(CUDA) # 这一句添加后 ,就会在vs中不需要再手动设置cuda
include_directories(${PROJECT_SOURCE_DIR}/include)
include_directories(D:\\TensorRT-7.0.0.11.Windows10.x86_64.cuda-10.2.cudnn7.6\\TensorRT-7.0.0.11\\include)
#-D_MWAITXINTRIN_H_INCLUDED 解决error: identifier "__builtin_ia32_mwaitx" is undefined
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -Ofast -D_MWAITXINTRIN_H_INCLUDED")
# 设置opencv的信息
set(OpenCV_DIR "D:\\opencv\\opencv346\\build")
find_package(OpenCV QUIET
NO_MODULE
NO_DEFAULT_PATH
NO_CMAKE_PATH
NO_CMAKE_ENVIRONMENT_PATH
NO_SYSTEM_ENVIRONMENT_PATH
NO_CMAKE_PACKAGE_REGISTRY
NO_CMAKE_BUILDS_PATH
NO_CMAKE_SYSTEM_PATH
NO_CMAKE_SYSTEM_PACKAGE_REGISTRY
)
message(STATUS "OpenCV library status:")
message(STATUS " version: ${OpenCV_VERSION}")
message(STATUS " libraries: ${OpenCV_LIBS}")
message(STATUS " include path: ${OpenCV_INCLUDE_DIRS}")
include_directories(${OpenCV_INCLUDE_DIRS})
link_directories(D:\\TensorRT-7.0.0.11.Windows10.x86_64.cuda-10.2.cudnn7.6\\TensorRT-7.0.0.11\\lib)
add_executable(dbnet ${PROJECT_SOURCE_DIR}/dbnet.cpp)
target_link_libraries(dbnet "nvinfer" "nvinfer_plugin" "nvparsers" "nvonnxparser")
#target_link_libraries(dbnet "cudart")
target_link_libraries(dbnet ${OpenCV_LIBS})
target_link_libraries(dbnet ${CUDA_LIBRARIES})
add_definitions(-pthread)
注意需要将tensorrt的库文件拷贝到生成的可执行程序目录,这些文件包括(此外还包括自己相应的opencv动态库):
将下面路径改成自己的路径
编译完成后先执行:
dbnet.exe -s
会在当前目录下生成DBNet.engine文件。
然后再执行:
dbnet.exe -d ./test
执行效果如下:
更多细节后续文章会加以说明。