darknet_vendor
This is a CMake wrapper around darknet, an open source neural network framework.
Building
cd darknet_vendor/
mkdir build/
cd build/
cmake -DCMAKE_BUILD_TYPE=Release ..
make
The package offers two CMake options to control how it is built.
DARKNET_CUDA(default:On) - Whether or not to build with CUDA supportDARKNET_OPENCV(default:On) - Whether or not to build with OpenCV support
These can be disabled prior to compiling if you don't need either of those dependencies.
cmake -DDARKNET_CUDA=Off -DDARKNET_OPENCV=Off ..
ROS 2 and Colcon
The package includes package.xml and should build fine using ROS 2 Dashing.
colcon build --cmake-args -DCMAKE_BUILD_TYPE=Release --packages-select darknet_vendor
Using darknet
If you are using ament_cmake, use ament_target_dependencies() as normal.
find_package(darknet_vendor REQUIRED)
#...
ament_target_dependencies(my_library_or_executable_target
darknet_vendor)Otherwise, use the exported target darknet_vendor::darknet_vendor
find_package(darknet_vendor REQUIRED)
#...
target_link_libraries(my_library_or_executable_target PUBLIC darknet_vendor::darknet_vendor)Then include the darknet.h header in your project to make darknet functions available.
#include <darknet.h>Alternatively, include darknet_vendor/darknet_vendor.h to get preprocessor definitions with info about the installed darknet_vendor version.
#include <darknet_vendor/darknet_vendor.h>