We will use vcpkg to manage dependencies. Instructions for installing can be found at the vcpkg getting-started page but these are the basic steps. We will install it into C:\ for simplicity.
chdir C:\
git clone https://github.com/Microsoft/vcpkg.git
.\vcpkg\bootstrap-vcpkg.bat
Enter the vcpkg root to access the executable
chdir vcpkg
Install the opencv3 package (with extra contrib) for x64 architecture
vcpkg.exe install opencv3[contrib]:x64-windows --recurse
Enter the vcpkg root to access the executable
chdir vcpkg
Install the dlib package for x64 architecture
vcpkg.exe install dlib:x64-windows --recurse
Navigate to the project root (where the CMakeLists.txt file is located) and initialize for x64 architecture
cmake -A x64 -S . -B "build64"
Then build the project (defaults to Debug)
cmake --build build64 --config Release
The executable and (required DLLs) will be found as:
build64/Release/face_detection.exe
Before running copy the file image1.jpg from assets. After running the file output.jpg contains the results.
The folder assets contains the DNN model files:
deploy.prototxtres10_300x300_ssd_iter_140000_fp16.caffemodel
These contain the network configuration and trained weights respectively and the detector is setup to use them in CMakeLists.txt via variables.
- FACE_DETECTION_CONFIGURATION
- FACE_DETECTION_WEIGHTS
At the moment the test images have to be copied from the assets folder into the executable folder. The filename is hardcoded in main.cpp. Output always goes to output.png