MTCNN for Tencent's ncnn, based on cpuimage's C++ project.
libmtcnn is a static library.
mtcnn is a shared library.
mtcnn-stdout is a usage example that links to the static library.
mtcnn-window is a usage example that links to the shared library.
libmtcnn and mtcnn depend on ncnn.
mtcnn-stdout depends on libmtcnn.
mtcnn-window depends on mtcnn, SDL2, and SDL2_image.
mtcnn-stdout and mtcnn-window accept command-line parameters:
- -n {minimum face size}
- -x {maximum face size}
- {image file name}
These can be a ratio between 0 and 1 (relative to the smaller side of the image) or an absolute number of pixels. The limit for minimum is 12 pixels, which is the size of the model used in the first convolution. The limit for maximum is the value of the smaller side of the image.
For now, see mtcnn.h, along with examples in mtcnn-stdout.cpp and mtcnn-window.cpp. More documentation will follow.