C++ Support

[SriramEmarose]

Hi All,

It is a great work. Thanks for the efforts. It is giving really good results.

One doubt is, Is there any API for the same library in C/C++? Though it is derived from Dlib, I hope the "FaceRecognition" function that Dlib offers is different from this implementation.

Would be a great help if you could point me to any such API calls if available.

Thanks, Sriram

[ageitgey]

If you want to use C++, just use dlib directly since C++ is what it's mainly designed for. See http://dlib.net/dnn_face_recognition_ex.cpp.html

The goal of this library is just to make face recognition super easy in Python and to provide an off-the-shelf solution that "just works" as much as possible (models automatically included, lots of examples, etc). But dlib is still doing all the face encoding and face detection work, so you should get the same results if you use the dlib's C++ api directly.

Some day this library may pull in alternate face detection/encoding models from outside dlib if needed, but that's not currently the case.

[SriramEmarose]

Hi,

Thank you very much for the prompt response. Will check it.

Thanks

[ageitgey]

Sure thing :)

[Jayhello]

@ageitgey , hello, I use dlib to do face detection (C++ code), core code like below:

dlib::frontal_face_detector detector = dlib::get_frontal_face_detector();

std::string img_path = "/data/service/face_rec/face_test/wlj1 (1).jpg";
dlib::array2d<unsigned char> img;
dlib::load_image(img, img_path);

std::vector<dlib::rectangle> dets = detector(img, 1);

I found the above code is about 5 times slow than python interface, do you know why?

[ghost]

@Jayhello Seriously?.This code is 5 times slower than python interface?.wow?.have you find the reason behind it?

[creatist]

@Jayhello Seriously?.This code is 5 times slower than python interface?.wow?.have you find the reason behind it?

I got the same mistake conclusion ago. But when i test Python API and C++ API again, i found C++ API is faster than Python API. Remember that the speed of face detector will be affected by the image size significantly.