1) As we know, the DLIB can detect frontal face with 99.95% accuracy and to an extent, tilted faces with approximatly 85.0% accuracy. But it cannot detected profile-faces. Can you let me know if you can develop detection of profile-faces. I need a minimum of 95% detection rate for both tiled-sideways-faces and profile-faces cases.
2) For training the DLIB for tiled-sideways-faces and profile-faces, which dataset you are planning to use? You have something in mind that can be used from internet?