Tried to play opencv on my old surface pro 3. Somehow face_recognition is running very slow on Anaconda python. I suspected that dlib was not optimized somehow.
Try out Windows subsystem for Linux and see how it goes. It is actually quite decent. Many has problem install opencv (with library link problem). It can be avoided by first setting up an virtualenv. I used
pip install opencv-python
pip install face_recognition
to install opencv and face_recognition module. Dlib will take a while to install. Python will complain can’t link to libsm6. But that can be fixed by just installing libsm6 via “sudo apt install”.
face_recognition.face_encoding actually runs 3 times faster under WSL than anaconda. Unfortunately, WSL does not support hardware for the moment. So I can’t actually grab any frame from a webcam and so is useless for CV applications. But it probably is a very promising environment if that part is fixed. For the moment, I probably just have to install Ubuntu along with Windows.