This code is an implementation of head pose estimation for a single image (HPE_image.py) and webcam (HPE_webcam.py). For head pose estimation, I used the algorithm explained here, and integrated it with dlib's facial landmark detector to make it real-time.
Please download dlib pre-trained facial landmark detector here.
-
Python 3.6.7
-
numpy 1.18.1
-
Opencv 4.2.0
You can install dlib by:
pip install dlib
“Pip install dlib” doesn’t work for Windows. Therefore, use the following procedure: First download “wheel” from here.
In the above link, you can see several versions. For example, for 64-bit Windows and Python 3.6, download the following wheel file:
dlib-19.8.1-cp36-cp36m-win_amd64.whl
Then, open Windows command prompt, go to the directory where the wheel file is saved by “cd” command. Then, install dlib using following command:
pip install [wheel file name]
example:
pip install dlib-19.8.1-cp36-cp36m-win_amd64.whl
If you would like to install dlib in a virtual environment, activate the desired virtual environment before running the above command.
Finally, close the command prompt and open a new command prompt. Now, you can import and use dlib.
Note: There is no wheel file for Python 3.7 for Windows environment. Please use Python 3.6.
Dlib can be installed by conda as explaniend here.