I am using the Facemesh model by media pipe which gives me back an array of landmarks. I want to extract the face only and remove all other surroundings from the image. How could I approach this in the most efficient way?