The width of the field the camera captures is relative to how far away any object is, more precisely:
which gives us:
So what's the view width at the point where the face is? We know that the proportion of the screen the face will fill is given by:
However, this is the same for pixels, so:
And this gives us:
Inserting this, we get:
In our case is the width of the canvas, is the width of the face on the canvas, and we assume is around 17 cm.