Hacker News Reader

You've created a nice little chicken and egg problem for yourself.

I think he means first detect that there is a face at all, and where in the image it is. When you know that, do further processing.

That's the idea.