First, I optimise the box collision method described in tutorial #3, then I show a different, shorter method that allows pixel-perfect detection by ignoring the transparency.