what does the model return if there is no face in the image?

[segalinc]

I was trying to pass a landascape image with no faces at all. the model seems to predict anyway the landmarks.. this seems wrong