I would very much hope that the available training data for "man fucking fish" is pretty sparse, maybe too sparse to build a training set. (I'm not as optimist

I would very much hope that the available training data for "man fucking fish" is pretty sparse, maybe too sparse to build a training set. (I'm not as optimistic about the sparseness of beheading videos, though)