    def from_existing(cls, confusion, *args, **kwargs):
        """Creates a confusion matrix from a DataFrame that already contains confusion counts (but not meta stats)
        >>> df = pd.DataFrame(np.matrix([[0,1,2,0,1,2,1,2,2,1],[0,1,2,1,2,0,0,1,2,0]]).T, columns=['True', 'Pred'])
        >>> c = Confusion(df)
        >>> c2 = pd.DataFrame(c)
        >>> hasattr(c2, '_binary_sensitivity')
        False
        >>> c3 = Confusion.from_existing(c2)
        >>> hasattr(c3, '_binary_sensitivity')
        True
        >>> (c3 == c).all().all()
        True
        >>> c3
        Pred  0  1  2
        True
        0     1  1  0
        1     2  1  1
        2     1  1  2
        """
        # Extremely brute-force to recreate data from a confusion matrix!

        df = []
        for t, p in product(confusion.index.values, confusion.columns.values):
            df += [[t, p]] * confusion[p][t]
        if confusion.index.name is not None and confusion.columns.name is not None:
            return Confusion(pd.DataFrame(df, columns=[confusion.index.name, confusion.columns.name]))
        return Confusion(pd.DataFrame(df))