Extract the one-hot encoding vector.

extract_onehot_vector.py

from sklearn.preprocessing import LabelEncoder, OneHotEncoder

X_str = np.array([['a', 'dog', 'red'], ['b', 'cat', 'green']])
# transform to integer
X_int = LabelEncoder().fit_transform(X_str.ravel()).reshape(*X_str.shape)
# transform to binary
X_bin = OneHotEncoder().fit_transform(X_int).toarray()

print(X_bin)
# [[ 1.  0.  0.  1.  0.  1.]
#  [ 0.  1.  1.  0.  1.  0.]]