I have an n-dimensional numpy array of shape: (3, 3, 3, 64). I would like to increase the count along axis 2 by inserting zeros, so that the new shape is (3, 3, 4, 64).
How do I insert zeros to increase a given axis value of a numpy array?
Add a comment
|
1 Answer
Create a zeros arrays with same shape as the input, but the third axis being of length same as the padding length, which is 1 for our case and concatenate with the input array along the same axis (third axis). For concatenation, we can employ np.concatenate or np.dstack(because its third axis).
Thus, the implementation would be -
z = np.zeros((3, 3, 1, 64),dtype=a.dtype)
out = np.concatenate((a,z),axis=2) # Or np.dstack((a,z))
Sample run -
In [182]: a = np.random.randint(11,99,(3, 3, 3, 64)) # Array with all nonzeros
In [183]: z = np.zeros((3, 3, 1, 64),dtype=a.dtype)
In [184]: out = np.concatenate((a,z),axis=2)
In [185]: (out[:,:,-1,:]==0).all()
Out[185]: True
In [186]: out.shape
Out[186]: (3, 3, 4, 64)
# Another way to verify
In [187]: (out==0).sum()
Out[187]: 576
In [188]: 3*3*64
Out[188]: 576