I am new to numpy and trying to understand how multi-dimensions work.
I have 300 numpy arrays of dimension (280, 190, 3). When I append all these arrays into a list and convert it into a numpy array (I think it's here where I am doing something wrong) I expect its shape to be (300, 280, 190, 3) but all I get is (300, ) as if it is a 1D array.
Could you please tell me what is my wrong step? Any additional information welcome. Thanks.
To illustrate the 1d ragged warning case:
In [461]: alist = [np.ones((10,5,3)), np.ones((10,5,3)), np.ones((9,5,3))]
In [463]: np.array(alist).shape
/usr/local/bin/ipython3:1: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray
#!/usr/bin/python3
Out[463]: (3,)
stack in the same case produces an error:
In [464]: np.stack(alist)
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-464-724d9c1d0554> in <module>
----> 1 np.stack(alist)
<__array_function__ internals> in stack(*args, **kwargs)
/usr/local/lib/python3.6/dist-packages/numpy/core/shape_base.py in stack(arrays, axis, out)
425 shapes = {arr.shape for arr in arrays}
426 if len(shapes) != 1:
--> 427 raise ValueError('all input arrays must have the same shape')
428
429 result_ndim = arrays[0].ndim + 1
ValueError: all input arrays must have the same shape
If shapes all match:
In [465]: alist = [np.ones((10,5,3)), np.ones((10,5,3)), np.ones((10,5,3))]
In [466]: np.array(alist).shape
Out[466]: (3, 10, 5, 3)
In [467]: np.stack(alist).shape
Out[467]: (3, 10, 5, 3)
To make a 1d object array from a list of equal size arrays, we have to do some extra work:
In [468]: arr = np.empty(3,object)
In [469]: arr[:] = alist
In [470]: arr.shape
Out[470]: (3,)
In [471]: np.array(arr).shape # np.array doesn't change this
Out[471]: (3,)
In [472]: np.stack(arr).shape # stack still works
Out[472]: (3, 10, 5, 3)
Here is an example:
data1=np.random.random(size=(280, 190, 3))
data2=np.random.random(size=(280, 190, 3))
data1.shape, data2.shape
((280, 190, 3), (280, 190, 3))
np.stack([data1,data2], axis=0).shape
(2, 280, 190, 3)
1.19 ragged warning, I also get an error message from np.stack about shapes that don't match.Use the numpy stack to stack arrays separately. Documentation: https://numpy.org/doc/stable/reference/generated/numpy.stack.html
dtype?objectis another indicator of ragged inputs. The newest version, 1.19, is now give a warning if this happens.VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray return array(a, dtype, copy=False, order=order)so i turned my conversion into thisnp.asarray(my_list, dtype=object)and the warning disappeared but the problem persists.