[FIXED] Change the dimension of an array

Issue

I have a number of images that have shapes (10,1134,1135). I am trying to change the shape to (10,1134,1134). I converted the image into NumPy and use the array. reshape but I get an error saying cannot reshape the array of size 12870900 into shape (10,1134,1134). Is there an alternate way to do this?

Solution

Since you need to shrink the size of your array you need a to drop a vector on some axis of data. There are a few ways to do this through slicing.

Example Array:

arr = np.zeros((10,1134,1135))
np.shape(arr)

#output
(10, 1134, 1135)

---

Drop First:

new_arr=arr[:,:,1:]
np.shape(new_arr)

#output
(10, 1134, 1134)

---

Drop Last:

new_arr=arr[:,:,0:1134]
np.shape(new_arr)

#output
(10, 1134, 1134)

---

Drop Chosen:

arr_not=list(range(0, np.shape(arr)[2]))
arr_not.remove(9) #choose by index

new_arr=arr[:,:,arr_not]
np.shape(new_arr)

#output
(10, 1134, 1134)

You can also use np.delete() as follows:

new_arr=np.delete(arr, [134], 2) #choose by index
np.shape(new_arr)

#output
(10, 1134, 1134)

np.delete() documentation

Answered By - j__carlson