I have a numpy array of shape (6, 100, 161). I want to convert it to (600, 161) and then trim 16 from the end, so it's (584, 161)?
Add a comment
|
2 Answers
You can call reshape followed by slicing
import numpy as np
a = np.zeros((6,100,161))
b = a.reshape(600,161)[:-16,:]
b.shape
out: (584,161)
Comments
You can use np.reshape:
>>> import numpy as np
>>> x = np.ones((6,100,161))
>>> x = x.reshape((-1, 161))
>>> x = x[:584, :]
>>> x.shape
(584, 161)
Here -1 works like this:
If the original shape was (x, y, z)
then if you write reshape(-1, z)
It infers that you want to dimensions, as you have given two numbers -1 and z, but you have specified size for only 1 dimension (i.e. z), so it infers the other size must be all that remains.
So total number of elements x * y * z, now after reshaping the number of elements must be same. You have specified you want z number of columns, so the number of rows must be x * y * z / z == x * y. -1 facilitates this.
For example:
>>> x = np.ones((3,4,3))
# number of elements = 3 * 4 * 3 == 36
>>> y = x.reshape(9, -1)
# you say you want 9 rows, so remaining elements
# would be, 36 / 9 == 4, so no. of cols == 4
>>> y.shape
(9, 4)
