Declaring 2D numpy array with unknown size

Question

I m new to numpy. I m trying to define a 2-d numpy array to read images, the size of which varies.So, i cant predefine the size of the array. My code is

np.ndarray(np.float64) I
for i in range(len(filename)):
    I=imread(filename)     //reading an image here
    I1=I.resize(256,256)   //resizing the image

Please suggest the corrections to the code.

Thank you in advance.

Solution

If you don't know the size before-hand, then don't use a numpy array to accumulate the values. Use a flexible container (e.g. a list) and then convert to a numpy array afterwards.

If you're working with something really large, there's also numpy.fromiter which will behave quite a bit more efficiently, but you'll have to jump through a couple of hoops to use it with >1D arrays.

As an example of the first suggestion, let's say we're creating an array that will always have 10 columns, but we have no way of knowing how many rows there are. We'll use a list to store each row, and then convert to a 2D array at the end:

import numpy as np

data = []

random_val = 1
while random_val > 0.05:
    data.append(np.arange(10))
    random_val = np.random.random()

data = np.array(data)
print data.shape

Comments

[Saullo G. P. Castro]

I'm wondering why the downvote...