# Reading .png image into a numerical vector

To be able to work with images reasonably (learn from them and classify them), we shall transform each image into a numerical vector. Each number in the vector will correspond to the intensity of certain pixel in the image.

## TLDR

With the use of Pillow and Numpy libraries, the whole thing is easy:

from PIL import Image
import numpy as np

impath = 'train_data/img_1112.png'
image_vector = np.array(Image.open(impath)).flatten()

## Explanation

If you are interested in what the above code actually does, let's do it step by step:

from PIL import Image
import numpy as np

impath = 'train_data/img_1112.png'
im = Image.open(impath)
print(type(im))
im2d = np.array(im)
print(type(im2d), im2d.shape)
im1d = im2d.flatten()
print(type(im1d), im1d.shape)
print(im1d)
After running the above code, you should get output similar to this:
<class 'PIL.PngImagePlugin.PngImageFile'>
<class 'numpy.ndarray'> (10, 10)
<class 'numpy.ndarray'> (100,)
[230 202 168 139 124 129 147 180 206 221 227 181 126  84  51  49  80 145
206 241 227 169 102  50  18   7  27  96 183 233 212 156  92  40  21  15
25  76 164 224 196 136  78  47  33  32  39  73 141 203 175 118  64  41
42  45  48  66 122 184 151  87  39  24  35  34  28  35  84 156 139  76
36  26  35  37  30  38  69 130 152 106  87  99 116 114  95  77  82 122
198 180 186 217 237 226 199 168 152 172]
Explanation:

• Function Image.open() reads in the image and returns an instance of class PngImageFile.
• Function np.array() converts the image into an instance of class numpy.ndarray, i.e., it creates a 2D array of size 10×10.
• Method ndarray.flatten() turns 10×10 matrix into a vector of length 100; its contents are at the end of the output.