Now, as we just finished learning some simple examples of using numpy array's reshape() function, let us now learn a more complex use of reshape() function.
For this, we will load a colored image, convert it into a grayscale image, and then will apply to reshape() function on this grayscale image.
Please follow the below steps:
(1) Please import the required libraries
import numpy as np
from sklearn.datasets import load_sample_image
(2) Let us use the load_sample_images function to load china.jpg image and store the loaded image in a variable called china
china = load_sample_image("china.jpg")
(3) Please check the shape of china array
china.<<your code comes here>>
the shape of china array is (427, 640, 3)
Let us see the structure of this china array and then we will try to understand each value in this shape.
china is a 3-dimensional numpy array, and it looks something like below
[
[ [174, 201, 231], [250, 251, 255],...........] - row 0
[ [172, 199, 229], [251, 252, 255],...........] - row 1
.
.
]
Hence, in china array, we see that
[') represents rows. Like this, we have 427 rows.[174, 201, 231] in the row represents a pixel). Like this we have 640 pixels in each row.174, 201, 231] ) represent colour codes for the colour of this pixel. e.g. 174, 201 and 231 combinedly may be representing a dark brown colour (colour of this pixel of the china image).
If you see the values above (marked in bold font), you will know what each value of the shape (427, 640, 3) represents in the china array.
(4) Just for simplifying the things, let us extract a small portion of this china numpy array and store it in a variable called image.
Here, we are extracting (slicing) a portion (image) of china array from row number 150 to 220 and column number 130 to 250.
image = china[150:220, 130:250]
Now, let us check the shape of this image array by using the below code
image.<<your code comes here>>
(5) Please store each value of shape 'tuple' separately in variables - height, width and channels
height, width, channels = image.shape
Here, channels are nothing but the number of colour codes i.e. 3.
(6) Let us convert the coloured image into a grayscale image (image_grayscale), by replacing the values which are along the axis=2, by their mean value.
Do you remember the concept of axis explained earlier?
axis=0 represents the first dimension of an array, i.e. it represents the rowaxis=1 represents the second dimension of an array, i.e. it represents the columnaxis=2 represents the third dimension of an array, in this case, the third dimension is channels.
channels (axis=2) has 3 values (e.g. [68, 39, 44] ), and, mean of these 3 values is 50.33
Hence, value [68, 39, 44] will be replaced with value 50.33 after the step - image.mean(axis=2)
image_grayscale = image.mean(axis=2).astype(np.float32)
Now, let us check the shape of this image_grayscale array by using the below code
image_grayscale.<<your code comes here>>
(7) Now, let us reshape this image_grayscale array into a 4-dimensional array (from existing 2-dimensions) and store the output in a variable called images.
<<your code comes here>> = image_grayscale.reshape(1, height, width, 1)
(8) Please check dimensions of images array
images.<<your code comes here>>
Loading comments...