load_image_data_numpy.py

#import all dependencies
import PIL
import keras
import numpy as np
from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img


print ("\n")
print("Your backend is ----------------------------------------------------------------->",keras.backend.backend())

#Path of image to augment 
image_file_path = "/train/dog/dog.01.jpg"

# Loading the image 
#install pil library using pip install Pillow
img = load_img(image_file_path) 

#Convert image to 3D array of shape (image height,image width,number of channels)
image_array = img_to_array(img) 
print("Shape of input image before reshape -------------------------------------------------->",image_array.shape)

#reshape the image_array into a size (1,image height,image width,number of channels)
#Here 1 stands for single input image.
x = image_array.reshape((1,) + image_array.shape)
print("The shape of input image after reshape ----------------------------------------------->",x.shape)

print ("\n")
print("1 in the shape, is the number of image to make x array compatible to be used as an input in the image augmenter function where the input size must always be a Numpy array of rank 4 or a tuple , followed by image height and width and then the number of channels,")
print("This data format is applied only if your backend is tensorflow.")

load_image_data_numpy_expand_dim.py

#import all dependencies
import PIL 
from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
import numpy as np
import keras

print ("\n")
print("Your backend is ----------------------------------------------------------------->",keras.backend.backend())

#Path of image to augment 
image_file_path = "/home/arya/workspace/LandT/dog.66.jpg"

# Loading the image 
#install pil library using pip install Pillow
img = load_img(image_file_path) 

#Convert image to 3D array of shape (image height,image width,number of channels)
image_array = img_to_array(img) 
print("Shape of image before reshape -------------------------------------------------->",image_array.shape)

#reshape the image_array into a size (1,image height,image width,number of channels)
#np.expand_dims adds an axis to the data. This is another way of writing the above code.
x = np.expand_dims(image_array,axis = 0)

print("The shape of image after reshape ----------------------------------------------->",x.shape)

print ("\n")
print("1 is the number of image, followed by image height and width and then the number of channels,")
print("This data format is applied only if your backend is tensorflow.")