Load Images in Tensorflow – Python

In this article, we are going to see how to load images in TensorFlow in Python.

Loading Images in Tensorflow 

For loading Images Using Tenserflow, we use tf.keras.utils.load_img function, which loads the image from a particular provided path in PIL Format. PIL is a Python Imaging Library that gives your Python interpreter access to image processing functions. This library offers a wide range of file format compatibility, a productive internal representation, and somewhat potent image processing features.

The primary image library is built for quick access to information held in a few basic pixel formats. It ought to serve as a strong starting point for a broad image processing tool.

tf.keras.utils.load_img

We can set various parameters in tf.keras.utils.load_img function for loading an Image.

path: Path of the required Image

grayscale: Set true if want to load an Image in grayscale format.

color_mode: Sets various color modes while loading images. By default RGB.

target_size: For loading an image in the required target size. Size format: (Image_height, Image_width)

interpolation: Set for required Interpolation. By default ‘nearest’.

keep_aspect_ratio: Boolean, indicating whether or not to resize photos without distorting their aspect ratio. 

Before resizing, the image is cropped in the middle to the desired aspect ratio.

Example 1: Load an image in Tensorflow

Python3

# Loading the required libraries
import tensorflow as tf

# Defining the path to the image
# which is to be loaded.
# The image_path shall be changed
# according to the requirement.
image_path = '/content/model_3.png'

# Calling the load_img function
# from the imported libraries.
# This return an image in PIL
# Format
image_loaded = tf.keras.utils.load_img(image_path)

# Printing the obtained image.
image_loaded

Output:

Example 2: Loading Images in Grayscale Format

The steps for loading an Image in grayscale are the same as that mentioned above, Just while loading an image we need to set the parameter grayscale = True.

Python3

# Loading the required libraries
import tensorflow as tf

# Defining the path to the image
# which is to be loaded.
image_path = '/content/model_3.png'

# Calling the load_img function from
# the imported libraries. This
# return an image in PIL
# Format
image_loaded = tf.keras.utils.load_img(image_path,
                                       grayscale=True)

# Printing the obtained image.
image_loaded

Output:

Example 3: Loading Images with a different target size

In this case, we will load our Image in different target size.

Python3

# Loading the required libraries
import tensorflow as tf

# Defining the path to the image
# which is to be loaded.
image_path = '/content/model_3.png'

# Calling the load_img function from the
# imported libraries. This return an
# image in PIL Format
image_loaded_1 = tf.keras.utils.load_img(image_path)

# Specifying size for the second Image
# Target size = (200,300)
image_loaded_2 = tf.keras.utils.load_img(image_path,
                                         target_size=(200,
                                                      300))

# Printing the obtained image.
print("Size of Image 1: ", image_loaded_1.size)
print("Size of Image 2: ", image_loaded_2.size)

image_loaded_2.save('/content/out.png')

Output: