(Please bear with me, I’m still a beginner in deep learning and this is my first upload to GitHub, I don’t know the way of the hub)
After doing a lot of mini projects, I realized the need to have a generalized input pipeline for machine learning models involving images. Further, I wanted to streamline data augmentation by creating certain common and easy to use image transformations.
This library is my solution to the problems posed above. I acknowledge that there will many bugs in the code provided. Also, the code may not have followed the best design philosophies. I welcome all changes to my repo. The class ImageReader serves to provide all of image reading and data augmentation functionalities in minimal lines of code.
Here is an example to show just how easy it is:

tfms = [Horizontal_flip(), Vertical_flip(), Gaussian_blur(3),
       Crop_and_resize(do_crop=False, sz=(300, 300))]
imr = ImageReader(PATH_TO_IMAGES, transforms=tfms)
image = imr.read_image_random()

Here, OpenCV is primarily used to read images and perform most transformation operations.

The file aug_transforms.py includes certain common transformations used in computer vision and their appropriate documentation is included in the functions.
The file Transformer.py provides a Transformer class that is used by the ImageReader class. It can be used to create objects that transform images read as numpy arrays. You can pass any number of your own transformations that behave as specified in the Transformer class’ documentation.

A typical way to use this library would be as follows:

from ImageReader import *

tfms = [Horizontal_flip(), Vertical_flip(), Gaussian_blur(3),
        Crop_and_resize(do_crop=False, sz=(300, 300))]
imr = ImageReader(PATH_TO_IMAGES, transforms=tfms)
image = imr.read_image_random()

Dependencies:
OpenCV, skimage, Numpy, Matplotlib, os.

Use as you please :-)