Image_path = 'data/dogs-vs-cats/train/dog/' + chosen_image We then create a variable called image_path and set that to the relative location on disk of the chosen image. Next, we choose a random image of a dog from disk.Ĭhosen_image = random.choice(os.listdir( 'data/dogs-vs-cats/train/dog')) For example, rotation_range is measured in degrees, while width_shift_range is measured as a fraction of the width of the image. Gen = ImageDataGenerator(rotation_range= 10, width_shift_range= 0.1, height_shift_range= 0.1, shear_range= 0.15, zoom_range= 0.1,Ĭhannel_shift_range= 10., horizontal_flip= True)ĭocumentation to understand the units used for each augmentation technique we've specified.

All the parameters being passed are the different ways we're telling Keras to augment the image. We'll now define this variable called gen as an ImageDataGenerator. TensorFlow's documentation to plot the processed images within our Jupyter notebook.įig, axes = plt.subplots( 1, 10, figsize=( 20, 20)) Next, we'll use this plotImages() function obtained from Let's now see how we can perform data augmentation using Keras.įirst, we import all the libraries we'll be using.įrom import ImageDataGenerator We can do that by augmenting our existing data and then adding that data to the training set.Īnother reason to use data augmentation is to reduce overfitting.

Maybe we have a small training set, or maybe we just want to make our training set larger. For image data specifically, data augmentation could consist of things like flipping the image horizontally or vertically, rotating the image, zooming in or out, cropping, or varying the color.įor starters, it will help us obtain more data for training. In our case, the data we'll work with will be images. We'll touch on the concept of data augmentation a bit more before we jump into the code, but for a more thorough presentation of the concept, check out theĭata augmentation episode from the Deep Learning Fundamentals course. In this episode, we'll demonstrate how to use data augmentation on images using TensorFlow's Keras API.ĭata augmentation occurs when new data is created based on modifications of existing data. Performing data augmentation with TensorFlow's Keras API

Keras image augmentation