Image data is classified as an unstructured type of data. One of the popular applications of deep learning networks involves developing image classification and recognition models. Image classification has various applications, such as face recognition on smartphones or on social media networks, classification of medical image data, classification of handwritten digits, and self-driving cars. Note that it is not possible to develop a classification model directly from unstructured data. The unstructured data needs to be first converted into a structured form before deep learning networks can be developed. For example, a black and white image may have dimensions of 21 x 21 and thus contain data on 441 (21 x 21) pixels. Once we convert an image into numbers representing all the pixels, it becomes feasible to develop image classification models. Although humans can classify a type of dress, a person, or certain object very easily, even when the images may have different sizes or orientation, training a computer to do so is a challenging task.

The Keras library provides several easy-to-use features for processing image data that helps in developing deep learning image classification networks. The effectiveness of having deep networks or neural networks with many layers especially comes to the fore when it comes to image recognition and classification problems. In Chapter 4, Image Classification and Recognition, we provide an illustration of applying a deep learning image classification model using R.