Image Data Preprocessing

In this section, we go over a few techniques that you can use as a data scientist to preprocess images. First, we look at image normalization, and then we learn how we can convert a color image into a greyscale image. Finally, we look at ways in which we can bring all images in a dataset to the same dimensions. Preprocessing images is needed because datasets do not contain images that are the same size; we need to convert them into a standard size to train machine learning models on them. Some image preprocessing techniques help by reducing the model's training time by either making the important features easier to identify for the model or by reducing the dimensions as in the case of a greyscale image.

Normalization

In the case of images, the scale of the pixels is of the same order and in the range 0 to 255. Therefore, this normalization step is optional, but it might help speed up the learning process. To reiterate, centering the data and scaling it to the...