A popular example of image classification is the MNIST dataset, which contains digits from 0 to 9 in different styles. Here, we'll use a drop-in replacement, called Fashion-MNIST, consisting of different pieces of clothing.
Fashion-MNIST is a dataset of Zalando's article images consisting of a training set of 60,000 examples and a test set of 10,000 examples: https://github.com/zalandoresearch/fashion-mnist.
Here are a few examples from the dataset:
In this recipe, we'll recognize clothing items with different models – we'll start with generic image features (Difference of Gaussians, or DoG) and a support vector machine; then we'll move on to a feedforward Multilayer Perceptron (MLP); then we'll use a Convolutional Neural Network (ConvNet); and finally, we'll look at transfer learning with MobileNet.
Before we can start, we have to install a library. In this recipe, we'll use scikit-image, a library...