Spotting outliers using autoencoders

Another great application of autoencoders is outlier detection. The idea behind this use case is that the autoencoder will learn an encoding with a very small error for the most common classes in a dataset, while its ability to reproduce scarcely represented categories (outliers) will be much more error-prone.

With this premise in mind, in this recipe, we'll rely on a convolutional autoencoder to detect outliers in a subsample of Fashion-MNIST.

Let's begin!

Getting ready

To install OpenCV, use the following pip command:

$> pip install opencv-contrib-python

We'll rely on TensorFlow's built-in convenience functions to load the Fashion-MNIST dataset.

How to do it…

Follow these steps to complete this recipe:

1. Import the required packages:

   import cv2
   import numpy as np
   from sklearn.model_selection import train_test_split
   from tensorflow.keras import Model
   from tensorflow.keras.datasets import...