For a specific project on the housing market (here), I had to analyze thousands of photos. To do that, I used a convolutional neural network (CNN), which is a fancy name for a complicated function that can be “trained” to recognize patterns in images. In this blog post, I would like to introduce the “Hello World” of computer vision and CNN: the classification of hand-written digits from the MNIST dataset. There are thousands of tutorials on the same topic using Python freely available on the Internet. Instead, let’s use Julia and the package Flux.jl. Why? Because Julia is fast, and if you have millions of images to analyze, the speed up could be substantial compared to Python. The Jupyter notebook used to generate this post can be found here.