Abstract

Image classification can have many practical, valuable and life-saving benefits. The “Hello World” of image classification is often considered MNIST and, more recently, Fashion MNIST. This article will use Fashion MNIST and store the images in a SingleStore DB database. We’ll also build an image classification model using Keras and Tensorflow and store the prediction results in SingleStore DB. Finally, we’ll build a quick visual front-end to our database system using Streamlit that enables us to retrieve an image and determine if the model correctly identified it.

The SQL scripts, Python code and notebook files used in this article are available on GitHub. The notebook files are available in DBC, HTML and iPython formats.