Exciting news! Gradient has launched a FREE GPU plan. Read More
Project Details

Image Compression Using Autoencoders in Keras

Compress and reconstruct images

The Gradient Team


Deep learning-based image compression techniques are a popular topic of current research, so much so that The Joint Photographic Experts Group (JPEG) committee has recently called for evidence on these techniques as of February 2020.

Autoencoders are a deep learning model for transforming data from a high-dimensional space to a lower-dimensional space, and are a commonly referenced model for image compression. In this tutorial we'll see a basic example of how to apply autoencoders to compress images from the MNIST dataset using TensorFlow and Keras. The notebook references a corresponding blog post which dives into the theory behind this.