Train a VAE model to generate images.
Description
This example trains a variational autoencoder to generate images that resemble those found in the fashion MNIST dataset.
This example trains the model in Node.js. See README.md for instructions. This page contains a viewer for the trained model.
This examples assumes that the latest space of the autoencoder is 2d.
Instructions
You can adjust the number of images in the grid as well as the start and end values used to generate each image. Experiment with them see the range of what the autoencoder can generate.
Controls
Results