A simple implementation of the Pix2Pix by Isola et al. paper, adapted to run in the browser using TensorFlow.js. It uses machine learning, namely, conditional adversarial networks, to create the pictures. The code runs in real time after you draw some edges, and you can switch between generating images of cats, facades or shoes.
Run the model via a laptop as mobile devices cannot handle the current models.
Check out the demo, and use the mouse to draw in the left hand box, and see the results on the right.