SD-GAN Art Demo

Chris Donahue, Julian McAuley

This is an interactive demo for the SD-DCGAN model from Disentangled representations of style and content for visual art with generative adversarial networks. This demo runs a neural network in your browser; it may take some time to render.

SD-GANs learn a latent representation that disentangles artistic style from the depicted content. They can generate images that appear to depict the same artistic style but with varying content. Furthermore, they can generate images that depict the same content in a variety of styles. Both style and content representations are imagined; no input images are required.

Instructions

The left matrix depicts all combinations between four distinct styles (rows) and four distinct contents (columns). The right matrix depicts a linear interpolation of both the style and content vectors for a pair of images selected from the left matrix.

Press the buttons on the rows/columns in the left matrix to randomize style/content.

Select two images in the left matrix to see their interpolation in the right.

Content
A
Content
B
Style
A
Style
B