Autoencoding beyond pixels using a learned similarity metric

Journal: arXiv preprint arXiv:1512.09300

We present an autoencoder that leverages learned representations to better measure similarities in data space. 

By combining a variational autoencoder with a generative adversarial network we can use learned feature representations in the GAN discriminator as basis for the VAE reconstruction objective.

Thereby, we replace element-wise errors with feature-wise errors to better capture the data distribution while offering invariance towards eg translation. We apply our method to images of faces and show that it outperforms VAEs with element-wise similarity measures in terms of visual fidelity.

Moreover, we show that the method learns an embedding in which high-level abstract visual features (eg wearing glasses) can be modified using simple arithmetic.

Download