Autoencoding beyond pixels using a learned similarity metric

Code

• github.com/clementchadebec/benchmark_VAE
  pytorch★ 1,983
• github.com/pravn/vaegan
  pytorch★ 10
• github.com/oadonca/ANVAE
  tf★ 6
• github.com/AlexanderBogatko/TensorFlow_Keras_VAEGAN
  tf★ 0
• github.com/manicman1999/Sword-GAN32
  none★ 0
• github.com/nguyenquangduc2000/AttGAN
  tf★ 0
• github.com/NanYi-hub/Deep-Generative-Network
  none★ 0
• github.com/seangal/dcgan_vae_pytorch
  pytorch★ 0
• github.com/andersbll/autoencoding_beyond_pixels
  torch★ 0
• github.com/gm3g11/VAE_GAN_pytorch
  pytorch★ 0

Abstract

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 e.g. 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 (e.g. wearing glasses) can be modified using simple arithmetic.

Tasks

Reproductions