Integrating Categorical Semantics into Unsupervised Domain Translation

Samuel Lavoie-Marchildon, Faruk Ahmed, Aaron Courville (privately revealed to you)

Keywords: Unsupervised Domain Translation, Unsupervised Learning, Image-to-Image Translation, Deep Learning, Representation Learning

Abstract: While unsupervised domain translation (UDT) has seen a lot of successes recently, we argue that allowing its translation to be mediated via categorical semantic features could enable wider applicability. In particular, we argue that categorical semantics are important when translating between domains with multiple object categories possessing distinctive styles, or even between domains that are simply too different but still share high-level semantics. We propose a method to learn, in an unsupervised manner, categorical semantic features (such as object labels) invariantly of the source and the target domains. We show that conditioning the style of a unsupervised domain translation methods on the learned categorical semantics leads to a considerably better high-level features preservation on tasks such as MNIST$\leftrightarrow$SVHN and to a more realistic stylization on Sketches$\to$Reals.

One-sentence Summary: We present a method for learning domain invariant categorial semantics which enable UDT on two setups.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics