One-Shot Unsupervised Cross Domain Translation
This addresses a key AI capability for cognitive agents to act in the world by enabling domain translation with minimal data, though it is incremental as it builds on existing unsupervised methods.
The paper tackles the problem of one-shot unsupervised cross-domain translation, where only a single image from the source domain is available, and shows that existing methods fail at this task. Their method, which adapts a variational autoencoder from the target domain, achieves performance comparable to existing methods that use multiple training samples from the source domain.
Given a single image x from domain A and a set of images from domain B, our task is to generate the analogous of x in B. We argue that this task could be a key AI capability that underlines the ability of cognitive agents to act in the world and present empirical evidence that the existing unsupervised domain translation methods fail on this task. Our method follows a two step process. First, a variational autoencoder for domain B is trained. Then, given the new sample x, we create a variational autoencoder for domain A by adapting the layers that are close to the image in order to directly fit x, and only indirectly adapt the other layers. Our experiments indicate that the new method does as well, when trained on one sample x, as the existing domain transfer methods, when these enjoy a multitude of training samples from domain A. Our code is made publicly available at https://github.com/sagiebenaim/OneShotTranslation