TraVeLGAN: Image-to-image Translation by Transformation Vector Learning
This addresses the limitation of existing models to homogeneous domains, potentially improving translation for real-world applications like medical imaging or autonomous driving, though it appears incremental as it builds on GAN frameworks.
The paper tackles image-to-image translation for complex domains with high-level shape and context differences, introducing TraVeLGAN which uses a siamese network to preserve intra-domain transformations, eliminating the need for cycle-consistency and enabling mappings between heterogeneous domains.
Interest in image-to-image translation has grown substantially in recent years with the success of unsupervised models based on the cycle-consistency assumption. The achievements of these models have been limited to a particular subset of domains where this assumption yields good results, namely homogeneous domains that are characterized by style or texture differences. We tackle the challenging problem of image-to-image translation where the domains are defined by high-level shapes and contexts, as well as including significant clutter and heterogeneity. For this purpose, we introduce a novel GAN based on preserving intra-domain vector transformations in a latent space learned by a siamese network. The traditional GAN system introduced a discriminator network to guide the generator into generating images in the target domain. To this two-network system we add a third: a siamese network that guides the generator so that each original image shares semantics with its generated version. With this new three-network system, we no longer need to constrain the generators with the ubiquitous cycle-consistency restraint. As a result, the generators can learn mappings between more complex domains that differ from each other by large differences - not just style or texture.