Exploring the Design Space of Diffusion Bridge Models
This work advances image-to-image translation by improving diffusion bridge models, though it appears incremental as it builds on existing Stochastic Interpolants.
The authors unified and expanded diffusion bridge models for image-to-image translation by extending Stochastic Interpolants with preconditioning, endpoint conditioning, and optimized sampling, achieving state-of-the-art performance in image quality and sampling efficiency across diverse tasks while addressing low sample diversity issues.
Diffusion bridge models and stochastic interpolants enable high-quality image-to-image (I2I) translation by creating paths between distributions in pixel space. However, the proliferation of techniques based on incompatible mathematical assumptions have impeded progress. In this work, we unify and expand the space of bridge models by extending Stochastic Interpolants (SIs) with preconditioning, endpoint conditioning, and an optimized sampling algorithm. These enhancements expand the design space of diffusion bridge models, leading to state-of-the-art performance in both image quality and sampling efficiency across diverse I2I tasks. Furthermore, we identify and address a previously overlooked issue of low sample diversity under fixed conditions. We introduce a quantitative analysis for output diversity and demonstrate how we can modify the base distribution for further improvements.