The Filtered-Guided Diffusion (FGD) is a novel approach developed to guide black-box diffusion processes based on an adaptive filter. The method is designed to improve the control over the strength of guidance in diffusion-based generative models, which are widely used for Image-to-Image translation and editing. The FGD method applies a filter to the input of each diffusion step adaptively, which does not depend on any specific architecture or sampler. This makes it easy to combine with other techniques, samplers, and diffusion architectures.
FGD offers a fast and strong baseline that is competitive with recent architecture-dependent approaches. Furthermore, FGD can also be used as a simple add-on to enhance the structural guidance of other state-of-the-art I2I methods. The method is designed to be architecture-independent, making it highly portable to different architectures and complementary to other guidance strategies. It allows for more continuous adjustment of guidance strength than other approaches, providing a more flexible and adaptable solution for image translation and editing tasks.
Publication date: June 29, 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2306.17141.pdf
