Teamwork: Collaborative Diffusion with Low-rank Coordination and Adaptation
2510.05532v1
cs.CV, cs.GR, cs.LG
2025-10-09
Авторы:
Sam Sartor, Pieter Peers
Abstract
Large pretrained diffusion models can provide strong priors beneficial for
many graphics applications. However, generative applications such as neural
rendering and inverse methods such as SVBRDF estimation and intrinsic image
decomposition require additional input or output channels. Current solutions
for channel expansion are often application specific and these solutions can be
difficult to adapt to different diffusion models or new tasks. This paper
introduces Teamwork: a flexible and efficient unified solution for jointly
increasing the number of input and output channels as well as adapting a
pretrained diffusion model to new tasks. Teamwork achieves channel expansion
without altering the pretrained diffusion model architecture by coordinating
and adapting multiple instances of the base diffusion model (\ie, teammates).
We employ a novel variation of Low Rank-Adaptation (LoRA) to jointly address
both adaptation and coordination between the different teammates. Furthermore
Teamwork supports dynamic (de)activation of teammates. We demonstrate the
flexibility and efficiency of Teamwork on a variety of generative and inverse
graphics tasks such as inpainting, single image SVBRDF estimation, intrinsic
decomposition, neural shading, and intrinsic image synthesis.
Ссылки и действия
Дополнительные ресурсы: