Search papers, labs, and topics across Lattice.
The paper introduces SKILD, a scale-invariant diffusion model that unifies image generation and continuous super-resolution within a single unconditional framework by making scale an explicit coordinate of the diffusion dynamics. SKILD's forward process attenuates image content from fine to coarse scales while injecting spectrum-matched Gaussian noise. The same trained reverse process then performs both generation and continuous super-resolution by varying only the starting timestep, eliminating the need for task-specific architectures or retraining.
Unconditional image diffusion models can now perform continuous super-resolution without task-specific architectures or retraining, simply by varying the starting timestep.
Creating images from noise is image generation; reconstructing fine details from coarse inputs is super-resolution. Despite their practical differences, both can be understood as reversing information loss across scales. We introduce $\textbf{SKILD}$, a $\textbf{S}$cale-invariant $\textbf{K}$-Space $\textbf{I}$mage $\textbf{L}$earning $\textbf{D}$iffusion model that unifies generation and continuous super-resolution within a single unconditional framework. Both natural images and critical physical systems exhibit scale invariance, and we leverage it to design a forward process that attenuates image content from fine to coarse scales while injecting spectrum-matched Gaussian noise, making scale an explicit coordinate of the diffusion dynamics. The same trained reverse process performs generation and continuous super-resolution by varying only the starting timestep: $\textit{no task-specific architecture, no conditioning branch, no classifier-free guidance, no retraining per scale factor}$. Empirically, SKILD reaches FID $2.65$ and Inception Score $9.63$ on unconditional CIFAR-10, performs $2\times$--$8\times$ super-resolution on ImageNet from a single unconditional checkpoint while outperforming conditional models across perceptual metrics, and reconstructs critical Ising models whose connected four-point correlations closely track the ground truth.
MIT CSAIL