ImageNet-128

ImageNet-128 is a benchmark dataset containing a subset of images from the larger ImageNet dataset, used to evaluate methods for solving linear inverse problems without the need for training.

Papers using ImageNet-128 (4)

Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion2024Multistep Consistency Models2024 · 3 citesTraining-free Linear Image Inverses via Flows2023EM Distillation for One-step Diffusion Models2024