AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

ShiftDDPMs: Exploring Conditional Diffusion Models by Shifting Diffusion Trajectories

Zijian Zhang, Zhou Zhao, Jun Yu, Qi Tian

AAAI 2023 · Image generation › Fast diffusion sampling · approach: Denoising diffusion probabilistic models · 9 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use Quadratic-Shift instead of conditional reverse-process conditioningCIFAR-10 image generation · CIFAR-10FID: 3.12 -> 3.05Improved
use Quadratic-Shift instead of conditional reverse-process conditioningCIFAR-10 image generation · CIFAR-10IS: 9.59 -> 9.67Improved
use Quadratic-Shift instead of conditional reverse-process conditioningCIFAR-10 image generation · CIFAR-10NLL (bits/dim): ≤ 3.74 -> ≤ 3.69Improved
use Prior-Shift instead of conditional reverse-process conditioningCIFAR-10 image generation · CIFAR-10FID: 3.12 -> 3.06Improved
use Data-Normalization rather than a conditional DDPMCIFAR-10 image generation · CIFAR-10FID: 3.12 -> 5.51Hurt
add class labels to the reverse process (ablation)Quadratic-Shift · CIFAR-10IS: 9.67 -> 9.74Improved
add class labels to the reverse process (ablation)Quadratic-Shift · CIFAR-10FID: 3.05 -> 3.02Improved
add class labels to the reverse process (ablation)Quadratic-Shift · CIFAR-10NLL (bits/dim): ≤ 3.69 -> ≤ 3.70Hurt
use Quadratic-Shift for inpainting rather than DDPM (score)Places2 image inpainting · Places2, 0-20% maskFID: 2.0665 -> 1.8314Improved
use Quadratic-Shift for inpainting rather than DDPM (score)Places2 image inpainting · Places2, 20-40% maskFID: 6.6129 -> 6.2915Improved
use Quadratic-Shift for inpainting rather than DDPM (score)Places2 image inpainting · Places2, 40-60% maskFID: 17.3601 -> 14.9667Improved
use Quadratic-Shift instead of conditional DDPMCUB text-to-image generation · CUBIS: 4.18 -> 4.42Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy