Research map / Image generation / Fast diffusion sampling
Blurring Diffusion Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add frequency-dependent blur to denoising diffusion | Denoising Diffusion · CIFAR10 | FID: 3.58 -> 3.17 | Helped |
| Add frequency-dependent blur to denoising diffusion | Denoising Diffusion · LSUN Churches 128×128 | FID: 4.68 -> 3.88 | Helped |
| Add frequency-dependent blur instead of inverse heat dissipation | IHDM · CIFAR10 | FID: 18.96 -> 3.17 | Helped |
| Add frequency-dependent blur instead of inverse heat dissipation | IHDM · LSUN Churches 128×128 | FID: 45.1 -> 3.88 | Helped |
| Increase maximum blur from 0 to 20 (ablation) | Blurring Diffusion · CIFAR10 | FID: 3.60 -> 3.17 | Helped |
| Increase maximum blur from 0 to 10 (ablation) | Blurring Diffusion · LSUN Churches 128×128 | FID: 4.68 -> 3.65 | Helped |
| Increase maximum blur from 10 to 20 (ablation) | Blurring Diffusion · LSUN Churches 128×128 | FID: 3.65 -> 3.88 | Hurt |
| Use a sin² rather than sin blur schedule at maximum blur 20 (ablation) | Blurring Diffusion · CIFAR10 | FID: 6.54 -> 3.17 | Helped |
| Replace epsilon prediction with direct x prediction (ablation) | Blurring Diffusion · CIFAR10 | FID: 3.17 -> 23.9 | Hurt |
| Train the UNet directly in frequency space rather than pixel space (ablation) | Blurring Diffusion · CIFAR10 | sample quality | Hurt |
| Add blur with σB,max = 20 instead of no blur (ablation) | Blurring Diffusion · LSUN Churches 64×64 | FID train: 2.70 -> 2.62 | Helped |
Closest papers by meaning
- Warm Diffusion: Recipe for Blur-Noise Mixture Diffusion Models (ICLR 2025 · similarity 0.91)
- TRIM: Scalable 3D Gaussian Diffusion Inference with Temporal and Spatial Trimming (NeurIPS 2025 · similarity 0.87)
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.87)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.87)
- Diffusion Bridge Implicit Models (ICLR 2025 · similarity 0.87)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.87)
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.