Research map / Image generation / Autoregressive and latent generation
SILO: Solving Inverse Problems with Latent Operators
Ron Raphaeli, Sean Man, Michael Elad
Read the paperCheck your idea against 137,217 papers
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add HQ image-specific prompt and CFG 4 (vs generic prompt, CFG 1) (ablation) | SILO (RV) · FFHQ SR x8 | FID: 27.10 -> 24.43 | Improved |
| add HQ image-specific prompt and CFG 4 (vs generic prompt, CFG 1) (ablation) | SILO (SD) · FFHQ SR x8 | FID: 30.71 -> 29.29 | Improved |
| use generic prompt instead of null prompt (ablation) | SILO (RV) · FFHQ SR x8 | FID: 27.63 -> 27.10 | Improved |
| use generic prompt instead of null prompt (ablation) | SILO (SD) · FFHQ SR x8 | FID: 32.55 -> 30.71 | Improved |
| use CFG 4 with generic prompt instead of CFG 1 (ablation) | SILO (RV) · FFHQ SR x8 | FID: 27.10 -> 26.21 | Improved |
| swap Realistic Vision v5.1 prior for SD v1.5 (ablation) | SILO · FFHQ SR x8 | FID: 30.71 -> 27.10 | Improved |
| replace Readout Guidance operator with time-independent CNN (ablation) | SILO (RV) · FFHQ SR x8 | FID: 27.10 -> 26.88 | Improved |
| replace Readout Guidance operator with time-independent CNN (ablation) | SILO (RV) · FFHQ SR x8 | runtime (sec): 149 -> 109 | Improved |
| replace pixel-space decoder guidance (PSLD) with latent operator | SILO (RV) · FFHQ SR x8 | FID: 41.58 -> 27.10 | Improved |
| replace ReSample with latent operator | SILO (RV) · FFHQ SR x8 | runtime (sec): 1418 -> 149 | Improved |
| increase measurement noise from 0.01 to 0.03 | ReSample · FFHQ SR x8 | FID: 131.75 -> 235.8 | Hurt |
| increase measurement noise from 0.01 to 0.03 | SILO (RV) · FFHQ SR x8 | FID: 27.10 -> 30.28 | Mixed |
Closest papers by meaning
- Prompt-tuning Latent Diffusion Models for Inverse Problems (ICML 2024 · similarity 0.90)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.90)
- Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models (ICML 2024 · similarity 0.90)
- Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency (ICLR 2024 · similarity 0.90)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.90)
- A Diffusion Model with State Estimation for Degradation-Blind Inverse Imaging (AAAI 2024 · similarity 0.89)
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.
