Research map / 3D vision / 3D generation
Improving 3D Imaging with Pre-Trained Perpendicular 2D Diffusion Models
Suhyeon Lee, Hyungjin Chung, Minyoung Park, Jonghyuk Park, Wi-Sun Ryu, Jong Chul Ye
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 auxiliary perpendicular 2D diffusion model (TPDM) versus 2D DPS alone | TPDM · MR-ZSR 5mm to 1mm (x5) | PSNR: 34.77 -> 35.97 | Improved |
| use sqrt(M) normalization in the forward kernel instead of mean (1/M) (ablation) | TPDM · MR-ZSR 5mm to 1mm (x5) | PSNR: 32.84 -> 35.97 | Improved |
| use DPS instead of MCG for measurement consistency in TPDM (ablation) | TPDM · MR-ZSR 5mm to 1mm (x5) | PSNR: 34.48 -> 35.97 | Improved |
| replace TV prior (DiffusionMBIR) with auxiliary perpendicular diffusion prior | 3D inverse solver · MR-ZSR | PSNR: N/W -> 35.97 | Improved |
| TPDM versus DiffusionMBIR on strong CS-MRI x48 | TPDM · CS-MRI Poisson x48 | PSNR: 34.83 -> 37.17 | Improved |
| TPDM versus DPS on easy CS-MRI x8 | TPDM · CS-MRI Poisson x8 | PSNR: 47.10 -> 44.96 | Hurt |
| TPDM versus DPS on x24 CS-MRI | TPDM · CS-MRI Poisson x24 | PSNR: 39.06 -> 40.34 | Improved |
| TPDM versus DPS on SV-CT with very small training set | TPDM · SV-CT 36-view | PSNR: 38.20 -> 38.25 | No clear effect |
| TPDM versus FBPConvNet on axial SSIM | TPDM · SV-CT 36-view | SSIM (axial): 0.945 -> 0.947 | No clear effect |
Closest papers by meaning
- DiffusionBlend: Learning 3D Image Prior through Position-aware Diffusion Score Blending for 3D Computed Tomography Reconstruction (NeurIPS 2024 · similarity 0.89)
- Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency (ICLR 2024 · similarity 0.87)
- Prompt-tuning Latent Diffusion Models for Inverse Problems (ICML 2024 · similarity 0.87)
- Ambient Diffusion Posterior Sampling: Solving Inverse Problems with Diffusion Models Trained on Corrupted Data (ICLR 2025 · similarity 0.87)
- Inference-Time Search Using Side Information for Diffusion-Based Image Reconstruction (NeurIPS 2026 · similarity 0.87)
- Learning Image Priors Through Patch-Based Diffusion Models for Solving Inverse Problems (NeurIPS 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.
