Research map / Label-efficient and robust learning / Noisy labels
Robust Equivariant Imaging: A Fully Unsupervised Framework for Learning To Image From Noisy and Partial Measurements
Check 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 |
|---|---|---|---|
| Replace EI's noisy measurement-consistency loss with SURE-based consistency and use robust equivariance | Equivariant Imaging · 4× accelerated MRI, σ=0.1–0.2 | PSNR: 3.6 dB at σ=0.1; 7 dB at σ=0.2 | Helped |
| Use REI instead of EI | Equivariant Imaging · Inpainting, Poisson noise | PSNR: more than 8 dB at high noise levels | Helped |
| Use REI instead of EI | Equivariant Imaging · Sparse-view CT, MPG noise | PSNR: 5 dB gain | Helped |
| Use REI instead of EI | Equivariant Imaging · Low-dose sparse-view CT, MPG noise | PSNR: 9 dB gain | Helped |
| Use the full REI loss rather than EI₁ (MC + robust equivariance) or EI₂ (SURE + non-robust equivariance) (ablation) | REI · 4× accelerated MRI | PSNR | Helped |
| Use noisy-input robust equivariance instead of the non-robust equivariance loss (ablation) | REI · 4× accelerated MRI | PSNR | Helped |
| Replace measurement consistency (MC) with SURE measurement consistency (ablation) | REI · 4× accelerated MRI | PSNR | Helped |
| Remove equivariance and use SURE alone (ablation) | REI · 4× accelerated MRI | PSNR | Hurt |
| Set equivariance strength α to 0.01 rather than 1 (ablation) | REI · Inpainting, γ=0.01–0.1 | PSNR: 21.0 -> 19.5; 18.2 -> 17.0; 16.6 -> 11.2 | Hurt |
| Set equivariance strength α to 0.1 rather than 1 (ablation) | REI · Inpainting, γ=0.01–0.1 | PSNR: 21.0 -> 19.66; 18.2 -> 18.1; 16.6 -> 16.0 | Mixed |
| Increase equivariance strength α to 10 or 100 rather than 1 (ablation) | REI · Inpainting, γ=0.01–0.1 | PSNR: α=10: 16.2, 13.5, 12.3; α=100: 10.6, 7.4, 8.0 | Hurt |
| Set finite-difference scale τ to 10⁻¹ rather than 10⁻² (ablation) | REI · 4× accelerated MRI, σ=0.1 | PSNR: 27.7 -> 24.6 | Hurt |
Closest papers by meaning
- A Regularization-Guided Equivariant Approach for Image Restoration (CVPR 2025 · similarity 0.88)
- Equivariant Plug-and-Play Image Reconstruction (CVPR 2024 · similarity 0.88)
- Rotation-Equivariant Self-Supervised Method in Image Denoising (CVPR 2025 · similarity 0.86)
- RESfM: Robust Deep Equivariant Structure from Motion (ICLR 2025 · similarity 0.86)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · similarity 0.85)
- Equivariant Multi-Modality Image Fusion (CVPR 2024 · similarity 0.85)
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.