Research map / Image generation / Super-resolution
DRIFT: Difficulty-aware Rectified Flows for Through-plane MRI Super-Resolution
Yoonseok Choi, Eun-Gyu Ha, Daniel Kim, Mohammed Al-masni, Ming-Hsuan Yang, Dong-Hyun Kim
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 CETA loss (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | PSNR: 36.92 -> 40.85 | Improved |
| add CETA loss (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | LPIPS: 0.0308 -> 0.0229 | Improved |
| add slice-thickness conditioning (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | PSNR: 34.82 -> 40.85 | Improved |
| add slice-thickness conditioning (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | SSIM: 0.8986 -> 0.9634 | Improved |
| add U-shaped (endpoint-biased) timestep sampling (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | PSNR: 38.12 -> 40.85 | Improved |
| add U-shaped timestep sampling (ablation) | DRIFT Stage 2 · HCP (avg 1.5/3.0/5.0mm) | LPIPS: 0.0342 -> 0.0229 | Hurt |
| use fixed 1.0 mm proximal CETA gap instead of random pairs (ablation) | DRIFT CETA · HCP (avg 1.5/3.0/5.0mm) | PSNR: 38.23 -> 40.85 | Improved |
| use 1.0 mm CETA gap instead of 0.5 mm (ablation) | DRIFT CETA · HCP (avg 1.5/3.0/5.0mm) | PSNR: 37.24 -> 40.85 | Improved |
| use 1.0 mm CETA gap instead of 2.0 mm (ablation) | DRIFT CETA · HCP (avg 1.5/3.0/5.0mm) | PSNR: 39.10 -> 40.85 | Improved |
| use 1.0 mm CETA gap instead of 4.0 mm (ablation) | DRIFT CETA · HCP (avg 1.5/3.0/5.0mm) | PSNR: 37.58 -> 40.85 | Improved |
| combine Stage 1 APN with Stage 2 rectified flow (vs either alone) (ablation) | DRIFT · HCP | PSNR/SSIM | Improved |
| use physics-aware difficulty (slice thickness) scheduling instead of image-dependent or hybrid scheduling (ablation) | DRIFT AIS · CC359 (multi-vendor T1w) | HF-PSNR | Improved |
Closest papers by meaning
- MiCo: Microstructure-Consistent Flow Matching for Diffusion MRI Angular Super-Resolution (NeurIPS 2026 · similarity 0.87)
- Rethinking Diffusion Model for Multi-Contrast MRI Super-Resolution (CVPR 2024 · similarity 0.85)
- Anatomical Consistency and Adaptive Prior-informed Transformation for Multi-contrast MR Image Synthesis via Diffusion Model (CVPR 2025 · similarity 0.84)
- D2SA: Dual-Stage Distribution and Slice Adaptation for Efficient Test-Time Adaptation in MRI Reconstruction (NeurIPS 2025 · similarity 0.84)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · similarity 0.84)
- Self-Adaptive Reality-Guided Diffusion for Artifact-Free Super-Resolution (CVPR 2024 · similarity 0.83)
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.
