Research map / Image generation / Fast diffusion sampling
FLOWER: A Flow-Matching Solver for Inverse Problems
Mehrsa Pourya, Bassam El Rawas, Michael Unser
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 independent source-target coupling when training the velocity network (instead of mini-batch OT) (ablation) | Flower · CelebA (5 tasks) | PSNR: Deblurring 34.98 -> 35.22 | Improved |
| average five Flower evaluations instead of one (ablation) | Flower-OT · CelebA deblurring | PSNR: 34.98 -> 35.67 | Improved |
| average five Flower evaluations instead of one (ablation) | Flower-OT · CelebA deblurring | LPIPS: 0.026 -> 0.032 | Hurt |
| set gamma=0 (ignore refinement uncertainty) instead of gamma=1 (ablation) | Flower · CelebA deblurring | PSNR: 30.84 -> 33.01 | Improved |
| use gamma=1 (theory-required sampling) with averaging (ablation) | Flower · CelebA deblurring | runs needed for comparable PSNR: 100 runs vs 5 runs | Hurt |
| gamma=0 in a 2D GMM toy posterior (ablation) | Flower · Gaussian mixture toy | posterior tail coverage | Hurt |
| replace PnP-Flow gradient update with proximal step | Flower vs PnP-Flow · CelebA | PSNR: Deblurring 34.48 -> 34.98 (single evaluation) | Improved |
| Flower versus PnP-Flow runtime | Flower · CelebA deblurring | time (s): 3.020 -> 5.622 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PnP-Flow1 — PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (ICLR 2025) · better on 12 of 16 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Flow Priors for Linear Inverse Problems via Iterative Corrupted Trajectory Matching (NeurIPS 2024 · similarity 0.91)
- Latent Refinement via Flow Matching for Training-free Linear Inverse Problem Solving (NeurIPS 2025 · similarity 0.90)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.89)
- Fast samplers for Inverse Problems in Iterative Refinement models (NeurIPS 2024 · similarity 0.89)
- PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (ICLR 2025 · similarity 0.87)
- Exact Posterior Score Estimation for Solving Linear Inverse Problems (NeurIPS 2026 · 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.
