Research map / Learning theory and algorithms / Matrix and tensor methods
DeepPhaseCut: Deep Relaxation in Phase for Unsupervised Fourier Phase Retrieval
Eunju Cha, Chanseok Lee, Mooseok Jang, 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 cycle consistency loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | PSNR (dB): 9.719 -> 24.284 | Improved |
| add cycle consistency loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | SSIM: 0.098 -> 0.748 | Improved |
| add PhaseCut loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | PSNR (dB): 21.513 -> 24.284 | Improved |
| add PhaseCut loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | SSIM: 0.676 -> 0.748 | Improved |
| add adversarial loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | PSNR (dB): 23.653 -> 24.284 | Improved |
| add adversarial loss (ablation) | DeepPhaseCut · Microscopy beads (1,000 test images) | SSIM: 0.741 -> 0.748 | Improved |
| use both phase network and image refinement network instead of a single generator (ablation) | DeepPhaseCut · fastMRI | PSNR (dB): 3.01 ~ 13.28 dB better than the ablated networks | Improved |
| replace HIO and PhaseCut with feed-forward DeepPhaseCut | PhaseCut · CelebA (2,478 test images) | runtime (sec): 1044.720 -> 0.048 | Improved |
| use DeepPhaseCut instead of the baselines | baselines · CelebA (2,478 test images) | PSNR (dB): 1.06 ~ 7.55 dB | Improved |
| use DeepPhaseCut instead of HIO and PhaseCut on complex-valued data | HIO · fastMRI (4,873 test images) | PSNR (dB): 18.806 -> 19.820 | Improved |
| use DeepPhaseCut instead of prDeep (pretrained denoiser regularizer) | prDeep · Microscopy beads (1,000 test images) | PSNR (dB): 23.849 -> 24.284 | Improved |
| initialize On-RED with HIO result | HIO · CelebA (2,478 test images) | PSNR (dB): 19.573 -> 19.796 | No clear effect |
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.
- prDeep — prDeep: Robust Phase Retrieval with a Flexible Deep Network (ICML 2018) · better on 6 of 6 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
- Solving Phase Retrieval with a Learned Reference (ECCV 2020 · similarity 0.89)
- When deep denoising meets iterative phase retrieval (ICML 2020 · similarity 0.89)
- prDeep: Robust Phase Retrieval with a Flexible Deep Network (ICML 2018 · similarity 0.87)
- Phase Retrieval Under a Generative Prior (NeurIPS 2018 · similarity 0.87)
- A Fast and Provable Algorithm for Sparse Phase Retrieval (ICLR 2024 · similarity 0.84)
- DiffFPR: Diffusion Prior for Oversampled Fourier Phase Retrieval (ICML 2024 · similarity 0.82)
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.
