Research map / 3D vision / Depth and stereo
Bridging RGB and RAW: Single-step Deterministic Flow with Homogeneous Representation Alignment
Diedong Feng, Peiyi Zeng, Zhen Liu, Zhongyang Li, Bing Zeng, Shuaicheng Liu
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 teacher weight loading and representation alignment (full HRA vs base) (ablation) | SHADE · FiveK-Nikon | PSNR: 28.05 -> 30.89 | Improved |
| add teacher weight loading (vs w/o Pretrain) (ablation) | SHADE · FiveK-Nikon | PSNR: 30.01 -> 30.89 | Improved |
| add representation alignment (vs w/o Align) (ablation) | SHADE · FiveK-Nikon | PSNR: 29.39 -> 30.89 | Improved |
| add representation alignment (vs w/o Align) (ablation) | SHADE · PASCALRAW | PSNR: 38.60 -> 39.07 | Improved |
| add L2 loss (ablation) | SHADE · FiveK-Nikon | PSNR: 29.62 -> 30.89 | Improved |
| add log-L1 loss (ablation) | SHADE · PASCALRAW | PSNR: 35.51 -> 39.07 | Improved |
| add log-L1 loss (ablation) | SHADE · FiveK-Nikon | SSIM: 0.8879 -> 0.8858 | Hurt |
| add AE reconstruction loss (ablation) | SHADE · PASCALRAW | PSNR: 36.83 -> 39.07 | Improved |
| add DiT reconstruction loss (ablation) | SHADE · FiveK-Nikon | PSNR: 29.21 -> 30.89 | Improved |
| single-step deterministic flow instead of standard flow matching (ablation) | SHADE · PASCALRAW | PSNR: 37.35 -> 39.07 | Improved |
| single-step deterministic flow instead of Mean Flow from noise (ablation) | SHADE · FiveK-Nikon | PSNR: 30.38 -> 30.89 | Improved |
| single-step deterministic flow instead of direct regression (ablation) | SHADE · PASCALRAW | PSNR: 37.69 -> 39.07 | Improved |
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.
- RAW-Flow — RAW-Flow: Advancing RGB-to-RAW Image Reconstruction with Deterministic Latent Flow Matching (AAAI 2026) · better on 3 of 3 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
- ReRAW: RGB-to-RAW Image Reconstruction via Stratified Sampling for Efficient Object Detection on the Edge (CVPR 2025 · similarity 0.87)
- ThermalGen: Style-Disentangled Flow-Based Generative Models for RGB-to-Thermal Image Translation (NeurIPS 2025 · similarity 0.85)
- ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency (AAAI 2025 · similarity 0.85)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.85)
- RGB-to-Polarization Estimation: A New Task and Benchmark Study (NeurIPS 2025 · similarity 0.84)
- GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · similarity 0.84)
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.
