Research map / Image generation / Image restoration
Flare7K++: Mixing Synthetic and Real Datasets for Nighttime Flare Removal and Beyond
Yuekun Dai, Chongyi Li, Shangchen Zhou, Ruicheng Feng, Yihang Luo, Chen Change Loy
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 Flare-R real flare data to Flare7K training (ablation) | Uformer · Flare7K++ real-captured test set | PSNR: 27.257 -> 27.633 | Improved |
| add Flare-R real flare data to Flare7K training (ablation) | Uformer · Flare7K++ real-captured test set | S-PSNR: 21.294 -> 22.603 | Improved |
| add Flare-R real flare data to Flare7K training (ablation) | Uformer · Flare7K++ real-captured test set | LPIPS: 0.0471 -> 0.0428 | Improved |
| add light source annotations to training (ablation) | Uformer · Flare7K++ real-captured test set | PSNR: 26.850 -> 27.633 | Improved |
| add light source annotations to training (ablation) | Uformer · Flare7K++ real-captured test set | S-PSNR: 21.909 -> 22.603 | Improved |
| add light source annotations to training (ablation) | Uformer · Flare7K++ real-captured test set | SSIM: 0.895 -> 0.894 | No clear effect |
| use end-to-end pipeline without subtraction instead of Wu et al. pipeline (ablation) | Uformer · Flare7K++ real-captured test set | PSNR: 27.166 -> 27.633 | Improved |
| use end-to-end pipeline without subtraction instead of Wu et al. pipeline (ablation) | Uformer · Flare7K++ real-captured test set | SSIM: 0.861 -> 0.894 | Improved |
| use end-to-end pipeline without subtraction instead of Wu et al. pipeline (ablation) | Uformer · Flare7K++ real-captured test set | S-PSNR: 22.118 -> 22.603 | Improved |
| train U-Net on Flare7K++ instead of Wu et al. dataset | U-Net · Flare7K++ real-captured test set | PSNR: 24.613 -> 27.189 | Improved |
| train on Flare7K++ with new pipeline instead of Flare7K baseline | Uformer · Flare7K++ real-captured test set | S-PSNR: 21.563 -> 22.603 | Improved |
| use Uformer backbone instead of U-Net on Flare7K++ | Flare7K++ network · Flare7K++ real-captured test set | PSNR: 27.189 -> 27.633 | Improved |
Closest papers by meaning
- Flare7K: A Phenomenological Nighttime Flare Removal Dataset (NeurIPS 2022 · similarity 0.98)
- FlareReal: A Real-Captured Paired Dataset for Nighttime Lens Flare Removal (NeurIPS 2026 · similarity 0.95)
- Disentangle Nighttime Lens Flares: Self-supervised Generation-based Lens Flare Removal (AAAI 2025 · similarity 0.91)
- FlareX: A Physics-Informed Dataset for Lens Flare Removal via 2D Synthesis and 3D Rendering (NeurIPS 2025 · similarity 0.91)
- Language-Guided One-Step Diffusion Model for Nighttime Flare Removal (CVPR 2026 · similarity 0.91)
- FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views (CVPR 2025 · 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.
