Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-LAR: Semi-supervised Contrastive Learning with Linear Attention for Removal of Nighttime Flares
Xiyu Zhu, Wei Wang, Kui Jiang, Zhengguo Li
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 dependable repository (DR) for pseudo labels (ablation) | Semi-LAR · Flare7K++ test | PSNR: 26.746 -> 27.620 | Improved |
| add flare-aware contrastive loss (ablation) | Semi-LAR · Flare7K++ test | PSNR: 26.711 -> 27.620 | Improved |
| add both DR and contrastive loss together (ablation) | Semi-LAR · Flare7K++ test | PSNR: 26.541 -> 27.620 | Improved |
| add momentum-based pseudo-label updating (ablation) | dependable repository · Flare7K++ test | PSNR: 26.746 -> 26.982 | Improved |
| add NR-IQA quality gating on top of momentum (ablation) | dependable repository · Flare7K++ test | PSNR: 26.982 -> 27.620 | Improved |
| add strong augmentations (color jitter, grayscale, blur) to unlabeled data (ablation) | Semi-LAR · Flare7K++ test | PSNR: 26.796 -> 27.620 | Improved |
| add Gaussian blur alone as strong augmentation (ablation) | Semi-LAR · Flare7K++ test | PSNR: 26.796 -> 27.159 | Improved |
| add 8K unlabeled images to 8K labeled training | Uformer · Flare7K++ test | PSNR: 27.392 -> 27.467 | Improved |
| add 8K unlabeled images to full supervision | Uformer · Flare7K++ test | PSNR: 27.633 -> 28.045 | Improved |
| add semi-supervised training on 8K labeled images versus SOTA limited-supervised SGSFT | Semi-LAR · Flare7K++ test | LPIPS: 0.0497 -> 0.0493 | No clear effect |
| train with 8K labeled plus 8K unlabeled instead of full labeled data | Semi-LAR · FlareX (cross-dataset) | PSNR: 19.496 -> 19.960 | Improved |
| semi-supervised 8K+8K versus Flare7Kpp full on real flare | Semi-LAR · FlareReal600 (cross-domain) | LPIPS: 0.257 vs 0.260 | Mixed |
Closest papers by meaning
- FlareReal: A Real-Captured Paired Dataset for Nighttime Lens Flare Removal (NeurIPS 2026 · similarity 0.91)
- Disentangle Nighttime Lens Flares: Self-supervised Generation-based Lens Flare Removal (AAAI 2025 · similarity 0.90)
- Improving Lens Flare Removal with General-Purpose Pipeline and Multiple Light Sources Recovery (ICCV 2023 · similarity 0.90)
- Language-Guided One-Step Diffusion Model for Nighttime Flare Removal (CVPR 2026 · similarity 0.88)
- FlareX: A Physics-Informed Dataset for Lens Flare Removal via 2D Synthesis and 3D Rendering (NeurIPS 2025 · similarity 0.87)
- FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views (CVPR 2025 · 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.
