Research map / Label-efficient and robust learning / Domain adaptation and generalization
S2R-HDR: A Large-Scale Rendered Dataset for HDR Fusion
Yujin Wang, Jiarui Wu, Yichen Bian, Fan Zhang, Tianfan Xue
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 |
|---|---|---|---|
| pretrain on S2R-HDR then adapt with S2R-Adapter instead of training on Challenge123 only | SAFNet · Challenge123 | PSNR-mu: 41.88 -> 43.43 | Improved |
| pretrain on S2R-HDR then adapt with S2R-Adapter instead of training on SCT only | SCTNet · SCT | PSNR-mu: 42.55 -> 43.24 | Improved |
| train on synthetic S2R-HDR instead of real SCT for cross-dataset testing | SAFNet · Challenge123 | PSNR-mu: 34.26 -> 42.75 | Improved |
| train on S2R-HDR instead of Challenge123 when testing on SCT | SAFNet · SCT | PSNR-mu: 35.14 -> 34.89 | Hurt |
| add test-time adaptation with S2R-Adapter on top of S2R-HDR pretraining | SAFNet · SCT (no GT) | PSNR-mu: 34.89 -> 36.28 | Improved |
| add test-time adaptation with S2R-Adapter | SAFNet · SCT (no GT) | PSNR-l: 43.85 -> 47.23 | Improved |
| add S2R-Adapter to SAFNet instead of simple fine-tuning to reduce forgetting (ablation) | SAFNet · S2R-HDR train set after SCT adaptation | PSNR-mu: 35.52 -> 35.95 | Improved |
| use S2R-Adapter instead of fine-tuning (ablation) | SCTNet · Challenge123 | PSNR-mu: 42.1 -> 42.58 | Improved |
| use S2R-Adapter instead of fine-tuning (ablation) | SAFNet · SCT | PSNR-mu: 43.03 -> 43.33 | Improved |
| add shared and transfer branches individually vs together (ablation) | S2R-Adapter on SAFNet · SCT | PSNR-mu | Improved |
| add teacher-student framework and uncertainty-based scale adjustment to test-time adaptation (ablation) | S2R-Adapter · SCT | PSNR | Improved |
| pretrain on S2R-HDR rather than Challenge123 before adapting with S2R-Adapter (ablation) | SAFNet · SCT | PSNR-mu: 42.92 -> 43.44 | Improved |
Closest papers by meaning
- HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting (NeurIPS 2024 · similarity 0.89)
- GaussHDR: High Dynamic Range Gaussian Splatting via Learning Unified 3D and 2D Local Tone Mapping (CVPR 2025 · similarity 0.89)
- Towards Real-World HDR Video Reconstruction: A Large-Scale Benchmark Dataset and A Two-Stage Alignment Network (CVPR 2024 · similarity 0.89)
- EvHDR-GS: Event-guided HDR Video Reconstruction with 3D Gaussian Splatting (AAAI 2025 · similarity 0.88)
- High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · similarity 0.88)
- LEDiff: Latent Exposure Diffusion for HDR Generation (CVPR 2025 · 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.
