Research map / 3D vision / Gaussian splatting
InstantHDR: Single-forward Gaussian Splatting for High Dynamic Range 3D Reconstruction
Dingqiang Ye, Jiacong Xu, Jianglu Ping, Yuxiang Guo, Chao Fan, Vishal Patel
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 exposure normalization (FiLM) (ablation) | InstantHDR · HDR-NeRF real (8 views, zero-shot) | PSNR: 13.72 -> 18.95 dB | Improved |
| add Meta Net for scene-specific tone mapping (ablation) | InstantHDR · HDR-NeRF real (8 views, zero-shot) | PSNR: 16.32 -> 18.95 dB | Improved |
| add geo-guided cross-view attention (ablation) | InstantHDR · HDR-NeRF real (8 views, zero-shot) | PSNR: 17.63 -> 18.95 dB | Improved |
| add DoG high-resolution upsampling (ablation) | InstantHDR · HDR-NeRF real (8 views, zero-shot) | PSNR / LPIPS: PSNR 19.20 -> 18.95; LPIPS 0.386 -> 0.269 | Mixed |
| one-time finetuning on real HDR-Plenoxels scenes (ablation) | InstantHDR · HDR-NeRF real (18 views) | PSNR: 18.08 -> 19.48 dB | Improved |
| use InstantHDR instead of exposure-aware AnySplat baseline | AnySplat · HDR-NeRF real (4 views) | PSNR: 14.59 -> 18.44 dB | Improved |
| add feed-forward initialization with 1K post-optimization vs GaussHDR | InstantHDR_1K · HDR-NeRF real (4 views) | PSNR: 19.26 -> 22.16 dB | Improved |
| same comparison with dense 18 views | InstantHDR_1K · HDR-NeRF real (18 views) | PSNR: 29.36 -> 29.19 dB | No clear effect |
| same comparison on synthetic scenes with 8 views | InstantHDR_1K · HDR-NeRF synthetic (8 views) | PSNR: 34.49 -> 32.75 dB | Hurt |
| post-optimize AnySplat for 1K iterations without HDR modeling | AnySplat_1K · HDR-NeRF real (4 views) | PSNR: 12.10 -> 11.84 dB | 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.
- HDR-GS — HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting (NeurIPS 2024) · better on 2 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
- GaussHDR: High Dynamic Range Gaussian Splatting via Learning Unified 3D and 2D Local Tone Mapping (CVPR 2025 · similarity 0.94)
- HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting (NeurIPS 2024 · similarity 0.93)
- High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · similarity 0.92)
- Single Shot HDR Recovery via a Video Diffusion Prior (NeurIPS 2026 · similarity 0.90)
- EvHDR-GS: Event-guided HDR Video Reconstruction with 3D Gaussian Splatting (AAAI 2025 · similarity 0.90)
- From Chaos to Clarity: 3DGS in the Dark (NeurIPS 2024 · similarity 0.90)
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.
