Research map / Detection and segmentation / CNN features
SplitHDR: Saturation-Aware HDR Recovery and Denoising for Real-Time Detection
Jaewon Kim, Sol Namkung, Dongsuk Jeon
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 |
|---|---|---|---|
| extend input stack from two to three exposures (ablation) | SplitHDR C1 · ROD synthetic HDR benchmark (ablation) | PSNR-µ: 40.67 -> 40.93 | Improved |
| extend input stack from two to three exposures (ablation) | SplitHDR C1 · ROD synthetic HDR benchmark (ablation) | AP: 37.88 -> 37.68 | Hurt |
| add dual-mode configuration with full model always executed (ablation) | SplitHDR C2 · ROD synthetic HDR benchmark (ablation) | PSNR-µ: 40.93 -> 44.01 | Improved |
| add dual-mode configuration with full model always executed (ablation) | SplitHDR C2 · ROD synthetic HDR benchmark (ablation) | AP: 37.68 -> 37.80 | No clear effect |
| add masked supervision to dual-mode (ablation) | SplitHDR C3 · ROD synthetic HDR benchmark (ablation) | AP: 37.80 -> 38.38 | Improved |
| add masked supervision to dual-mode (ablation) | SplitHDR C3 · ROD synthetic HDR benchmark (ablation) | PSNR-µ: 44.01 -> 44.96 | Improved |
| add saturation-aware conditional execution (ablation) | SplitHDR C4 · ROD synthetic HDR benchmark (ablation) | GMACs: 10.23 -> 3.63 | Improved |
| add saturation-aware conditional execution (ablation) | SplitHDR C4 · ROD synthetic HDR benchmark (ablation) | AP: 38.38 -> 38.38 | No clear effect |
| use soft mask instead of binary mask at 90% threshold (ablation) | SplitHDR · ROD synthetic HDR benchmark (mask/threshold sweep) | AP: 38.35 -> 38.38 | Improved |
| raise saturation threshold from 90% to 99% (ablation) | SplitHDR soft mask · ROD synthetic HDR benchmark (mask/threshold sweep) | AP: 38.38 -> 37.75 | Hurt |
| use 128x128 tiles with conditional execution vs without (ablation) | SplitHDR · ROD synthetic HDR benchmark (tile size) | Total latency (ms): 17.70 -> 15.24 | Improved |
| use 32x32 tiles instead of 128x128 (ablation) | SplitHDR · ROD synthetic HDR benchmark (tile size) | Total latency (ms): 15.24 -> 21.24 | Hurt |
Closest papers by meaning
- RawHDR: High Dynamic Range Image Reconstruction from a Single Raw Image (ICCV 2023 · similarity 0.89)
- Single Shot HDR Recovery via a Video Diffusion Prior (NeurIPS 2026 · similarity 0.88)
- Neural Exposure Fusion for High-Dynamic Range Object Detection (CVPR 2024 · similarity 0.87)
- High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · similarity 0.87)
- Real-Time Scene-Adaptive Tone Mapping for High-Dynamic Range Object Detection (NeurIPS 2025 · similarity 0.87)
- GaussHDR: High Dynamic Range Gaussian Splatting via Learning Unified 3D and 2D Local Tone Mapping (CVPR 2025 · similarity 0.86)
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.
