Research map / Architectures and efficiency / Attention and transformers
From Local to Global: A Progressive Reconstruction Network for Diffractive Snapshot Spectral Imaging
Zhengyue Zhuge, Shiqi Chen, Chi Zhang, Jiahui Xu, Tianchen Qiu, Dingchuan Yu, Yueting Chen
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 |
|---|---|---|---|
| use local-to-global two-stage design instead of single-stage global (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.35 -> 39.76 | Improved |
| use local-to-global two-stage design instead of single-stage global (ablation) | LGSRNet · ICVL+VNIRSR (combined) | GFLOPs: 118.07 -> 102.34 | Improved |
| arrange local and global modules in parallel instead of local-to-global (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.76 -> 38.34 | Hurt |
| arrange local and global modules alternately instead of local-to-global (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.76 -> 38.92 | Hurt |
| replace local-first stage with a second global stage (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.76 -> 39.65 | Hurt |
| add SGCA channel attention instead of SCA (no FF) (ablation) | LGSRNet w/o FF · ICVL+VNIRSR (combined) | PSNR (dB): 38.63 -> 39.00 | Improved |
| add SCA channel attention without feature fusion (ablation) | LGSRNet w/o FF · ICVL+VNIRSR (combined) | PSNR (dB): 38.64 -> 38.63 | Hurt |
| add EAGF feature fusion (no channel attention) (ablation) | LGSRNet w/o CA · ICVL+VNIRSR (combined) | PSNR (dB): 38.64 -> 39.14 | Improved |
| combine SGCA and EAGF versus neither (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 38.64 -> 39.76 | Improved |
| use EAGF instead of CSFF fusion (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.59 -> 39.76 | Improved |
| use ECSS sample selection instead of gradient-based selection (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.44 -> 39.50 | Improved |
| use ECSS sample selection instead of no selection (ablation) | LGSRNet · ICVL+VNIRSR (combined) | PSNR (dB): 39.26 -> 39.50 | Improved |
Closest papers by meaning
- Spectral Compressive Imaging via Chromaticity-Intensity Decomposition (NeurIPS 2025 · similarity 0.90)
- Prior Image Guided Snapshot Compressive Spectral Imaging (TPAMI 2023 · similarity 0.90)
- DiffSCI: Zero-Shot Snapshot Compressive Imaging via Iterative Spectral Diffusion Model (CVPR 2024 · similarity 0.87)
- Improving Spectral Snapshot Reconstruction with Spectral-Spatial Rectification (CVPR 2024 · similarity 0.87)
- Dual-Window Multiscale Transformer for Hyperspectral Snapshot Compressive Imaging (AAAI 2024 · similarity 0.86)
- FRN: Fractal-Based Recursive Spectral Reconstruction Network (NeurIPS 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.
