Research map / Architectures and efficiency / Attention and transformers
Coarse-to-Fine Sparse Transformer for Hyperspectral Image Reconstruction
Yuanhao Cai, Jing Lin, Xiaowan Hu, Haoqian Wang, Xin Yuan, Yulun Zhang, Radu Timofte, Luc Van Gool
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 SAH-MSA hashing attention (ablation) | CST-M baseline · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 32.57 -> 35.53 | Improved |
| add SASM patch screening on top of SAH-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | FLOPS (G): 24.60 -> 16.91 | Improved |
| add SASM patch screening on top of SAH-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.53 -> 35.31 | Hurt |
| use SASM instead of random sparsity (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 34.37 -> 35.31 | Improved |
| use SASM instead of uniform sparsity (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 34.33 -> 35.31 | Improved |
| use SAH-MSA instead of G-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.04 -> 35.53 | Improved |
| use SAH-MSA instead of W-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.02 -> 35.53 | Improved |
| use SAH-MSA instead of Swin-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.12 -> 35.53 | Improved |
| use SAH-MSA instead of S-MSA (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.21 -> 35.53 | Improved |
| restrict hash clustering to local patches rather than global image (ablation) | CST-M · KAIST 10 scenes ablation (CST-M base) | PSNR (dB): 35.33 -> 35.53 | Improved |
| increase sparsity ratio sigma beyond 50% (ablation) | CST-M · KAIST 10 scenes | PSNR (dB) | Hurt |
| increase hashing rounds R from 1 to 2 and beyond (ablation) | CST-M · KAIST 10 scenes | PSNR (dB) | Mixed |
Closest papers by meaning
- Mask-Guided Spectral-Wise Transformer for Efficient Hyperspectral Image Reconstruction (CVPR 2022 · similarity 0.93)
- Dual-Window Multiscale Transformer for Hyperspectral Snapshot Compressive Imaging (AAAI 2024 · similarity 0.92)
- In2SET: Intra-Inter Similarity Exploiting Transformer for Dual-Camera Compressive Hyperspectral Imaging (CVPR 2024 · similarity 0.88)
- Improving Spectral Snapshot Reconstruction with Spectral-Spatial Rectification (CVPR 2024 · similarity 0.88)
- SPECAT: SPatial-spEctral Cumulative-Attention Transformer for High-Resolution Hyperspectral Image Reconstruction (CVPR 2024 · similarity 0.87)
- Learning Multi-Modal Cross-Scale Deformable Transformer Network for Unregistered Hyperspectral Image Super-resolution (AAAI 2024 · 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.
