Research map / Architectures and efficiency / Attention and transformers
Uncertainty-Driven Spectral Compressive Imaging with Spatial-Frequency Transformer
Lintao Peng, Siyu Xie, Liheng Bian
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 FWSA to the baseline (ablation) | Specformer baseline · KAIST (trained on CAVE) | PSNR (dB): 31.08 -> 35.97 | Improved |
| add LWSA to the baseline (ablation) | Specformer baseline · KAIST (trained on CAVE) | PSNR (dB): 31.08 -> 33.34 | Improved |
| combine LWSA and FWSA in parallel instead of serially (ablation) | SF-block · KAIST (trained on CAVE) | PSNR (dB): 37.58 -> 39.82 | Improved |
| parallel instead of serial design reduces cost (ablation) | SF-block · KAIST (trained on CAVE) | Params / GFLOPs: 2.82M -> 2.48M; 46.73 -> 39.85 | Improved |
| add uncertainty-driven loss (two-stage training) (ablation) | Specformer · KAIST (trained on CAVE) | PSNR (dB): 39.82 -> 40.61 | Improved |
| add uncertainty-driven loss (ablation) | Specformer · KAIST (trained on CAVE) | Params / GFLOPs at test time: 2.48M, 39.85 unchanged | No clear effect |
| two-stage UDL training | Specformer · training cost | training time: about +30% | Hurt |
Closest papers by meaning
- SPECAT: SPatial-spEctral Cumulative-Attention Transformer for High-Resolution Hyperspectral Image Reconstruction (CVPR 2024 · similarity 0.91)
- Mask-Guided Spectral-Wise Transformer for Efficient Hyperspectral Image Reconstruction (CVPR 2022 · similarity 0.91)
- Coarse-to-Fine Sparse Transformer for Hyperspectral Image Reconstruction (ECCV 2022 · similarity 0.91)
- Dual-Window Multiscale Transformer for Hyperspectral Snapshot Compressive Imaging (AAAI 2024 · similarity 0.90)
- Improving Spectral Snapshot Reconstruction with Spectral-Spatial Rectification (CVPR 2024 · similarity 0.90)
- Spectral Compressive Imaging via Chromaticity-Intensity Decomposition (NeurIPS 2025 · similarity 0.88)
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.
