AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Degradation-Aware Unfolding Half-Shuffle Transformer for Spectral Compressive Imaging

Yuanhao Cai, Jing Lin, Haoqian Wang, Xin Yuan, Henghui Ding, Yulun Zhang, Radu Timofte, Luc Van Gool

NeurIPS 2022 · Architectures and efficiency › Attention and transformers · approach: Efficient image transformers · 63 citations (OpenAlex)

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 changedWhereMeasuredEffect
add DAUF (degradation-aware parameter estimation) to baseline (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.05 -> 35.37Improved
add HS-MSA to baseline (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.05 -> 35.49Improved
combine DAUF and HS-MSA (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.05 -> 37.21Improved
HS-MSA instead of global MSA (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.63 -> 34.05Improved
HS-MSA instead of Swin window MSA (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.75 -> 34.05Improved
HS-MSA instead of spectral-wise MSA (ablation)DAUHST-3stg · KAIST simulationPSNR: 33.82 -> 34.05Improved
DAUF instead of DNU unfolding framework (ablation)DAUHST-3stg · KAIST simulationPSNR: 34.62 -> 37.21Improved
DAUF instead of GAP framework (ablation)DAUHST-3stg · KAIST simulationPSNR: 35.58 -> 37.21Improved
estimate alpha (projection scale) with estimator (ablation)DAUHST-3stg · KAIST simulationPSNR: 36.49 -> 36.94Improved
feed beta (noise level) into denoiser (ablation)DAUHST-3stg · KAIST simulationPSNR: 36.49 -> 36.83Improved
use both alpha and beta (ablation)DAUHST-3stg · KAIST simulationPSNR: 36.49 -> 37.21Improved
increase unfolding stages from 3 to 9 (ablation)DAUHST-3stg · KAIST simulationPSNR: 37.21 -> 38.36Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy