AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Memory-Efficient Hierarchical Neural Architecture Search for Image Restoration

Haokui Zhang, Ying Li, Hao Chen, Chengrong Gong, Zongwen Bai, Chunhua Shen

IJCV 2021 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 14 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 layer-wise architecture sharing (ablation)HiNAS · BSD500 denoising (sigma=30)PSNR: 29.15 -> 29.17Improved
add residual learning (ablation)HiNAS · BSD500 denoising (sigma=30)PSNR: 29.15 -> 29.20Improved
add both LWAS and residual learning (ablation)HiNAS · BSD500 denoising (sigma=30)PSNR: 29.15 -> 29.23Improved
add residual learning (ablation)HiNAS · BSD100 x4 super-resolutionPSNR: 26.92 -> 27.32Improved
add layer-wise architecture sharing (ablation)HiNAS · BSD100 x4 super-resolutionPSNR: 26.92 -> 27.03Improved
add both strategies (ablation)HiNAS · BSD100 x4 super-resolutionPSNR: 26.92 -> 27.35Improved
search the outer layer width (w40 instead of fixed width) (ablation)HiNAS · BSD500 denoisingPSNR: 29.25 -> 29.14Hurt
add l_ssim loss to MSE (ablation)HiNAS · BSD500 denoising (sigma=30)SSIM: 0.8254 -> 0.8403Improved
add l_ssim loss to MSE (ablation)HiNAS · BSD100 x4 super-resolutionPSNR: 27.38 -> 27.35Hurt
add l_ssim loss to MSE (ablation)HiNAS · BSD100 x4 super-resolutionSSIM: 0.7367 -> 0.7467Improved
replace searched operations with other convolutions (R1) (ablation)HiNAS · BSD500 denoisingPSNR: 29.25 -> 29.21Hurt
change cell topology (R2) (ablation)HiNAS · BSD500 denoisingSSIM: 0.8420 -> 0.8304Hurt

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