Research map / Architectures and efficiency / Neural architecture search
Memory-Efficient Hierarchical Neural Architecture Search for Image Denoising
Haokui Zhang, Ying Li, Hao Chen, Chunhua Shen
Check 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 |
|---|---|---|---|
| search the outer layer widths rather than manually fixing them (ablation) | HiNAS · BSD500, σ=30 | PSNR: 29.14 -> 29.15 (HiNAS-w40); 29.14 -> 28.89 (HiNAS-wm) | Mixed |
| search the outer layer widths rather than manually fixing them (ablation) | HiNAS · BSD500, σ=30 | SSIM: 0.8403 -> 0.8406 (HiNAS-w40); 0.8403 -> 0.8370 (HiNAS-wm) | Mixed |
| add the SSIM loss term to MSE (ablation) | HiNAS · BSD500, σ=30 | SSIM: 0.8254 -> 0.8403 | Improved |
| add the SSIM loss term to MSE (ablation) | HiNAS · BSD500, σ=30 | PSNR: 29.03 -> 29.14 | Improved |
| add the SSIM loss term to MSE (ablation) | HiNAS · BSD500, σ=50 | SSIM: 0.7498 -> 0.7635 | Improved |
| add the SSIM loss term to MSE (ablation) | HiNAS · BSD500, σ=70 | SSIM: 0.6962 -> 0.7129 | Improved |
| replace conventional convolutions with deformable convolutions (ablation) | searched HiNAS architecture · BSD500, σ=30 | PSNR: 29.14 -> 29.06 | Hurt |
| replace conventional convolutions with deformable convolutions (ablation) | searched HiNAS architecture · BSD500, σ=30 | SSIM: 0.8403 -> 0.8398 | Hurt |
| change the connection relationships between nodes inside each cell (ablation) | searched HiNAS architecture · BSD500, σ=30 | PSNR: 29.14 -> 29.13 | Hurt |
| change the connection relationships between nodes inside each cell (ablation) | searched HiNAS architecture · BSD500, σ=30 | SSIM: 0.8403 -> 0.8400 | Hurt |
| use HiNAS instead of N3Net | image denoising · SIM1800 test | PSNR: 27.62 -> 27.23 | Hurt |
| use HiNAS instead of N3Net | image denoising · SIM1800 test | SSIM: 0.8191 -> 0.8326 | Improved |
Closest papers by meaning
- Memory-Efficient Hierarchical Neural Architecture Search for Image Restoration (IJCV 2021 · similarity 0.97)
- ISNAS-DIP: Image-Specific Neural Architecture Search for Deep Image Prior (CVPR 2022 · similarity 0.90)
- Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective (ICLR 2021 · similarity 0.90)
- DDPNAS: Efficient Neural Architecture Search via Dynamic Distribution Pruning (IJCV 2023 · similarity 0.90)
- DA-NAS: Data Adapted Pruning for Efficient Neural Architecture Search (ECCV 2020 · similarity 0.89)
- Towards Neural Architecture Search through Hierarchical Generative Modeling (ICML 2024 · 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.
