Research map / Image generation / Image restoration
Learning Enriched Features for Fast Image Restoration and Enhancement
Syed Waqas Zamir, Aditya Arora, Salman Hameed Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming–Hsuan Yang, Ling Shao
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 skip connections (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 28.21 -> 30.97 | Improved |
| add residual contextual block (RCB) (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.79 -> 30.97 | Improved |
| add intermediate SKFF information exchange (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.85 -> 30.97 | Improved |
| use SKFF instead of sum for feature aggregation (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.76 -> 30.97 | Improved |
| use SKFF instead of concatenation (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.83 -> 30.97 | Improved |
| use RCB instead of EDSR Resblock baseline (g=2) (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.84 -> 30.97 | Improved |
| add the transform part to RCB (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.92 -> 30.97 | Improved |
| use regular convolutions (g=1) instead of group convolutions (g=2) in RCB (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.97 -> 31.05; params 5.9M -> 9.7M | Mixed |
| use progressive learning instead of fixed 128 patches (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.97 -> 31.06 | Improved |
| use progressive learning instead of fixed 224 patches (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 31.08 -> 31.06; train time 33h -> 22h | No clear effect |
| increase parallel streams from 1 to 3 (3 columns) (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.47 -> 31.03 | Improved |
| increase columns of RCBs from 1 to 3 (3 streams) (ablation) | MIRNet-v2 · RealSR x3 (ablation) | PSNR (dB): 30.73 -> 31.03 | Improved |
Closest papers by meaning
- Learning Enriched Features for Real Image Restoration and Enhancement (ECCV 2020 · similarity 0.99)
- Revitalizing Convolutional Network for Image Restoration (TPAMI 2024 · similarity 0.91)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · similarity 0.90)
- Unfolding Taylor's Approximations for Image Restoration (NeurIPS 2021 · similarity 0.90)
- Boosting Image Restoration via Priors from Pre-trained Models (CVPR 2024 · similarity 0.87)
- Adapt or Perish: Adaptive Sparse Transformer with Attentive Feature Refinement for Image Restoration (CVPR 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.
