Research map / Image generation / Super-resolution
ARM: Any-Time Super-Resolution Method
Bohong Chen, Mingbao Lin, Kekai Sheng, Mengdan Zhang, Peixian Chen, Ke Li, Liujuan Cao, Rongrong Ji
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 |
|---|---|---|---|
| use ARM-L subnet selection instead of the static backbone | FSRCNN supernet · F2K | PSNR: 27.91 -> 27.98 (468M -> 380M FLOPs) | Improved |
| use ARM instead of ClassSR (no extra parameters) | FSRCNN · Test2K | PSNR: 25.61 -> 25.64 PSNR; 311M -> 366M FLOPs | Improved |
| use ARM-S smallest budget instead of the backbone | FSRCNN · F2K | PSNR: 27.91 -> 27.65 (468M -> 184M FLOPs) | Hurt |
| use ARM-M instead of the CARN backbone | CARN · Test2K | PSNR: 25.95 -> 26.02 (1.15G -> 831M FLOPs) | Improved |
| use ARM-L instead of ClassSR | CARN · Test4K | PSNR: 27.42 -> 27.45 | Improved |
| use ARM-M instead of SRResNet backbone | SRResNet · F2K | PSNR: 29.01 -> 29.01 (5.20G -> 3.59G FLOPs) | No clear effect |
| use ARM lower latency subnets instead of ClassSR | FSRCNN · Test2K | latency: 549.48 -> 531.36 ms (ARM-M) | Improved |
| choose subnets with Edge-to-PSNR lookup tables instead of manual edge thresholds (ablation) | ARM-FSRCNN · DIV2K-trained, test sets | FLOPs-PSNR tradeoff | Improved |
| use Sobel or Prewitt instead of Laplacian edge operator (ablation) | ARM · FLOPs-PSNR | tradeoff | No clear effect |
| increase the number of subnets beyond 3 (ablation) | ARM · FLOPs-PSNR | tradeoff | Improved |
| sample subnets by FLOPs^2 instead of uniform sampling (ablation) | ARM supernet · FLOPs-PSNR | tradeoff | Mixed |
| use FLOPs^3 sampling instead of FLOPs^2 (ablation) | ARM supernet · FLOPs-PSNR | tradeoff | Mixed |
Closest papers by meaning
- Accelerating Image Super-Resolution Networks with Pixel-Level Classification (ECCV 2024 · similarity 0.88)
- Image Processing GNN: Breaking Rigidity in Super-Resolution (CVPR 2024 · similarity 0.87)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.87)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.87)
- Sequence Matters: Harnessing Video Models in 3D Super-Resolution (AAAI 2025 · similarity 0.86)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.86)
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.
