Research map / Image generation / Super-resolution
Compiler-Aware Neural Architecture Search for On-Mobile Real-Time Super-Resolution
Yushu Wu, Yifan Gong, Pu Zhao, Yanyu Li, Zheng Zhan, Wei Niu, Hao Tang, Minghai Qin, Bin Ren, Yanzhi Wang
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 |
|---|---|---|---|
| tighten latency threshold from 100ms to 40ms | Compiler-aware NAS SR · Set5 x2 | PSNR: 37.64 -> 37.19 | Hurt |
| tighten latency threshold from 100ms to 40ms | Compiler-aware NAS SR · Urban100 x2 | PSNR: 31.08 -> 30.15 | Hurt |
| tighten latency threshold from 100ms to 40ms | Compiler-aware NAS SR · Set5 x4 | PSNR: 32.02 -> 31.73 | Hurt |
| search model with latency-aware NAS instead of using WDSR | WDSR · 720p x2 latency | latency (ms): 1973.31 -> 34.92 | Improved |
| use searched models rather than heuristic depth/width reduction of WDSR (ablation) | WDSR · Set5 x2 | latency at equal PSNR | Improved |
| add depth search to width search (ablation) | Compiler-aware NAS SR · Set5 and Urban100 x2 | latency, PSNR, SSIM | Improved |
| add depth search versus width search only (ablation) | Compiler-aware NAS SR · Urban100 x2 | PSNR, SSIM | Improved |
| search for DSP with its own speed model | Compiler-aware NAS SR · Set5 x2 (mobile DSP) | PSNR: 37.34 PSNR, 32.51 ms | Improved |
| use our 11 GMACs model versus SMSR on A100 GPU | SMSR · x2 SR | latency (ms): 52 -> 11.3 | Improved |
Closest papers by meaning
- Achieving On-Mobile Real-Time Super-Resolution With Neural Architecture and Pruning Search (ICCV 2021 · similarity 0.95)
- Edge-SD-SR: Low Latency and Parameter Efficient On-device Super-Resolution with Stable Diffusion via Bidirectional Conditioning (CVPR 2025 · similarity 0.87)
- Data Overfitting for On-Device Super-Resolution with Dynamic Algorithm and Compiler Co-Design (ECCV 2024 · similarity 0.87)
- PocketSR: The Super-Resolution Expert in Your Pocket Mobiles (NeurIPS 2025 · similarity 0.85)
- AdaBM: On-the-Fly Adaptive Bit Mapping for Image Super-Resolution (CVPR 2024 · similarity 0.84)
- Achieving Lightweight Super-Resolution for Real-Time Computer Graphics (AAAI 2025 · similarity 0.84)
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.
