Research map / Video / Video segmentation and tracking
Data Overfitting for On-Device Super-Resolution with Dynamic Algorithm and Compiler Co-Design
Gen Li, zhihao shu, Jie Ji, Minghai Qin, Fatemeh Afghah, Wei Niu, Xiaolong Ma
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 dynamic multi-path network with uneven patch splitting (one model) instead of STDO multi-model | Dy-DCA · UVG and VSD4K videos | PSNR: 47.50 -> 47.64 | Improved |
| reduce transmitted models to one with uneven patches | Dy-DCA vs STDO · mobile deployment | model switching and I/O overhead: 4x saving on model switching, 7x saving on I/O | Improved |
| add compiler data-flow analysis with fusion and static execution planning | Dy-DCA on MNN · mobile CPU/GPU latency and memory | speedup and memory: 1.7x speedup and 1.61x memory reduction | Improved |
| increase number of paths from 2 to 4 (ablation) | Dy-DCA (WDSR) · PSNR vs latency | PSNR and latency: PSNR 45.86 -> 46.51; max latency 32 -> 42 ms | Mixed |
| Dy-DCA instead of single-model SRVC | WDSR x2 · game-45s | PSNR: 31.03 -> 45.86 | Improved |
| apply to longer videos (2-min game, combined 90s) (ablation) | Dy-DCA (WDSR x4) · VSD4K long video | PSNR: game-2min 44.72; combine-90s 47.34 | No clear effect |
Closest papers by meaning
- Towards High-Quality and Efficient Video Super-Resolution via Spatial-Temporal Data Overfitting (CVPR 2023 · similarity 0.94)
- Deep Video Super-Resolution Network Using Dynamic Upsampling Filters Without Explicit Motion Compensation (CVPR 2018 · similarity 0.88)
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network (CVPR 2016 · similarity 0.87)
- Accelerating the Super-Resolution Convolutional Neural Network (ECCV 2016 · similarity 0.87)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.86)
- Arbitrary-Scale Video Super-resolution Guided by Dynamic Context (AAAI 2024 · similarity 0.85)
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.
