Research map / Image generation / Super-resolution
Dynamic Dual Trainable Bounds for Ultra-Low Precision Super-Resolution Networks
Yunshan Zhong, Mingbao Lin, Xunchao Li, Ke Li, Yunhang Shen, Fei Chao, Yongjian Wu, 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 |
|---|---|---|---|
| add DDTB dynamic dual bounds for activations (ablation) | PAMS 2-bit EDSR x4 · Urban100 | PSNR (dB): 23.72 -> 24.06 | Improved |
| add dual bounds for weights (DW) (ablation) | PAMS 2-bit EDSR x4 · Urban100 | PSNR (dB): 23.72 -> 23.98 | Improved |
| add DDTB initializer to DDTB (ablation) | DDTB 2-bit EDSR x4 · Urban100 | PSNR (dB): 24.06 -> 24.23 | Improved |
| add DW to DDTB (ablation) | DDTB 2-bit EDSR x4 · Urban100 | PSNR (dB): 24.06 -> 24.64 | Improved |
| add initializer to DDTB+DW (full method) (ablation) | DDTB+DW 2-bit EDSR x4 · Urban100 | PSNR (dB): 24.64 -> 24.82 | Improved |
| add DDTB and DW to PAMS together (ablation) | PAMS 2-bit EDSR x4 · Set5 | PSNR (dB): 29.51 -> 30.80 | Improved |
| apply dynamic gate to top 30% layers (best P) versus no gate (ablation) | DDTB 2-bit EDSR x4 · Set14 | PSNR (dB): +0.16dB | Improved |
| apply gate to more layers than P=30 (ablation) | DDTB 2-bit EDSR x4 · Set14 | PSNR (dB) | Hurt |
| use full-precision gate instead of 2-bit gate (ablation) | DDTB 2-bit EDSR x4 · Set14 | PSNR (dB) | No clear effect |
| compare DDTB to PAMS at 2-bit | EDSR x4 · Urban100 | PSNR (dB): 23.72 -> 24.82 | Improved |
| compare DDTB to PAMS at 4-bit | EDSR x4 · Urban100 | PSNR (dB): 25.32 -> 25.69 | Improved |
| compare DDTB to FQSR in fully quantized setting | EDSR x4 4-bit · Urban100 | PSNR (dB): 24.93 -> 25.70 | Improved |
Closest papers by meaning
- QuantSR+: Pushing the Limit of Quantized Image Super-Resolution Networks (TPAMI 2026 · similarity 0.90)
- QuantSR: Accurate Low-bit Quantization for Efficient Image Super-Resolution (NeurIPS 2023 · similarity 0.89)
- Thinking in Granularity: Dynamic Quantization for Image Super-Resolution by Intriguing Multi-Granularity Clues (AAAI 2025 · similarity 0.88)
- 2DQuant: Low-bit Post-Training Quantization for Image Super-Resolution (NeurIPS 2024 · similarity 0.88)
- AdaBM: On-the-Fly Adaptive Bit Mapping for Image Super-Resolution (CVPR 2024 · similarity 0.87)
- Edge-SD-SR: Low Latency and Parameter Efficient On-device Super-Resolution with Stable Diffusion via Bidirectional Conditioning (CVPR 2025 · 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.
