Research map / Video / Video segmentation and tracking
QuantVSR: Low-Bit Post-Training Quantization for Real-World Video Super-Resolution
Bowen Chai, Zheng Chen, Libo Zhu, Wenbo Li, Yong Guo, Yulun Zhang
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 STCA skip branch instead of SVDQuant-style fixed-rank branch (ablation) | QuantVSR W4A4 · SPMCS | PSNR (dB): 18.94 -> 22.75 | Improved |
| add FP skip-connection branch (SVDSC) to W4A4 model (ablation) | QuantVSR W4A4 · SPMCS | PSNR (dB): 17.13 -> 18.94 | Improved |
| add learnable bias alignment (LBA) without skip branch (ablation) | QuantVSR W4A4 · SPMCS | PSNR (dB): 17.13 -> 21.38 | Improved |
| add LBA to SVDSC (ablation) | QuantVSR W4A4 · SPMCS | PSNR (dB): 18.94 -> 22.58 | Improved |
| add LBA to STCA (ablation) | QuantVSR W4A4 · SPMCS | PSNR (dB): 22.75 -> 22.76 | No clear effect |
| add LBA to STCA (ablation) | QuantVSR W4A4 · SPMCS | LPIPS: 0.2914 -> 0.2857 | Improved |
| add LBA to STCA (ablation) | QuantVSR W4A4 · SPMCS | DOVER: 0.6886 -> 0.6969 | Improved |
| add LBA to STCA (ablation) | QuantVSR W4A4 · SPMCS | CLIP-IQA: 0.4586 -> 0.4553 | Hurt |
| use layer-specific ranks (avg 24) instead of fixed rank 32 (ablation) | QuantVSR W4A4 · MGLD-VSR UNet | average FP-branch rank: 32 -> 24 | Improved |
| use QuantVSR instead of SVDQuant at W4A4 | MGLD-VSR · REDS4 | PSNR (dB): 21.19 -> 23.31 | Improved |
| use QuantVSR instead of SVDQuant at W4A4 | MGLD-VSR · MVSR4x | PSNR (dB): 21.70 -> 22.90 | Improved |
| use QuantVSR instead of SVDQuant at W4A4 | MGLD-VSR · MVSR4x | CLIP-IQA: 0.6394 -> 0.4339 | Hurt |
Closest papers by meaning
- LSGQuant: Layer-Sensitivity Guided Quantization for One-Step Diffusion Real-World Video Super-Resolution (ICML 2026 · similarity 0.92)
- 2DQuant: Low-bit Post-Training Quantization for Image Super-Resolution (NeurIPS 2024 · similarity 0.90)
- PassionSR: Post-Training Quantization with Adaptive Scale in One-Step Diffusion based Image Super-Resolution (CVPR 2025 · similarity 0.89)
- $\text{S}^2$Q-VDiT: Accurate Quantized Video Diffusion Transformer with Salient Data and Sparse Token Distillation (NeurIPS 2025 · similarity 0.88)
- DSAQuant: Denoising-Stage-Aligned Quantization-Aware Training for Video Generation (NeurIPS 2026 · similarity 0.87)
- Q-VDiT: Towards Accurate Quantization and Distillation of Video-Generation Diffusion Transformers (ICML 2025 · 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.
