Research map / Video / Video segmentation and tracking
Learning Data-Driven Vector-Quantized Degradation Model for Animation Video Super-Resolution
Zixi Tuo, Huan Yang, Jianlong Fu, Yujie Dun, Xueming Qian
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 |
|---|---|---|---|
| replace basic-operator-only degradation with VQ degradation codebook (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3768 -> 0.3857 | Improved |
| use multi-scale VQGAN instead of single-scale (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3466 -> 0.3857 | Improved |
| use two-stage training for multi-scale VQGAN instead of single-stage (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3430 -> 0.3857 | Improved |
| use stochastic top-k VQ instead of nearest-neighbor search (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3770 -> 0.3857 | Improved |
| increase max degradation level K in top-k from 50 to 100 (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3857 -> 0.3756 | Hurt |
| add VQ degradation to Real-ESRGAN* backbone | Real-ESRGAN* · AVC-RealLQ | MANIQA: 0.3825 -> 0.3967 | Improved |
| add VQ degradation to RealBasicVSR* backbone | RealBasicVSR* · AVC-RealLQ | MANIQA: 0.3504 -> 0.3849 | Improved |
| add VQ degradation to AnimeSR backbone | AnimeSR · AVC-RealLQ | MANIQA: 0.3832 -> 0.3857 | Improved |
| add HR-SR enhancement of training HR videos (Real-ESRGAN S) (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.3857 -> 0.4096 | Improved |
| add HR-SR enhancement when training RealBasicVSR | RealBasicVSR · AVC-RealLQ | MANIQA: 0.3504 -> 0.3855 | Improved |
| add HR-SR enhancement when training AnimeSR | AnimeSR · AVC-RealLQ | MANIQA: 0.3832 -> 0.4079 | Improved |
| use larger Real-ESRGAN (L) instead of small (S) for HR enhancement (ablation) | VQD-SR (base) · AVC-RealLQ | MANIQA: 0.4096 -> 0.4211 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AnimeSR — AnimeSR: Learning Real-World Super-Resolution Models for Animation Videos (NeurIPS 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AnimeSR: Learning Real-World Super-Resolution Models for Animation Videos (NeurIPS 2022 · similarity 0.93)
- Investigating Tradeoffs in Real-World Video Super-Resolution (CVPR 2022 · similarity 0.91)
- Real-World Video Super-Resolution: A Benchmark Dataset and a Decomposition Based Learning Scheme (ICCV 2021 · similarity 0.89)
- NegVSR: Augmenting Negatives for Generalized Noise Modeling in Real-world Video Super-Resolution (AAAI 2024 · similarity 0.89)
- DSAQuant: Denoising-Stage-Aligned Quantization-Aware Training for Video Generation (NeurIPS 2026 · similarity 0.87)
- VideoGigaGAN: Towards Detail-rich Video Super-Resolution (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.
