CAASR: A Real-World Animation Super-Resolution Benchmark with Color Degradation and Multi-Scale Multi-Frequency Alignment (TIP 2025)

CAASR is a benchmark designed to advance the frontier of animation super-resolution.
It features a high-quality dataset and a dedicated training pipeline for both 2D and 3D animated content, with a focus on color degradation, frequency alignment, and scale adaptability.

---

🎯 Updates

• 👌 Tools Coming Soon
• ✅ 2025.07.26 – Pretrained Weights Added
• ✅ 2025.07.25 – Dataset Added
• ✅ 2025.07.21 – Code Released

---

📖 Visualization

(Coming Soon)

---

🏄 Installation

(Instructions Coming Soon)

---

😇 ADASR Dataset

• Full Training Dataset (2D & 3D animation sequences):
  Baidu Drive (code: a135)

---

♥️ Fast Inference

• Evaluation Dataset:
  Baidu Drive (code: a135)

---

♣️ Pretrained Weights

• Pretrained Models (2D & 3D animation):
  Baidu Drive (code: a135) | Google Drive
• Weights for Comparison Methods:
  Baidu Drive (code: a135) | Google Drive

---

🎩 Training

• Configure degradation strategies and choose scale-aware architectures according to your animation content and downstream tasks.

---

🍺 Testing

• We adopt PYIQA for perceptual quality assessment.
• 2D Animation: Default configurations are applied.
• 3D Animation: Fine-tuned MANIQA and TReS models are provided.
  Baidu Drive (code: a135) | Google Drive

---

🔧 Tools

(Coming Soon)

---

🍫 Citation

If you find our work useful, please consider citing:

@article{animationSR,
  title   = {{A Real-World Animation Super-Resolution Benchmark with Color Degradation and Multi-Scale Multi-Frequency Alignment}},
  author  = {Jiang, Yu and Zhang, Yongji and Li, Siqi and Huang, Yang and Wang, Yuehang and Yao, Yutong and Gao, Yue},
  journal = {IEEE Transactions on Image Processing},
  year    = {2025}
}