Research map / Image generation / Editing and style transfer
A Geometric Perspective on Optimizing Vector Quantized Latent Diffusion Model for Image Restoration
Chen Hang, Haoming Chen, Xuwei Fang, Weisheng Xie, Xiangxiang Gao, Faming Fang, Guixu Zhang, Haichuan Song
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 zq with interpolation zlerp as diffusion input (train-free) (ablation) | ResShift · Super Resolution | PSNR: 23.0160 -> 23.0356 | Improved |
| replace zq with interpolation zlerp (train-free, lambda 0.5) (ablation) | ResShift · Blind Face Restoration | PSNR: 21.7634 -> 20.6089 | Hurt |
| use interpolation zlerp during training instead of zq (ablation) | VQ-LDM-4 · CSD | PSNR: 25.96 -> 27.80 | Improved |
| use interpolation zlerp instead of zq (ablation) | VQ-LDM-4 · RainDrop | SSIM: 0.8113 -> 0.7950 | Hurt |
| use interpolation zlerp instead of zq (ablation) | VQ-LDM-4 · RESIDE | SSIM: 0.9138 -> 0.9054 | Hurt |
| use interpolation zlerp instead of zq (ablation) | VQ-LDM-4 · NTURain | PSNR: 17.58 -> 26.96 | Improved |
| use Chebyshev center instead of zq (ablation) | VQ-LDM-4 · CSD | PSNR: 25.96 -> 28.16 | Improved |
| use Chebyshev center instead of interpolation zlerp (ablation) | VQ-LDM-4 · RESIDE | SSIM: 0.9054 -> 0.9178 | Improved |
| use Chebyshev center instead of interpolation zlerp (ablation) | VQ-LDM-4 · DAVIS | PSNR: 26.58 -> 27.94 | Improved |
| use Chebyshev center instead of interpolation zlerp (ablation) | VQ-LDM-4 · GoPro | PSNR: 26.20 -> 26.53 | Improved |
| use Chebyshev center instead of interpolation (training cost) | VQ-LDM-4 · training time | training time: approx 2x increase | Hurt |
| use zq instead of ze as latent initial state (ablation) | VQ-LDM-4 · NTURain | PSNR: 18.75 -> 17.58 | Hurt |
Closest papers by meaning
- EfficientDM: Efficient Quantization-Aware Fine-Tuning of Low-Bit Diffusion Models (ICLR 2024 · similarity 0.89)
- TFMQ-DM: Temporal Feature Maintenance Quantization for Diffusion Models (CVPR 2024 · similarity 0.88)
- LMD: Faster Image Reconstruction with Latent Masking Diffusion (AAAI 2024 · similarity 0.88)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.88)
- TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models (AAAI 2025 · similarity 0.88)
- Diffusion Priors for Variational Likelihood Estimation and Image Denoising (NeurIPS 2024 · similarity 0.88)
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.
