Research map / Image generation / Super-resolution
DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution
Zheng-Peng Duan, Jiawei Zhang, Xin Jin, Ziheng Zhang, Zheng Xiong, Dongqing Zou, Jimmy S. Ren, Chunle Guo, Chongyi Li
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 |
|---|---|---|---|
| integrate LR stream into the attention of DiT blocks (bidirectional interaction) (ablation) | DiT4SR · RealLQ250 | MUSIQ: 66.963 -> 71.832 | Improved |
| add LR Residual shortcut around LR-stream attention (ablation) | DiT4SR · RealLQ250 | MUSIQ: 70.887 -> 71.832 | Improved |
| add LR Injection between MLPs via depth-wise conv (ablation) | DiT4SR · RealLQ250 | MANIQA: 0.610 -> 0.632 | Improved |
| use 3x3 depth-wise conv instead of linear layer for LR injection (ablation) | DiT4SR · RealLQ250 | MUSIQ: 71.607 -> 71.832 | No clear effect |
| replace ControlNet-style one-way injection with DiT4SR bidirectional design | SD3 · RealLQ250 | MUSIQ: 66.385 -> 71.832 | Improved |
| DiT4SR versus SeeSR and DiffBIR on small paired benchmarks | DiT4SR · RealSR | LIQE: 3.977 vs 4.123 (SeeSR) | Hurt |
| DiT4SR fidelity LPIPS versus UNet baselines | DiT4SR · DrealSR | LPIPS: 0.365 vs 0.317 (SeeSR) | Hurt |
Later papers that compared with this method
7 later accepted papers report DiT4SR as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Low-Resolution Editing is All You Need for High-Resolution Editing (CVPR 2026 · better on 6 of 6 shared results)
- Fill2SR: Repurposing Inpainting Diffusion Transformers for Real-World Super-Resolution (ECCV 2026 · better on 4 of 5 shared results)
- FRAMER: Frequency-Aligned Self-Distillation with Adaptive Modulation Leveraging Diffusion Priors for Real-World Image Super-Resolution (CVPR 2026 · better on 4 of 4 shared results)
- LucidNFT: LR-Anchored Multi-Reward Preference Optimization for Flow-Based Real-World Super-Resolution (NeurIPS 2026 · better on 3 of 3 shared results)
- DreamSR: Towards Ultra-High-Resolution Image Super-Resolution via a Receptive-Field Enhanced Diffusion Transformer (CVPR 2026 · better on 1 of 3 shared results)
- VOSR: A Vision-Only Generative Model for Image Super-Resolution (CVPR 2026 · better on 2 of 2 shared results)
- Coloring the Noise: Adversarial Sobolev Alignment for Faithful Image Super Resolution (ICML 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SeeSR — SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution (CVPR 2024) · better on 2 of 5 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
- Effective Diffusion Transformer Architecture for Image Super-Resolution (AAAI 2025 · similarity 0.94)
- Unleashing the Power of One-Step Diffusion based Image Super-Resolution via a Large-Scale Diffusion Discriminator (NeurIPS 2025 · similarity 0.92)
- When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution (NeurIPS 2026 · similarity 0.92)
- Taming Diffusion Prior for Image Super-Resolution with Domain Shift SDEs (NeurIPS 2024 · similarity 0.91)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.91)
- One Diffusion Step to Real-World Super-Resolution via Flow Trajectory Distillation (ICML 2025 · similarity 0.91)
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.
