Research map / Image generation / Super-resolution
Inf-DiT: Upsampling any-resolution image with memory-efficient diffusion transformer.
Zhuoyi Yang, Heyang Jiang, Wenyi Hong, Jiayan Teng, Wendi Zheng, Yuxiao Dong, Ming Ding, Jie Tang
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 unidirectional block attention between blocks (ablation) | Inf-DiT · HPDv2 2048x2048 | FID_crop: 83.7 -> 71.9 | Improved |
| add nearby LR cross attention (ablation) | Inf-DiT · HPDv2 2048x2048 | FID_crop: 73.6 -> 71.9 | Improved |
| add CLIP image embedding semantic input (ablation) | Inf-DiT · HPDv2 2048x2048 | FID_crop: 72.8 -> 71.9 | Improved |
| add semantic input (ablation) | Inf-DiT · HPDv2 2048x2048 | FID: 66.2 -> 66.0 | No clear effect |
| use Inf-DiT instead of BSRGAN at 4096x4096 | SDXL upsampling · HPDv2 4096x4096 | FID: 66.0 -> 67.0 | Hurt |
| use Inf-DiT instead of UNet SDXL architecture for 4096x4096 inference | upsampler architecture · memory at 4096x4096 | GPU memory: more than 5x memory saved | Improved |
| use Inf-DiT instead of Real-ESRGAN on DIV2K variable resolution | super-resolution · DIV2K | FID: 35.0 -> 20.2 (vs BSRGAN) | Improved |
| add iterative upsampling for 64x upscale from 32x32 | Inf-DiT · iterative upsampling (qualitative) | error accumulation | Hurt |
Later papers that compared with this method
1 later accepted paper reports Inf-DiT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- InfGen: A Resolution-Agnostic Paradigm for Scalable Image Synthesis (ICCV 2025 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DemoFusion — DemoFusion: Democratising High-Resolution Image Generation With No $$$ (CVPR 2024) · better on 4 of 4 shared results
- ScaleCrafter — ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models (ICLR 2024) · better on 1 of 1 shared result
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 0.88)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.87)
- FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion (CVPR 2025 · similarity 0.87)
- SEGA: Spectral-Energy Guided Attention for Resolution Extrapolation in Diffusion Transformers (NeurIPS 2026 · similarity 0.86)
- $\infty$-Diff: Infinite Resolution Diffusion with Subsampled Mollified States (ICLR 2024 · similarity 0.86)
- CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up (NeurIPS 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.
