Research map / Image generation / Autoregressive and latent generation
InfGen: A Resolution-Agnostic Paradigm for Scalable Image Synthesis
Tao Han, Wanghan Xu, Junchao Gong, Xiaoyu Yue, Song Guo, Luping Zhou, Lei Bai
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 VAE decoder with InfGen (16x upscaling) | DiT-XL/2 · ImageNet 32x32 latent 1024x1024 | FID_P: 61.52 -> 41.75 | Improved |
| replace VAE decoder with InfGen (4x upscaling) | DiT-XL/2 · ImageNet 32x32 latent 512x512 | FID_P: 44.17 -> 39.81 | Improved |
| replace VAE decoder with InfGen | SiT-XL/2 · ImageNet 32x32 latent 1024x1024 | FID_P: 67.98 -> 46.50 | Improved |
| replace VAE decoder with InfGen | DiT-XL/2 · ImageNet 64x64 latent 3072x3072 | FID_P: 77.84 -> 45.94 | Improved |
| replace VAE decoder with InfGen | SD1.5 · text-to-image 3072x3072 | FID_P: 73.13 -> 40.75 | Improved |
| replace VAE decoder with InfGen | DiT-XL/2 · ImageNet 64x64 latent 1024x1024 | FID_P: 43.17 -> 39.06 | Improved |
| replace VAE decoder with InfGen | SD1.5 · text-to-image 1024x1024 | FID_P: 21.58 -> 16.92 | Improved |
| use InfGen+SDXL-B-1 instead of InfGen+SD1.5 | InfGen · LAION text-to-image 2048x2048 | FID_P: 139.14 -> 96.41 | Improved |
| use InfGen+SD1.5 instead of UltraPixel | InfGen · LAION text-to-image 2048x2048 | FID_P: 127.26 -> 139.14 | Hurt |
| use InfGen+SD1.5 instead of UltraPixel | InfGen · LAION text-to-image 1024x1024 | FID_P: 48.37 -> 44.85 | Improved |
| increase output resolution from 256 to 512 in reconstruction | InfGen · ImageNet-50k reconstruction | rFID: 1.07 -> 0.61 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- ScaleCrafter — ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models (ICLR 2024) · better on 2 of 2 shared results
- UltraPixel — UltraPixel: Advancing Ultra High-Resolution Image Synthesis to New Peaks (NeurIPS 2024) · better on 2 of 2 shared results
- Inf-DiT — Inf-DiT: Upsampling any-resolution image with memory-efficient diffusion transformer. (ECCV 2024) · better on 2 of 2 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
- DemoFusion: Democratising High-Resolution Image Generation With No $$$ (CVPR 2024 · similarity 0.90)
- DiffuseHigh: Training-Free Progressive High-Resolution Image Synthesis Through Structure Guidance (AAAI 2025 · similarity 0.89)
- FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion (CVPR 2025 · similarity 0.89)
- ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models (ICLR 2024 · similarity 0.88)
- Arbitrary-Scale Image Generation and Upsampling using Latent Diffusion Model and Implicit Neural Decoder (CVPR 2024 · similarity 0.88)
- RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations (NeurIPS 2026 · 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.
