Research map / Image generation / Text-to-image diffusion
PanGu-Draw: Advancing Resource-Efficient Text-to-Image Synthesis with Time-Decoupled Training and Reusable Coop-Diffusion
Guansong Lu, Yuanfan Guo, Jianhua Han, Minzhe Niu, Yihan Zeng, Songcen Xu, Zeyi Huang, Zhao Zhong, Wei Zhang, Hang Xu
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 |
|---|---|---|---|
| use time-decoupled structure and texture generators instead of a single resolution-boost model (ablation) | PanGu-Draw · COCO (ablation, LAION subset) | FID: 106.12 -> 87.66 | Improved |
| train structure generator on all data including up-scaled low-res images instead of only high-resolution (ablation) | PanGu-Draw structure generator · COCO (ablation) | FID: 89.52 -> 87.66 | Improved |
| train texture generator at 512 instead of 256 (ablation) | PanGu-Draw texture generator · COCO (ablation) | FID: 87.66 -> 90.98 | Hurt |
| set timestep splitting point T_struct to 500 versus 200 (ablation) | PanGu-Draw · COCO (ablation) | FID: 105.08 -> 87.66 | Improved |
| increase T_struct from 500 to 700 (ablation) | PanGu-Draw · COCO (ablation) | FID: 87.66 -> 89.48 | Hurt |
| use bilingual text encoder (Chinese plus English) in PanGu-Draw | PanGu-Draw · COCO-CN | FID: 23.99 -> 21.81 | Improved |
| PanGu-Draw 5B versus DALL-E 3 in user study | PanGu-Draw · ImageEval-prompt user study | average score: 4.69 vs 4.58 | Hurt |
| PanGu-Draw versus SDXL on COCO FID | PanGu-Draw · COCO (30k) | FID: 11.93 -> 7.99 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- AltDiffusion — AltDiffusion: A Multilingual Text-to-Image Diffusion Model (AAAI 2024) · better on 2 of 2 shared results
- RAPHAEL — RAPHAEL: Text-to-Image Generation via Large Mixture of Diffusion Paths (NeurIPS 2023) · better on 0 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
- SwiftBrush: One-Step Text-to-Image Diffusion Model with Variational Score Distillation (CVPR 2024 · similarity 0.90)
- CtrLoRA: An Extensible and Efficient Framework for Controllable Image Generation (ICLR 2025 · similarity 0.90)
- Würstchen: An Efficient Architecture for Large-Scale Text-to-Image Diffusion Models (ICLR 2024 · similarity 0.90)
- Any-Size-Diffusion: Toward Efficient Text-Driven Synthesis for Any-Size HD Images (AAAI 2024 · similarity 0.90)
- PixArt-$\alpha$: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis (ICLR 2024 · similarity 0.90)
- DiffuseHigh: Training-Free Progressive High-Resolution Image Synthesis Through Structure Guidance (AAAI 2025 · similarity 0.89)
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.
