Research map / Image generation / Text-to-image diffusion
SpeedUpNet: A Plug-and-Play Adapter Network for Accelerating Text-to-Image Diffusion Models
Weilong Chai, Dandan Zheng, Jiajiong Cao, Zhiquan Chen, Changbao Wang, Chenguang Ma
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 Multi-Step Consistency loss (p=0.1) vs none (ablation) | SUN · Realistic Vision v5.1, 4-step FID | FID: 22.41 -> 19.60 | Improved |
| add Multi-Step Consistency loss (p=0.1) vs none (ablation) | SUN · Anything v5, 8-step FID | FID: 20.98 -> 16.17 | Improved |
| raise MSC replacement probability p from 0.1 to 0.25 (ablation) | SUN · Realistic Vision v5.1, 8-step FID | FID: 15.73 -> 17.22 | Hurt |
| add Attention Normalization to the adapter (ablation) | SUN · Realistic Vision V5.1 and Rev Animated | generation stability (qualitative) | Improved |
| plug SUN adapter into fine-tuned SD models without training vs LCM-LoRA | SUN · Anything v5, 4-step FID vs 25-step reference | FID: 26.34 -> 22.52 | Improved |
| use 18.5M-param adapter instead of full-network distillation (LCM, 860M) | SUN on SD v1.5 · LAION-Aesthetics-6+, 4-step FID | FID: 13.23 vs LCM 11.10 at 4 steps; 11.98 vs 12.02 at 12 steps | Mixed |
| reduce from 25-step DPM-Solver++ with CFG to 4-step SUN | SD v1.5 inference · V100 FP32 UNet time | seconds: 3.16 -> 0.274 | Improved |
| SUN vs LCM-LoRA inference speed | SUN · V100 FP32 UNet time | seconds: 0.317 -> 0.274 | Improved |
Closest papers by meaning
- Not All Prompts Are Made Equal: Prompt-based Pruning of Text-to-Image Diffusion Models (ICLR 2025 · similarity 0.91)
- FlashEval: Towards Fast and Accurate Evaluation of Text-to-image Diffusion Generative Models (CVPR 2024 · similarity 0.90)
- Würstchen: An Efficient Architecture for Large-Scale Text-to-Image Diffusion Models (ICLR 2024 · similarity 0.90)
- On improved Conditioning Mechanisms and Pre-training Strategies for Diffusion Models (NeurIPS 2024 · similarity 0.90)
- On Discrete Prompt Optimization for Diffusion Models (ICML 2024 · similarity 0.89)
- On the Scalability of Diffusion-based Text-to-Image Generation (CVPR 2024 · 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.
