Research map / Image generation / Text-to-image diffusion
Cocktail: Mixing Multi-Modality Control for Text-Conditional Image Generation
Minghui Hu, Jianbin Zheng, Daqing Liu, Chuanxia Zheng, Chaoyue Wang, Dacheng Tao, Tat-Jen Cham
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 ControlNorm to Cocktail (ablation) | Cocktail · COCO5k validation | LPIPS: 0.4900 $\pm 0.00141$ -> 0.4836 $\pm 0.00133$ | Improved |
| add ControlNorm to Cocktail (ablation) | Cocktail · COCO5k validation | Sketch L2 Distance: 7.18413 $\pm 0.01453$ -> 7.28929 $\pm 0.01385$ | Hurt |
| add ControlNorm to Cocktail (ablation) | Cocktail · COCO5k validation | Segmentation mPA: 48.26 $\pm 0.287$ -> 49.20 $\pm 0.289$ | Improved |
| add ControlNorm to Cocktail (ablation) | Cocktail · COCO5k validation | Segmentation mIoU: 32.66 $\pm 0.272$ -> 33.27 $\pm 0.271$ | Improved |
| add ControlNorm to Cocktail (ablation) | Cocktail · COCO5k validation | Pose mAP: 61.93 $\pm 0.775$ -> 61.99 $\pm 0.778$ | Improved |
| use Cocktail instead of ControlNet | Text + Edge generation · COCO validation | FID: 18.73 -> 16.66 | Improved |
| use Cocktail instead of ControlNet | Text + Segmentation Map generation · COCO validation | FID: 20.53 -> 23.88 | Hurt |
| use Cocktail instead of ControlNet | Text + Pose generation · COCO validation | FID: 35.13 -> 39.64 | Hurt |
| use Cocktail instead of T2I-Adapter | Text + Edge generation · COCO validation | CLIP Score: 0.2638 -> 0.2561 | Hurt |
| use Cocktail instead of ControlNet | Text + Segmentation Map generation · COCO validation | Aesthetic Score: 5.43 -> 5.64 | Improved |
| use Cocktail instead of T2I-Adapter | Text + Pose generation · COCO validation | ImageReward: 0.3810 -> 0.4488 | Improved |
| apply spatial guidance sampling | Cocktail · Spatially controlled image generation | regional preservation | Improved |
Closest papers by meaning
- FlexControl: Computation-Aware Conditional Control with Differentiable Router for Text-to-Image Generation (ICML 2025 · similarity 0.89)
- EMControl: Adding Conditional Control to Text-to-Image Diffusion Models via Expectation-Maximization (AAAI 2025 · similarity 0.89)
- Simplifying Control Mechanism in Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.89)
- CtrLoRA: An Extensible and Efficient Framework for Controllable Image Generation (ICLR 2025 · similarity 0.89)
- Local Conditional Controlling for Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.88)
- Universal Few-shot Spatial Control for Diffusion Models (NeurIPS 2025 · 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.
