Research map / Image generation / Text-to-image diffusion
InstanceControl: Controllable Complex Image Generation without Instance Labeling
XIAOYU LIU, Huan Wang, FAN LI, Zhixing Wang, Jiaqi Xu, Ming LIU, Wangmeng Zuo
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 VLM-derived instance correspondences (no manual labels) | FLUX ControlNet · MIG-Eval Canny | MIoU: 0.6526 -> 0.8250 | Improved |
| add VLM correspondences with predicted masks, no refinement (ablation) | FLUX ControlNet · MIG-Eval | average accuracy: 77.78% -> 87.97% | Improved |
| add mask refinement module (MRM) (ablation) | InstanceControl · MIG-Eval | average accuracy: 87.97% -> 90.10% | Improved |
| add interactive mask correction | InstanceControl · MIG-Eval | average accuracy: 90.10% -> 93.19% | Improved |
| add Shared SEG Token (ablation) | InstanceControl · MIG-Eval Canny | MIoU: 0.8214 -> 0.8250 | Improved |
| use MRM refined mask instead of raw predicted mask (ablation) | InstanceControl · MIG-Eval Canny | FID: 10.9284 -> 10.0264 | Improved |
| use MRM instead of score-based fusion of masks (ablation) | InstanceControl · MIG-Eval Canny | MIoU: 0.8210 -> 0.8250 | Improved |
| use InstanceControl instead of unified models (Qwen-Image ControlNet, Nano Banana) | generation · MIG-Eval 300 images | MIoU: 0.8307 -> 0.8834 | Improved |
| use InstanceControl vs FLUX ControlNet under depth | FLUX ControlNet · MIG-Eval Depth | ImageReward: 0.3216 -> 0.2961 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DreamRenderer — DreamRenderer: Taming Multi-Instance Attribute Control in Large-Scale Text-to-Image Models (ICCV 2025) · 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
- InstanceDiffusion: Instance-level Control for Image Generation (CVPR 2024 · similarity 0.90)
- MIGC: Multi-Instance Generation Controller for Text-to-Image Synthesis (CVPR 2024 · similarity 0.90)
- ROICtrl: Boosting Instance Control for Visual Generation (CVPR 2025 · similarity 0.89)
- Generating compositional scenes via Text-to-image RGBA Instance Generation (NeurIPS 2024 · similarity 0.85)
- VersaGen: Unleashing Versatile Visual Control for Text-to-Image Synthesis (AAAI 2025 · similarity 0.85)
- InstanceAssemble: Layout-Aware Image Generation via Instance Assembling Attention (NeurIPS 2025 · similarity 0.85)
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.
