Research map / Image generation / Text-to-image diffusion
MIGC++: Advanced Multi-Instance Generation Controller for Image Synthesis
Dewei Zhou, You Li, Fan Ma, Zongxin Yang, Yi Yang
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 Enhance Attention (ablation) | MIGC · COCO-Position | Success Ratio: 34.7 -> 80.2 | Improved |
| add Layout Attention to Enhance Attention + SAC (ablation) | MIGC · COCO-Position | Success Ratio: 80.2 -> 80.3 | No clear effect |
| add Shading Aggregation Controller (ablation) | MIGC · COCO-Position | Success Ratio: 78.1 -> 80.3 | Improved |
| add Refined Shader (MIGC++ over MIGC) | MIGC · COCO-MIG-BOX | Instance Success Ratio: 69.2 -> 71.4 | Improved |
| add Refined Shader (MIGC++ over MIGC) | MIGC · COCO-Position | Success Ratio: 80.29 -> 82.54 | Improved |
| deploy Instance Shader in mid and up-1 instead of mid only (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 46.9 -> 71.4 | Improved |
| also deploy Instance Shader in up-2 (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 71.4 -> 66.5 | Hurt |
| also deploy Instance Shader in down-2 (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 71.4 -> 66.4 | Hurt |
| deploy Instance Shader in all blocks (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 71.4 -> 65.7 | Hurt |
| control first 50% of sampling steps instead of 30% (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 66.9 -> 71.4 | Improved |
| extend control from 50% to 60% of steps (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 71.4 -> 69.8 | Hurt |
| deploy Refined Shader in all blocks vs none (ablation) | MIGC++ · COCO-MIG-BOX | Instance Success Ratio: 67.2 -> 71.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MIGC — MIGC: Multi-Instance Generation Controller for Text-to-Image Synthesis (CVPR 2024) · better on 7 of 7 shared results
- InstanceDiffusion — InstanceDiffusion: Instance-level Control for Image Generation (CVPR 2024) · better on 6 of 6 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
- MIGC: Multi-Instance Generation Controller for Text-to-Image Synthesis (CVPR 2024 · similarity 0.96)
- InstanceDiffusion: Instance-level Control for Image Generation (CVPR 2024 · similarity 0.88)
- InstanceAssemble: Layout-Aware Image Generation via Instance Assembling Attention (NeurIPS 2025 · similarity 0.86)
- Composing Parts for Expressive Object Generation (CVPR 2025 · similarity 0.86)
- ROICtrl: Boosting Instance Control for Visual Generation (CVPR 2025 · similarity 0.86)
- Generating compositional scenes via Text-to-image RGBA Instance Generation (NeurIPS 2024 · 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.
