Research map / Image generation / Autoregressive and latent generation
Envisioning Beyond the Few: Disentangled Semantics and Primitives for Few-Shot Atypical Layout-to-Image Generation
Nan Bao, Yifan Zhao, Wenzhuang Wang, Jia Li
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 Semantic Anchoring, Primitive Imbuing and Conceptual Steering together (ablation) | layout-to-image SD1.5 baseline · DIOR 5-shot | Bootstrap FID: 94.96 -> 74.34 | Improved |
| add Conceptual Steering on top of SA and PI (ablation) | SA + PI model · DIOR 5-shot | Bootstrap FID: 85.00 -> 74.34 | Improved |
| add Conceptual Steering on top of SA and PI (ablation) | SA + PI model · DIOR 5-shot | mAP: 25.28 -> 26.06 | Improved |
| enable SA or PI individually and then combine them (ablation) | SD1.5 baseline · DIOR 5-shot | Bootstrap FID: 94.96 -> 85.00 (SA+PI) | Improved |
| use full number of SA injection layers rather than reduced (ablation) | Ours full model · DIOR 5-shot | mAP: 17.31 -> 26.06 | Improved |
| use full number of PI injection layers rather than reduced (ablation) | Ours full model · DIOR 5-shot | Bootstrap FID: 76.48 -> 74.34 | Improved |
| keep SA in middle/early and PI in late upsampling blocks rather than swapping them (ablation) | Ours full model · DIOR 5-shot | mAP: 11.89 -> 26.06 | Improved |
| learn primitives with ridge-regression alternating optimization instead of random init (ablation) | Primitive Imbuing · DIOR 5-shot | Bootstrap FID: 75.08 -> 74.34 | Improved |
| learn primitives with ridge-regression alternating optimization instead of fixed K-means (ablation) | Primitive Imbuing · DIOR 5-shot | Bootstrap FID: 78.23 -> 74.34 | Improved |
| vary number of visual primitives (32, 64, 256 vs 128) (ablation) | Primitive Imbuing · DIOR 5-shot | Bootstrap FID: 76.38 (32), 76.03 (64), 74.34 (128), 76.30 (256) | Mixed |
| use proposed method vs CC-Diff on aerial domain | Ours · DIOR 5-shot | Bootstrap FID: 82.51 -> 74.34 | Improved |
| evaluate with YOLOv8 detector instead of Faster R-CNN | Ours vs CC-Diff · DIOR 5-shot | mAP: 19.50 -> 20.80 | Improved |
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.
- MIGC — MIGC: Multi-Instance Generation Controller for Text-to-Image Synthesis (CVPR 2024) · 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
- PLACE: Adaptive Layout-Semantic Fusion for Semantic Image Synthesis (CVPR 2024 · similarity 0.88)
- Rethinking The Training And Evaluation of Rich-Context Layout-to-Image Generation (NeurIPS 2024 · similarity 0.88)
- Think, Plan, Paint: Layout-Aware Reasoning for Controllable Image Generation in Unified Models (NeurIPS 2026 · similarity 0.87)
- InstanceAssemble: Layout-Aware Image Generation via Instance Assembling Attention (NeurIPS 2025 · similarity 0.86)
- Control and Realism: Best of Both Worlds in Layout-to-Image without Training (ICML 2025 · similarity 0.86)
- SSMG: Spatial-Semantic Map Guided Diffusion Model for Free-Form Layout-to-Image Generation (AAAI 2024 · similarity 0.86)
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.
