Research map / Image generation / Text-to-image diffusion
InstaPano: Zero-shot Instance Layout Controlled Panorama Generation Via Global Attention Fusion
Zejian Li, Rui Huang, Lefan Hou, Pei Chen, Heyuan Xu, Shengyuan Zhang, Kewen Zhu, Huanghuang Deng, Li Liu, Lingyun Sun
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 global self-attention fusion for 10 steps (ablation) | InstaPano · Pano-Layout-Bench | Intra-LPIPS: 0.6140 -> 0.5752 | Improved |
| lengthen global self-attention fusion from 10 to 30 steps (ablation) | InstaPano · Pano-Layout-Bench | CLIP score: 32.19 -> 31.53 | Hurt |
| lengthen global self-attention fusion from 10 to 30 steps (ablation) | InstaPano · Pano-Layout-Bench | Intra-LPIPS: 0.5752 -> 0.5478 | Improved |
| lengthen global self-attention fusion from 0 to 30 steps (ablation) | InstaPano · Pano-Layout-Bench | AP50: 0.37 -> 0.46 | Improved |
| add global self-attention fusion for 10 steps (ablation) | InstaPano · Pano-Layout-Bench | Aesthetic score: 6.18 -> 6.12 | Hurt |
| replace MAD post-hoc sample fusion with holistic layout-aware attention fusion (ablation) | IFAdapter backbone, holistic prompt · Pano-Layout-Bench | mIoU: 0.56 -> 0.71 | Improved |
| replace MAD post-hoc sample fusion with holistic layout-aware attention fusion (ablation) | IFAdapter backbone, holistic prompt · Pano-Layout-Bench | CLIP score: 30.02 -> 32.37 | Improved |
| use InstaPano joint diffusion instead of direct IFAdapter panoramic inference | IFAdapter · Pano-Layout-Bench | mIoU: 0.69 -> 0.71 | Improved |
| use InstaPano joint diffusion instead of direct IFAdapter panoramic inference | IFAdapter · Pano-Layout-Bench | AP50: 0.47 -> 0.45 | Hurt |
| use InstaPano joint diffusion instead of direct IFAdapter panoramic inference | IFAdapter · Pano-Layout-Bench | Aesthetic score: 5.97 -> 6.12 | Improved |
| use holistic prompt instead of background-only prompt | InstaPano · Pano-Layout-Bench | mIoU: 0.63 -> 0.71 | Improved |
| use InstaPano instead of MAD | MAD · Pano-Layout-Bench | Intra-LPIPS: 0.6007 -> 0.5665 | Improved |
Closest papers by meaning
- L-MAGIC: Language Model Assisted Generation of Images with Coherence (CVPR 2024 · similarity 0.87)
- DiffPano: Scalable and Consistent Text to Panorama Generation with Spherical Epipolar-Aware Diffusion (NeurIPS 2024 · similarity 0.86)
- HORIZON: High-Resolution Semantically Controlled Panorama Synthesis (AAAI 2024 · similarity 0.86)
- Conditional Panoramic Image Generation via Masked Autoregressive Modeling (NeurIPS 2025 · similarity 0.86)
- ViewPoint: Panoramic Video Generation with Pretrained Diffusion Models (NeurIPS 2025 · similarity 0.86)
- CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation (ICLR 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.
