Research map / Image generation / Text-to-image diffusion
Linear Fusion MultiDiffusion for Fast Training-Free Spherical Panorama Generation
Akio Hayakawa, Yusuke Mukuta, Tatsuya Harada
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 |
|---|---|---|---|
| replace direct-sampling MultiDiffusion with linear-fusion least-squares aggregation | SphereDiff-style training-free panorama · Panorama generation | runtime: 32m 15s -> 2m 6s | Improved |
| use LF-MultiDiffusion instead of SphereDiff | Flux panorama generation · Panorama generation | CLIP score: 27.65 -> 29.19 | Improved |
| use LSMR instead of PCG solver (ablation) | LF-MultiDiffusion · Panorama generation | QAlign: 3.04 -> 3.34 | Improved |
| increase solver iterations from 10 to 30 (ablation) | LF-MultiDiffusion (LSMR) · Panorama generation | QAlign: 3.27 -> 3.34 | Improved |
| increase solver iterations from 30 to 100 (ablation) | LF-MultiDiffusion (LSMR) · Panorama generation | QAlign: 3.34 -> 3.30 | Hurt |
| add Laplacian regularizer lambda=1e-4 (ablation) | LF-MultiDiffusion · Panorama generation | QAlign: 3.282 -> 3.341 | Improved |
| ridge vs Laplacian regularizer (ablation) | LF-MultiDiffusion · Panorama generation | QAlign: 3.340 vs 3.341 | No clear effect |
| increase views from 6 to 14 (ablation) | LF-MultiDiffusion · Panorama generation | Distortion: 3.98 -> 4.64 | Improved |
| increase views from 6 to 14 (ablation) | LF-MultiDiffusion · Panorama generation | QAlign: 3.549 -> 3.341 | Hurt |
| increase views from 14 to 26 (ablation) | LF-MultiDiffusion · Panorama generation | runtime: 2m 6s -> 3m 18s | Hurt |
| use backward instead of forward projection (ablation) | LF-MultiDiffusion ERP · Panorama generation | runtime: 4m 42s -> 2m 6s | Improved |
| use spherical latent instead of ERP latent (ablation) | LF-MultiDiffusion · Panorama generation | Aesthetic: 0.616 -> 0.631 | 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.
- SphereDiff — SphereDiff: Tuning-free 360° Static and Dynamic Panorama Generation via Spherical Latent Representation (AAAI 2026) · better on 6 of 7 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
- DiffPano: Scalable and Consistent Text to Panorama Generation with Spherical Epipolar-Aware Diffusion (NeurIPS 2024 · similarity 0.88)
- SphereDiffusion: Spherical Geometry-Aware Distortion Resilient Diffusion Model (AAAI 2024 · similarity 0.87)
- MultiDiff: Consistent Novel View Synthesis from a Single Image (CVPR 2024 · similarity 0.87)
- CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation (ICLR 2025 · similarity 0.87)
- Spherical-Nested Diffusion Model for Panoramic Image Outpainting (ICML 2025 · similarity 0.87)
- Conditional Panoramic Image Generation via Masked Autoregressive Modeling (NeurIPS 2025 · 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.
