Research map / 3D vision / 3D generation
OmniX: From Unified Panoramic Generation and Perception To Graphics-Ready 3D Scenes
Yukun Huang, Jiwen Yu, Yanning Zhou, Jianan Wang, Xintao Wang, Pengfei Wan, Xihui Liu
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 |
|---|---|---|---|
| use separate per-modality LoRA adapters instead of a shared branch (ablation) | OmniX perception · PanoX-Test+OutDomain albedo | PSNR: 15.29 -> 21.68 | Improved |
| use separate per-modality LoRA adapters instead of a shared adapter (ablation) | OmniX perception · PanoX-Test+OutDomain albedo | PSNR: 20.46 -> 21.68 | Improved |
| use separate adapters instead of shared adapter (ablation) | OmniX perception · PanoX-Test+OutDomain distance | AbsRel: 0.219 -> 0.154 | Improved |
| initialize from 2D generative pre-trained weights instead of scratch (ablation) | Separate-Adapter · PanoX-Test+OutDomain albedo | PSNR: 17.74 -> 21.68 | Improved |
| initialize from 2D generative pre-trained weights instead of scratch (ablation) | Separate-Adapter · PanoX-Test+OutDomain distance | AbsRel: 0.461 -> 0.154 | Improved |
| add camera ray maps as extra condition (ablation) | OmniX perception · PanoX-Test+OutDomain normal | Mean angular error: 20.58 -> 19.92 | Improved |
| add camera ray maps as extra condition (ablation) | OmniX perception · PanoX-Test+OutDomain albedo | PSNR: 21.68 -> 21.29 | No clear effect |
| model roughness and metallic jointly with cross-attention instead of channel concat (ablation) | OmniX PBR perception · PanoX-Test+OutDomain metallic | LPIPS: 0.323 -> 0.138 | Improved |
| model roughness and metallic independently instead of joint cross-attention (ablation) | OmniX PBR perception · PanoX-Test+OutDomain average | PSNR: 21.43 -> 21.40 | No clear effect |
| model distance and normal jointly instead of independently (ablation) | OmniX geometry · PanoX-Test+OutDomain distance | AbsRel: 0.155 -> 0.163 | Hurt |
| model distance and normal jointly instead of independently (ablation) | OmniX geometry · PanoX-Test+OutDomain normal | Mean angular error: 19.92 -> 20.80 | Hurt |
| apply circular synchronization instead of circular blending or latent rotation (ablation) | Flux.1-dev · panorama seam continuity | visual seam quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- MoGe — MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision (CVPR 2025) · better on 0 of 2 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
- DreamCube: RGB-D Panorama Generation via Multi-plane Synchronization (ICCV 2025 · similarity 0.88)
- 4K4DGen: Panoramic 4D Generation at 4K Resolution (ICLR 2025 · similarity 0.88)
- Omni-View: Unlocking How Generation Facilitates Understanding in Unified 3D Model based on Multiview images (ICLR 2026 · similarity 0.88)
- Pano3DComposer: Feed-Forward Compositional 3D Scene Generation from Single Panoramic Image (CVPR 2026 · similarity 0.88)
- HORIZON: High-Resolution Semantically Controlled Panorama Synthesis (AAAI 2024 · similarity 0.86)
- ViewPoint: Panoramic Video Generation with Pretrained Diffusion Models (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.
