Research map / 3D vision / 3D generation
Uni-3D: A Universal Model for Panoptic 3D Scene Reconstruction
Xiang Zhang, Zeyuan Chen, Fangyin Wei, Zhuowen Tu
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 Pano-Re 2D segmentation with Uni-3D segmentation (ablation) | Pano-Re · 3D-Front | PRQ: 42.60 -> 43.10 | Improved |
| replace Pano-Re depth with Uni-3D depth (ablation) | Pano-Re · 3D-Front | PRQ: 42.60 -> 32.64 | Hurt |
| replace Pano-Re 2D network with Uni-3D 2D network (ablation) | Pano-Re · 3D-Front | PRQ: 42.60 -> 46.00 | Improved |
| use unified 3D segments with 2D queries (Uni-3D 3D network) over Pano-Re 3D network, same 2D network (ablation) | Pano-Re w/ our 2D network · 3D-Front | PRQ: 46.00 -> 52.48 | Improved |
| add multi-scale feature lifting (ablation) | Uni-3D · 3D-Front | PRQ: 52.48 -> 53.54 | Improved |
| add instance propagation (ablation) | Pano-Re · 3D-Front | PRQ: 19.76 -> 42.60 | Improved |
| add 2D semantic segmentation map to back-projection (ablation) | Uni-3D · 3D-Front | PRQ: 52.15 -> 52.48 | Improved |
| add depth distance-field embedding (ablation) | Uni-3D · 3D-Front | PRQ: 52.44 -> 52.48 | No clear effect |
| add mask feature P^S(F4) (ablation) | Uni-3D · 3D-Front | PRQ: 52.03 -> 52.48 | Improved |
| add depth feature P^D(F4) (ablation) | Uni-3D · 3D-Front | PRQ: 52.32 -> 52.48 | Improved |
| use ground-truth depth instead of predicted depth (ablation) | Uni-3D · 3D-Front | PRQ: 52.48 -> 56.13 | Improved |
| add learned 2D queries (vs random learnable queries) (ablation) | Uni-3D · 3D-Front | PRQ: 51.67 -> 52.48 | Improved |
Later papers that compared with this method
2 later accepted papers report Uni-3D as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- PixARMesh: Autoregressive Mesh-Native Single-View Scene Reconstruction (CVPR 2026 · better on 2 of 2 shared results)
- Describe, Adapt and Combine: Empowering CLIP Encoders for Open-set 3D Object Retrieval (ICCV 2025 · better on 4 of 4 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
- PanoRecon: Real-Time Panoptic 3D Reconstruction from Monocular Video (CVPR 2024 · similarity 0.89)
- A Unified Framework for 3D Scene Understanding (NeurIPS 2024 · similarity 0.89)
- UniK3D: Universal Camera Monocular 3D Estimation (CVPR 2025 · similarity 0.88)
- Uni3D: Exploring Unified 3D Representation at Scale (ICLR 2024 · similarity 0.88)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.88)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · similarity 0.87)
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.
