Research map / 3D vision / Depth and stereo
OmniX: Any-view and Any-time 4D reconstruction via Feed-forward Trajectory Fields
Yanqin Jiang, Tengfei Wang, Zhenwei Wang, Chenjie Cao, Junta Wu, Wenhan Luo, Weiming Hu, Jin Gao, Chunchao Guo
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 proposed temporal embedding instead of AdaLN (ablation) | OmniX SSA · SSA ablation monocular | FG APD3D: 0.260 -> 0.293 | Improved |
| use proposed temporal embedding instead of AdaLN (ablation) | OmniX SSA · SSA ablation hybrid | FG APD3D: 0.291 -> 0.324 | Improved |
| add self attention among dynamic tokens (ablation) | OmniX SSA · SSA ablation monocular | FG APD3D: 0.297 -> 0.293 | No clear effect |
| add self attention among dynamic tokens (ablation) | OmniX SSA · SSA ablation disjoint and hybrid | FG APD3D: 0.298 -> 0.298; 0.324 -> 0.324 | No clear effect |
| add cross attention (ablation) | OmniX SSA · SSA ablation monocular | FG APD3D: 0.171 -> 0.297 | Improved |
| add cross attention (ablation) | OmniX SSA · SSA ablation disjoint | FG APD3D: 0.197 -> 0.298 | Improved |
| add cross attention (ablation) | OmniX SSA · SSA ablation hybrid | FG APD3D: 0.205 -> 0.324 | Improved |
| add sparse design (ablation) | OmniX · 16-frame efficiency | Runtime (s): 4.10 -> 2.15 | Improved |
| add sparse design (ablation) | OmniX · 16-frame efficiency | FLOPs (T): 15.25 -> 12.07 | Improved |
| increase top-rho percent of dynamic tokens from 10 to 30 (ablation) | OmniX · 4-frame monocular | APD3D: 0.3194 -> 0.3454 | Improved |
| increase training data size (ablation) | OmniX · 16-frame monocular | APD3D: 0.182 -> 0.243 | Improved |
| train with large camera motion data and disentangle dynamics | OmniX versus VDPM · Dense trajectories, disjoint pairs (first frame) | FG APD3D: 0.164 -> 0.322 | 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.
- VDPM — V-DPM: 4D Video Reconstruction with Dynamic Point Maps (CVPR 2026) · better on 13 of 13 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
- Omnidirectional 3D Scene Reconstruction from Single Image (NeurIPS 2025 · similarity 0.89)
- ODGS: 3D Scene Reconstruction from Omnidirectional Images with 3D Gaussian Splattings (NeurIPS 2024 · similarity 0.86)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.86)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.86)
- OmniLocalRF: Omnidirectional Local Radiance Fields from Dynamic Videos (CVPR 2024 · similarity 0.86)
- Omni-Scene: Omni-Gaussian Representation for Ego-Centric Sparse-View Scene Reconstruction (CVPR 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.
