Research map / 3D vision / Gaussian splatting
MoRe: Motion-aware Feed-forward 4D Reconstruction Transformer
Juntong Fang, Zequn Chen, Weiqi Zhang, Donglin Di, Xuancheng Zhang, Chengmin Yang, Yu-Shen 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 |
|---|---|---|---|
| add attention forcing during training (ablation) | MoRe · Sintel camera pose | ATE: 0.163 -> 0.147 | Improved |
| add attention forcing during training (ablation) | MoRe · TUM-dynamics camera pose | ATE: 0.028 -> 0.026 | Improved |
| add BA-like refinement with duplicated camera tokens (ablation) | MoRe · Sintel camera pose | ATE: 0.155 -> 0.147 | Improved |
| add BA-like refinement (ablation) | MoRe · TUM-dynamics camera pose | ATE: 0.027 -> 0.026 | Improved |
| use grouped causal attention instead of standard causal attention (ablation) | MoRe · Sintel video depth | Abs Rel: 0.277 -> 0.254 | Improved |
| use grouped causal attention instead of standard causal attention (ablation) | MoRe · kitti video depth | Abs Rel: 0.079 -> 0.072 | Improved |
| use grouped causal attention (ablation) | MoRe · Bonn video depth | Abs Rel: 0.070 -> 0.068 | No clear effect |
| add attention forcing to VGGT-style backbone (full attention) | VGGT · Sintel camera pose | ATE: 0.1715 -> 0.0877 | Improved |
| add attention forcing to VGGT-style backbone (full attention) | VGGT · ScanNet (static) camera pose | ATE: 0.0347 -> 0.0375 | Hurt |
| use MoRe instead of Stream3R | Stream3R · TUM-dynamics video depth | Abs Rel: 0.151 -> 0.173 | Hurt |
| use MoRe streaming instead of Stream3R | Stream3R · Sintel camera pose | ATE: 0.2144 -> 0.1474 | 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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 3 of 5 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
- Efficiently Reconstructing Dynamic Scenes One D4RT at a Time (CVPR 2026 · similarity 0.92)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.91)
- PAGE-4D: Disentangled Pose and Geometry Estimation for 4D Perception (ICLR 2026 · similarity 0.91)
- Flux4D: Flow-based Unsupervised 4D Reconstruction (NeurIPS 2025 · similarity 0.90)
- Fused View-Time Attention and Feedforward Reconstruction for 4D Scene Generation (NeurIPS 2025 · similarity 0.89)
- 4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos (NeurIPS 2025 · similarity 0.89)
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.
