Research map / 3D vision / Gaussian splatting
UniFusion: Sparse-View 4D Reconstruction via Unified Spatio-temporal Depth Alignment
Yongzhe Lyu, Shaofei Wang, Yixin Chen, Siyuan Huang
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 temporal residual weights to per-frame alignment (vs naive time embedding) (ablation) | MonoFusion · Ego-Exo4D | PSNR: 28.53 -> 30.90 | Improved |
| add temporal residual weights to naive time embedding variant (ablation) | Def-GS · Ego-Exo4D | PSNR: 32.60 -> 33.61 | Improved |
| add naive time embedding to per-frame MAtCha (ablation) | MonoFusion · Ego-Exo4D | PSNR: 30.14 -> 28.53 | Hurt |
| add naive time embedding to per-frame MAtCha (ablation) | Def-GS · Ego-Exo4D | PSNR: 32.60 -> 32.60 | No clear effect |
| add depth-order loss (ablation) | MonoFusion · Ego-Exo4D | PSNR: 30.90 -> 31.52 | Improved |
| add depth-order loss (ablation) | Def-GS · Ego-Exo4D | PSNR: 33.61 -> 33.90 | Improved |
| add scale-and-shift-invariant loss (ablation) | Def-GS · Ego-Exo4D | PSNR: 33.90 -> 34.66 | Improved |
| add SSI loss (ablation) | MonoFusion · Ego-Exo4D | PSNR: 31.52 -> 31.63 | Improved |
| add temporal residual weights (ablation) | Def-GS · EgoHuman | PSNR: 20.93 -> 21.80 | Mixed |
| add SSI loss (ablation) | Def-GS · EgoHuman | PSNR: 22.15 -> 22.76 | Improved |
| increase residual rank R from 4 to 8 (ablation) | Def-2DGS + Ours · Ego-Exo4D | PSNR: 33.76 -> 34.66 | Improved |
| increase residual rank R from 8 to 16 (ablation) | Def-2DGS + Ours · Ego-Exo4D | PSNR: 34.66 -> 34.51 | No clear effect |
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.
- MonoFusion — MonoFusion: Sparse-View 4D Reconstruction via Monocular Fusion (ICCV 2025) · better on 1 of 1 shared result
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
- Seurat: From Moving Points to Depth (CVPR 2025 · similarity 0.88)
- Align3R: Aligned Monocular Depth Estimation for Dynamic Videos (CVPR 2025 · similarity 0.88)
- Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · similarity 0.87)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.87)
- Fused View-Time Attention and Feedforward Reconstruction for 4D Scene Generation (NeurIPS 2025 · similarity 0.87)
- MonoInstance: Enhancing Monocular Priors via Multi-view Instance Alignment for Neural Rendering and Reconstruction (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.
