Research map / 3D vision / 3D generation
Trust It or Not: Evidential Uncertainty for Feed-Forward 3D Reconstruction with Trust3R
Zihao Zhu, Wenyuan Zhao, Nuo Chen, Chao Tian, Zhiwen Fan
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 MASt3R confidence with evidential NIW uncertainty head | MASt3R · ScanNet++ | AURC: 0.1649 -> 0.1233 | Improved |
| replace MASt3R confidence with evidential NIW uncertainty head | MASt3R · ScanNet++ | Spearman rho: 0.2837 -> 0.4930 | Improved |
| replace MASt3R confidence with evidential NIW uncertainty head | MASt3R · KITTI | AURC: 1.0277 -> 0.9868 | Improved |
| evidential head versus single-pass heteroscedastic regression | MASt3R · KITTI | AURC: 0.9749 -> 0.9868 | Hurt |
| add gated residual mean refinement (ablation) | Trust3R · ScanNet++ | AURC: 0.1788 -> 0.1349 | Improved |
| add gated residual mean refinement (ablation) | Trust3R · KITTI | MAE: 1.6108 -> 1.6436 | Hurt |
| add gated residual mean refinement (ablation) | Trust3R · KITTI | AURC: 1.1170 -> 0.9942 | Improved |
| use full-covariance NIW head instead of factorized NIG (ablation) | Trust3R · ETH3D | AURC: 0.3213 -> 0.3040 | Improved |
| use full-covariance NIW head instead of factorized NIG (ablation) | Trust3R · ScanNet++ | AURC: 0.1237 -> 0.1233 | No clear effect |
| use epistemic readout instead of aleatoric (ablation) | Trust3R · ETH3D | AURC: 0.3175 -> 0.3040 | Improved |
| use total uncertainty instead of aleatoric (ablation) | Trust3R · ETH3D | AURC: 0.3175 -> 0.3064 | Improved |
| attach evidential head to VGGT backbone | VGGT · uncertainty ranking | AUSE: 0.0452 -> 0.0209 | Improved |
Closest papers by meaning
- Auto3R: Automated 3D Reconstruction and Scanning via Data-driven Uncertainty Quantification (ECCV 2026 · similarity 0.87)
- Dens3R: A Foundation Model for 3D Geometry Prediction (ICLR 2026 · similarity 0.87)
- Creat3r: Confidence Reaggregation for Exploration-aware Active 3D Reconstruction (ICML 2026 · similarity 0.86)
- Variational Multi-scale Representation for Estimating Uncertainty in 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.86)
- Test3R: Learning to Reconstruct 3D at Test Time (NeurIPS 2025 · similarity 0.85)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.85)
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.
