Research map / 3D vision / Depth and stereo
Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction
Chin-Yang Lin, Yang-Che Sun, Cheng Sun, Fu-En Yang, Min-Hung Chen, Yen-Yu Lin, Wei-Chen Chiu, Yu-Lun 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 |
|---|---|---|---|
| replace global first-frame pose regression with multi-reference relative pose querying plus PGO | CUT3R · KITTI | ATE avg: 207.2 -> 69.7 | Improved |
| add multi-reference training (3 references) versus no reference supervision (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 33.318 -> 5.632 | Improved |
| use multiple references instead of a single reference during training (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 15.764 -> 5.632 | Improved |
| add keyframe selection (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 38.258 -> 5.632 | Improved |
| add online pose-graph optimization (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 20.089 -> 5.632 | Improved |
| increase inference references from 4 to 12 (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 15.748 -> 5.632 | Improved |
| increase inference references from 12 to 24 (beyond training count) (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 5.63 -> 12.16 | Hurt |
| add loop closure (ablation) | Scal3R (CUT3R) · KITTI | ATE avg: 143.45 -> 75.01 | Improved |
| use asymmetric instead of symmetric (VPT) attention injection (ablation) | Scal3R (CUT3R) · vKITTI | ATE: 57.78 -> 5.63 | Improved |
| use asymmetric instead of symmetric attention injection (ablation) | Scal3R (CUT3R) · Sintel / ScanNet (indoor) | ATE: Sintel 0.154 vs 0.168; ScanNet 0.083 vs 0.092 | Mixed |
| add TTT3R on top of Scal3R poses | Scal3R · KITTI | ATE: 69.73 -> 62.05 | Improved |
| remove Sim(3) scale alignment at evaluation (ablation) | Scal3R (CUT3R) · KITTI | ATE: 69.73 -> 88.66 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- TTT3R — TTT3R: 3D Reconstruction as Test-Time Training (ICLR 2026) · better on 4 of 4 shared results
- MUSt3R — MUSt3R: Multi-view Network for Stereo 3D Reconstruction (CVPR 2025) · better on 1 of 2 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
- OnlineSplatter: Pose-Free Online 3D Reconstruction for Free-Moving Objects (NeurIPS 2025 · similarity 0.87)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.86)
- Pow3R: Empowering Unconstrained 3D Reconstruction with Camera and Scene Priors (CVPR 2025 · similarity 0.86)
- Rethinking the State Update Gate for Long-Sequence Recurrent 3D Reconstruction (NeurIPS 2026 · similarity 0.86)
- No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images (ICLR 2025 · similarity 0.86)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 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.
