Research map / 3D vision / Depth and stereo
LASER: Layer-wise Scale Alignment for Training-Free Streaming 4D Reconstruction
Tianye Ding, Yiming Xie, Yiqing Liang, Moitreya Chatterjee, Pedro Miraldo, Huaizu Jiang
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 layer-wise scale alignment (LSA) (ablation) | LASER (pi3) · Video depth (Sintel) | Abs Rel: 0.328 -> 0.247 | Improved |
| add layer-wise scale alignment (LSA) (ablation) | LASER (pi3) · Video depth (Sintel) | delta<1.25: 51.4 -> 68.8 | Improved |
| add layer-wise scale alignment (LSA) (ablation) | LASER (pi3) · Video depth (Bonn) | Abs Rel: 0.123 -> 0.048 | Improved |
| add layer-wise scale alignment (LSA) (ablation) | LASER (pi3) · Video depth (Bonn) | delta<1.25: 85.6 -> 97.4 | Improved |
| replace the efficient segmentation algorithm with SAM 2 (ablation) | LASER (pi3) · Video depth (Sintel) | Abs Rel: 0.247 -> 0.251 | No clear effect |
| add temporal propagation along intra-window edges (ablation) | LASER (pi3) · Video depth (Sintel) | Abs Rel: 0.261 -> 0.247 | Improved |
| add temporal propagation along intra-window edges (ablation) | LASER (pi3) · Video depth (Sintel) | delta<1.25: 64.7 -> 68.8 | Improved |
| raise IoU threshold tau from 0.2 to 0.3 (ablation) | LASER (pi3) · Video depth (Bonn) | delta<1.25: 94.7 -> 97.4 | Improved |
| vary IoU threshold tau from 0.3 to 0.6 (ablation) | LASER (pi3) · Video depth (Sintel) | Abs Rel: 0.247 -> 0.248 | No clear effect |
| use IRLS instead of closed-form scale estimation (ablation) | LASER (pi3) · Video depth (Sintel) | Abs Rel: 0.328 -> 0.247 | Improved |
| use IRLS instead of closed-form scale estimation (ablation) | LASER (pi3) · Camera pose (Sintel) | ATE: 0.107 -> 0.061 | Improved |
| use scaled camera anchors instead of scaled point maps for rigid transform (ablation) | LASER (pi3) · Camera pose (Sintel) | RPE_rot: 0.742 -> 0.249 | 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.
- TTT3R — TTT3R: 3D Reconstruction as Test-Time Training (ICLR 2026) · better on 2 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
- LongStream: Long-Sequence Streaming Autoregressive Visual Geometry (CVPR 2026 · similarity 0.88)
- CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction (NeurIPS 2026 · similarity 0.87)
- GHOST: Geometry-Hierarchical Online Streaming Token Eviction for Efficient 3D Reconstruction (NeurIPS 2026 · similarity 0.86)
- IGGT4D: Streaming 4D Instance-Grounded Geometry Transformer (NeurIPS 2026 · similarity 0.86)
- FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry (NeurIPS 2026 · similarity 0.86)
- World Tracing: Pixel-Aligned Geometry Beyond the Visible (NeurIPS 2026 · 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.
