Research map / Point clouds / Registration and matching
DRS-VPT: Directly Re-localizing in Scenes using a Vision and Point Transformer
Lanke Fu, Maurice Fallon
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 raw pose head with feature-metric direct alignment refinement (ablation) | DRS-VPT · KITTI | mean translation error (m): 0.50 -> 0.09 | Improved |
| add ICP refinement on point maps after pose head (ablation) | DRS-VPT · KITTI | mean translation error (m): 0.50 -> 0.52 | No clear effect |
| add Kabsch alignment on point maps after pose head (ablation) | DRS-VPT · KITTI | mean translation error (m): 0.50 -> 0.61 | Hurt |
| add ICP refinement on point maps (ablation) | DRS-VPT · KITTI | mean rotation error (deg): 1.45 -> 1.25 | Improved |
| increase input point count from 5,000 to 50,000 (ablation) | DRS-VPT · KITTI | mean translation error (m): 0.14 -> 0.08 | Improved |
| train one model on diverse multi-dataset corpus vs KITTI-only training | DRS-VPT · KITTI to nuScenes generalization | recall (%): 2.9 -> 66.74 | Improved |
| ingest 3D map at inference instead of scene-specific training | DRS-VPT · 7Scenes | overall recall (%): 85.2 (DSAC*) vs 66.4 | Mixed |
| use DRS-VPT zero-shot on unseen 12Scenes | DRS-VPT · 12Scenes | overall recall (%): 53.6 | Mixed |
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.
- TrafficLoc — TrafficLoc: Localizing Traffic Surveillance Cameras in 3D Scenes (ICCV 2025) · better on 4 of 4 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
- LINK: Learning to Localize from Known to Unknown Scenes (NeurIPS 2026 · similarity 0.86)
- Local Supports Global: Deep Camera Relocalization With Sequence Enhancement (ICCV 2019 · similarity 0.86)
- RegFormer: An Efficient Projection-Aware Transformer Network for Large-Scale Point Cloud Registration (ICCV 2023 · similarity 0.86)
- A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration (NeurIPS 2024 · similarity 0.85)
- On the Limits of Pseudo Ground Truth in Visual Camera Re-Localisation (ICCV 2021 · similarity 0.85)
- Living Scenes: Multi-object Relocalization and Reconstruction in Changing 3D Environments (CVPR 2024 · 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.
