Research map / 3D vision / Depth and stereo
Multi-View 3D Point Tracking
Frano Rajič, Haofei Xu, Marko Mihajlovic, Siyuan Li, Irem Demir, Emircan Gündoğdu, Lei Ke, Sergey Prokudin, Marc Pollefeys, Siyu Tang
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 kNN offset vector to point correlation (ablation) | MVTracker · Multi-View Kubric | AJ: 21.3 -> 53.6 | Improved |
| add neighbor world coordinates alongside the offset (ablation) | MVTracker · Multi-View Kubric | AJ: 53.6 -> 48.7 | Hurt |
| replace world-aligned triplane correlation with kNN correlation | MVTracker · DexYCB (DUSt3R depth) | AJ: 57.5 -> 71.6 | Improved |
| add more input views from 1 to 8 | MVTracker · DexYCB (DUSt3R depth) | AJ: 64.0 -> 79.2 | Improved |
| add depth-source augmentation (GT plus estimated) to variable-view training (ablation) | MVTracker · DexYCB | AJ: 49.9 -> 65.2 | Improved |
| add variable-view augmentation to depth-source training (ablation) | MVTracker · DexYCB | AJ: 62.3 -> 65.2 | Improved |
| add depth-source augmentation (ablation) | MVTracker · Panoptic Studio | AJ: 80.4 -> 80.7 | No clear effect |
| add variable-view augmentation (ablation) | MVTracker · Multi-View Kubric | AJ: 77.2 -> 80.4 | Improved |
| use sensor depth instead of DUSt3R depth | MVTracker · DexYCB | AJ: 71.6 -> 82.7 | Improved |
| use multi-view fusion instead of single-view tracker | MVTracker vs SpaTracker · Panoptic Studio | AJ: 61.5 -> 86.0 | Improved |
| use VGGT instead of ground-truth depth | MVTracker · Panoptic Studio | AJ: 86.0 -> 47.7 | Hurt |
| add more views for single-view baselines on Kubric | TAPIP3D · Multi-View Kubric | AJ: 76.4 -> 70.6 | Hurt |
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.
- TAPIP3D — TAPIP3D: Tracking Any Point in Persistent 3D Geometry (NeurIPS 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
- TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels (NeurIPS 2025 · similarity 0.90)
- TAPVid-3D: A Benchmark for Tracking Any Point in 3D (NeurIPS 2024 · similarity 0.89)
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.89)
- TAPIP3D: Tracking Any Point in Persistent 3D Geometry (NeurIPS 2025 · similarity 0.89)
- Tracking Everything Everywhere across Multiple Cameras (AAAI 2025 · similarity 0.89)
- TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking (NeurIPS 2026 · similarity 0.89)
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.
