Research map / 3D vision / Human pose estimation
Large-Displacement 3D Object Tracking with Hybrid Non-local Optimization
Xuhui Tian, Xinran Lin, Fan Zhong, Xueying Qin
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 hybrid non-local search over out-of-plane rotation to the local method (ablation) | Ours · RBOT regular (S=4) | accuracy (%): 30.7 -> 81.7 | Improved |
| add grid pretermination to naive non-local search (ablation) | Ours non-local · RBOT regular (S=4) | time and iterations: 1085 -> 718 iterations; 47.7ms -> 34.0ms | Improved |
| add path pretermination to naive non-local search (ablation) | Ours non-local · RBOT regular (S=4) | time and iterations: 1085 -> 823 iterations; 47.7ms -> 37.9ms | Improved |
| add grid and path pretermination plus near-to-far search (ablation) | Ours non-local · RBOT regular (S=4) | time: 47.7ms -> 22.3ms; 85.1 -> 81.7 accuracy | Improved |
| increase alpha from 0.125 to 1.5 (ablation) | Ours (local) · RBOT regular (S=1) | accuracy (%): 94.2 -> 72.3 | Hurt |
| increase number of search directions D from 4 to 16 (ablation) | Ours (local) · RBOT regular (S=1) | accuracy (%): 88.1 -> 94.2 | Improved |
| increase D from 16 to 20 (ablation) | Ours (local) · RBOT regular (S=1) | accuracy (%): 94.2 -> 94.3 | No clear effect |
| increase candidate correspondences M from 1 to 3 (ablation) | Ours · RBOT regular (S=1) | accuracy (%): 93.6 -> 95.2 | Improved |
| increase M from 3 to 15 (ablation) | Ours · RBOT regular (S=1) | accuracy (%): 95.2 -> 95.6 | No clear effect |
| increase sampled contour points N from 50 to 200 (ablation) | Ours · RBOT regular (S=1) | accuracy (%): 93.9 -> 95.2 | Improved |
| increase N from 200 to 400 (ablation) | Ours · RBOT regular (S=1) | accuracy (%): 95.2 -> 95.2 | No clear effect |
| use new search-line model instead of prior methods (local variant, no non-local search) | Ours (local) vs SRT3D · RBOT regular (S=4) | accuracy (%): 19.4 -> 30.7 | 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.
- SRT3D — SRT3D: A Sparse Region-Based 3D Object Tracking Approach for the Real World (IJCV 2022) · 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
- SRT3D: A Sparse Region-Based 3D Object Tracking Approach for the Real World (IJCV 2022 · similarity 0.89)
- Joint Spatial-Temporal Optimization for Stereo 3D Object Tracking (CVPR 2020 · similarity 0.88)
- Prior-free 3D Object Tracking (CVPR 2025 · similarity 0.87)
- SpatialTracker: Tracking Any 2D Pixels in 3D Space (CVPR 2024 · similarity 0.85)
- TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels (NeurIPS 2025 · similarity 0.85)
- 4DTAM: Non-Rigid Tracking and Mapping via Dynamic Surface Gaussians (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.
