Research map / 3D vision / Depth and stereo
Neural Prior for Trajectory Estimation
Chaoyang Wang, Xueqian Li, Jhony Kaesemodel Pontes, Simon Lucey
Check 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 |
|---|---|---|---|
| Use NTP instead of the strongest listed baseline | NRSfM · drink, pickup, stretch, and dance | normalized mean 3D error ×100: drink: 0.47 -> 0.34; pickup: 2.03 -> 1.84; stretch: 1.62 -> 1.51; dance: 10.22 -> 9.78 | Improved |
| Use NTP instead of PAUL | NRSfM · yoga | normalized mean 3D error ×100: 1.71 -> 2.10 | Hurt |
| Use NTP instead of PND | NRSfM · shark | normalized mean 3D error ×100: 1.35 -> 9.42 | Hurt |
| Add NTP to PAUL | NRSfM · drink, pickup, yoga, stretch, dance, and shark | normalized mean 3D error ×100: drink: 0.47 -> 0.28; pickup: 2.03 -> 1.16; yoga: 1.71 -> 1.20; stretch: 1.62 -> 1.32; dance: 10.22 -> 8.53; shark: 0.37 | Improved |
| Use NTP rather than JM | Dense NRSfM · Synthetic face traj. A | normalized mean 3D error: 0.028 -> 0.031 | Hurt |
| Use NTP rather than EM-FEM | Dense NRSfM · Synthetic face traj. B | normalized mean 3D error: 0.030 -> 0.034 | Hurt |
| Replace NSFP Euler integration with direct NTP trajectory estimation | Lidar scene-flow integration · Argoverse full | Acc 0.5 (%): 45.28 -> 52.28 | Improved |
| Replace NSFP Euler integration with direct NTP trajectory estimation | Lidar scene-flow integration · Argoverse dynamic regions | Acc 0.5 (%): 63.56 -> 69.49 | Improved |
| Replace NSFP Euler integration with direct NTP trajectory estimation | Lidar scene-flow integration · Argoverse full | Acc 1 (%): 59.52 -> 69.88 | Improved |
| Replace NSFP Euler integration with direct NTP trajectory estimation | Lidar scene-flow integration · Argoverse full | Out. (%): 29.33 -> 20.95 | Improved |
| Replace NSFP Euler integration with direct NTP trajectory estimation | Lidar scene-flow integration · Argoverse full | cd-24 (m): 14.09 -> 9.77 | Improved |
| Add continuous depth field fz and trajectory code field fϕ (ablation) | Dense NRSfM · Synthetic face traj. A | normalized mean 3D error: 0.097 -> 0.032 | 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.
- PAUL — PAUL: Procrustean Autoencoder for Unsupervised Lifting (CVPR 2021) · better on 5 of 5 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
- Neural Scene Flow Prior (NeurIPS 2021 · similarity 0.89)
- PriorMotion: Generative Class-Agnostic Motion Prediction with Raster-Vector Motion Field Priors (ICCV 2025 · similarity 0.87)
- Fast Neural Scene Flow (ICCV 2023 · similarity 0.87)
- Normalizing Trajectory Models (NeurIPS 2026 · similarity 0.85)
- E-MoFlow: Learning Egomotion and Optical Flow from Event Data via Implicit Regularization (NeurIPS 2025 · similarity 0.84)
- Optimizing Diffusion Noise Can Serve As Universal Motion Priors (CVPR 2024 · similarity 0.84)
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.
