Research map / 3D vision / Depth and stereo
ParticleSfM: Exploiting Dense Point Trajectories for Localizing Moving Cameras in the Wild
Wang Zhao, Shaohui Liu, Hengkai Guo, Wenping Wang, Yong-Jin Liu
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 trajectory-based motion segmentation to dense trajectories with global BA (ablation) | Traj + Global BA · Sintel, successful subset | ATE: 0.071 -> 0.046 | Improved |
| add trajectory-based motion segmentation (ablation) | Traj + Global BA · Sintel, successful subset | RPE rot: 0.969 -> 0.212 | Improved |
| add trajectory optimization with path consistency (ablation) | Traj + Seg + Global BA · Sintel, full set | ATE: 0.146 -> 0.129 | Improved |
| add trajectory optimization without segmentation (ablation) | Traj + Global BA · Sintel, successful subset | ATE: 0.071 -> 0.072 | No clear effect |
| add MAT motion masks to SIFT global BA (ablation) | SIFT + Global BA · Sintel, successful subset | ATE: 0.060 -> 0.054 | Mixed |
| use dense trajectories with segmentation and optimization instead of SIFT features (ablation) | SIFT + Global BA · Sintel, successful subset | ATE: 0.060 -> 0.042 | Improved |
| add relative depth (MiDaS) input to segmentation network (ablation) | Trajectory segmentation network · Sintel motion segmentation | mIoU: 54.6 -> 60.6 | Improved |
| use transformer encoder instead of MLP encoder (ablation) | Trajectory segmentation network · Sintel motion segmentation | mIoU: 54.8 -> 60.6 | Improved |
| use OANet decoder instead of PointNet decoder (ablation) | Trajectory segmentation network · Sintel motion segmentation | mIoU: 46.3 -> 60.6 | Improved |
| ParticleSfM vs COLMAP on static ScanNet subset | ParticleSfM · ScanNet COLMAP subset | ATE: 0.171 vs 0.319 | Hurt |
| ParticleSfM vs COLMAP on static ScanNet subset | ParticleSfM · ScanNet COLMAP subset | RPE rot: 2.900 -> 0.632 | Improved |
| ParticleSfM vs DROID-SLAM on Sintel | ParticleSfM · Sintel full set | ATE: 0.175 -> 0.129 | 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.
- DROID-SLAM — DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (NeurIPS 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
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025 · similarity 0.88)
- Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos (ECCV 2026 · similarity 0.88)
- Degrees of Freedom Matter: Inferring Dynamics from Point Trajectories (CVPR 2024 · similarity 0.86)
- Dense Optical Tracking: Connecting the Dots (CVPR 2024 · similarity 0.86)
- SLAM3R: Real-Time Dense Scene Reconstruction from Monocular RGB Videos (CVPR 2025 · similarity 0.86)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.86)
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.
