Research map / 3D vision / Human pose estimation
Object Reprojection Error (ORE): Camera pose benchmarks from lightweight tracking annotations
Xingyu Chen, Weiyao Wang, Hao Tang, Matt Feiszli
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 exhaustive matching with vocabulary tree matching (ablation) | COLMAP · EgoStatic | ORE: 0.066 -> 0.041 | Improved |
| replace exhaustive matching with vocabulary tree matching (ablation) | COLMAP · EgoStatic | runtime: about 1.5x faster | Improved |
| run on raw fisheye videos without undistortion (ablation) | COLMAP-Tuned · EgoStatic | ORE: 0.041 -> 0.044 | No clear effect |
| filter dynamic components with hand-object masks (ablation) | COLMAP · EgoStatic | ORE: 0.066 -> 0.049 | Improved |
| remove global backend bundle adjustment (ablation) | DROID-SLAM · EgoStatic | ORE: 0.185 -> 0.214 | Hurt |
| remove global backend bundle adjustment (ablation) | DROID-SLAM · ScanNet | ORE: 0.029 -> 0.053 | Hurt |
| remove global backend bundle adjustment (ablation) | DROID-SLAM · ScanNet | ATEtrans: 0.147 -> 0.218 | Hurt |
| remove global backend bundle adjustment (ablation) | DROID-SLAM · ScanNet | ATErot: 0.135 -> 0.232 | Hurt |
| shrink object bounding box for ORE from 1000% to 20% (ablation) | ORE metric · ScanNet | Spearman correlation with ATErot: 0.384 -> 0.800 | Improved |
| shrink object bounding box for ORE from 1000% to 20% (ablation) | ORE metric · ScanNet | Spearman correlation with ATEtrans: 0.262 -> 0.716 | Improved |
| use point tracklets (box size 0%) instead of 20% box (ablation) | ORE metric · ScanNet | Spearman correlation with ATEtrans: 0.716 -> 0.720 | No clear effect |
| move from ScanNet to egocentric EgoStatic videos | DROID-SLAM · EgoStatic | ORE: 0.029 -> 0.185 | Hurt |
Closest papers by meaning
- Princeton365: A Diverse Dataset with Accurate Camera Pose (ICCV 2025 · similarity 0.87)
- SCRREAM : SCan, Register, REnder And Map: A Framework for Annotating Accurate and Dense 3D Indoor Scenes with a Benchmark (NeurIPS 2024 · similarity 0.85)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.85)
- InFlux: A Benchmark for Self-Calibration of Dynamic Intrinsics of Video Cameras (NeurIPS 2025 · similarity 0.85)
- 3DSPA: A 3D Semantic Point Autoencoder for Evaluating Video Realism (NeurIPS 2026 · similarity 0.85)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · 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.
