Research map / Detection and segmentation / Tracking
A Framework for Evaluating 6-DOF Object Trackers
Mathieu Garon, Denis Laurendeau, Jean-Francois Lalonde
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 |
|---|---|---|---|
| increase training translation range Delta t from 10 to 50 mm (ablation) | Ours specific tracker · 6-DOF tracking dataset, stability (near) | translation jitter (mm/frame): 0.42 -> 0.95 | Hurt |
| increase training translation range Delta t from 10 to 40 mm (ablation) | Ours specific tracker · 6-DOF tracking dataset, interaction speed (30, 40] mm | translation error (mm): 18.8 -> 6.6 | Improved |
| increase training rotation range Delta r from 15 to 35 deg (ablation) | Ours specific tracker · 6-DOF tracking dataset, occlusion 60% | rotation error (degrees): 15.7 -> 36.7 | Hurt |
| increase training rotation range Delta r from 15 to 35 deg (ablation) | Ours specific tracker · 6-DOF tracking dataset, occlusion 75% | rotation error (degrees): 26.0 -> 66.1 | Hurt |
| train on the object to track plus other objects (multi-object) instead of the object only | Ours tracker · 6-DOF tracking dataset, stability (near) | translation jitter (mm/frame): 0.56 -> 0.38 | Improved |
| train on the object to track plus other objects (multi-object) instead of the object only | Ours tracker · 6-DOF tracking dataset, stability (occluded) | rotation jitter (degree/frame): 0.76 -> 1.09 | Hurt |
| train on multiple objects, none of them the test object (generic) instead of the specific object | Ours tracker · 6-DOF tracking dataset, occlusion 45% | rotation error (degrees): 10.0 -> 26.8 | Hurt |
| use the generic tracker (unseen objects) instead of Garon and Lalonde trained per object | Ours generic vs Garon and Lalonde · 6-DOF tracking dataset, occlusion 45% | translation error (mm): 26.8 -> 25.9 | No clear effect |
| use the improved object-specific tracker instead of Garon and Lalonde | Garon and Lalonde · 6-DOF tracking dataset, stability (near) | translation jitter (mm/frame): 0.93 -> 0.56 | Improved |
| use the improved object-specific tracker instead of Tan et al | Tan et al · 6-DOF tracking dataset, occlusion 45% | rotation error (degrees): 70.3 -> 10.0 | Improved |
| use the improved object-specific tracker instead of Garon and Lalonde | Garon and Lalonde · 6-DOF tracking dataset, interaction free-hard | tracking failures (count): 53 -> 37 | Improved |
| remove Vicon marker artifacts from depth with the post-processing correction | Kinect depth capture · 6-DOF tracking dataset (marker patches) | RMSE (mm): 139.8 -> 4.7 | Improved |
Closest papers by meaning
- Need for Speed: A Benchmark for Higher Frame Rate Object Tracking (ICCV 2017 · similarity 0.87)
- Omni6DPose: A Benchmark and Model for Universal 6D Object Pose Estimation and Tracking (ECCV 2024 · similarity 0.87)
- GOT-10k: A Large High-Diversity Benchmark for Generic Object Tracking in the Wild (TPAMI 2019 · similarity 0.87)
- 6D Object Pose Tracking in Internet Videos for Robotic Manipulation (ICLR 2025 · similarity 0.87)
- ProxyPose: 6-DoF Pose Tracking via Video-to-Video Translation (NeurIPS 2026 · similarity 0.86)
- KPA-Tracker: Towards Robust and Real-Time Category-Level Articulated Object 6D Pose Tracking (AAAI 2024 · 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.
