Research map / Video / Video segmentation and tracking
Segmentation-Guided Homography Estimation for Long-Term Planar Tracking
Jonáš Šerých, Jiri Matas
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 SAM-H mask-based homography with SAM 2 masks | planar tracker on PlanarTrack TST · PlanarTrack TST (original GT) | p@5: 43.6 (WOFT) -> 62.0 | Improved |
| add SAM-H re-detection cascade to WOFT (WOFTSAM) | WOFT · PlanarTrack TST (original GT) | p@15: 64.8 -> 77.2 | Improved |
| add SAM-H re-detection cascade to WOFT (WOFTSAM) | WOFT · POT-210 | p@15: 95.4 -> 97.5 | Improved |
| use SAM-H alone instead of WOFT on precise-alignment benchmark | SAM-H · POT-210 | p@5: 90.0 (WOFT) -> 64.4 | Hurt |
| add DINOv2 appearance-based symmetry disambiguation (ablation) | SAM-H · PlanarTrack TST (re-annotated) | p@15: 68.0 -> 80.5 | Improved |
| require multi-frame consistency (Theta_r=5) for symmetry disambiguation (ablation) | SAM-H · PlanarTrack TST | p@15: 76.6 -> 80.0 | Improved |
| increase Theta_r from 5 to 8 (ablation) | SAM-H · PlanarTrack TST | p@15: 80.0 -> 80.1 | No clear effect |
| use SAM2+ (stride 5, no memory update when absent) instead of plain SAM 2 (ablation) | SAM-H · PlanarTrack TST | p@15: 77.2 -> 80.0 | Improved |
| use full DAM4SAM instead of SAM2+ (ablation) | SAM-H · PlanarTrack TST | p@15: 80.0 -> 77.7 | Hurt |
| replace Hough corner extraction with min-area rectangle or RDP quad (ablation) | SAM-H · PlanarTrack TST (re-annotated) | p@15: 80.5 -> 39.2 (min-area), 70.1 (RDP) | Hurt |
| vary WOFTSAM inlier threshold from 20% to 10% or 30% (ablation) | WOFTSAM · PlanarTrack TST / POT-210 | p@15: PT 75.0 / 77.2 / 77.6 for 10/20/30% | No clear effect |
| re-annotate PlanarTrack initial frames precisely | WOFTSAM benchmark evaluation · PlanarTrack TST | p@5: WOFTSAM 51.5 -> 57.4; SAM-H 62.0 -> 62.0 | Improved |
Closest papers by meaning
- PlanarTrack: A Large-scale Challenging Benchmark for Planar Object Tracking (ICCV 2023 · similarity 0.89)
- "HVC-Net: Unifying Homography, Visibility, and Confidence Learning for Planar Object Tracking" (ECCV 2022 · similarity 0.88)
- Adapting Dense Matching for Homography Estimation with Grid-based Acceleration (CVPR 2025 · similarity 0.86)
- Noisy One-point Homographies are Surprisingly Good (CVPR 2024 · similarity 0.85)
- KPA-Tracker: Towards Robust and Real-Time Category-Level Articulated Object 6D Pose Tracking (AAAI 2024 · similarity 0.85)
- ProxyPose: 6-DoF Pose Tracking via Video-to-Video Translation (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.
