Research map / 3D vision / Depth and stereo
Detect by Track: Making Detector-Free Matcher Trackable
Yusuke Sekikawa, HIDEKI SHIRAI, Ruka Eto, Yuzhe Hao, Kengo Mitsui, Nakamasa Inoue
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 DeT (detect by tracking) instead of NN association (ablation) | JamMA · IMC Reichstag 5-frame pose | pose AUC@5: 35.0 -> 84.0 | Improved |
| add DeT instead of NN association (ablation) | JamMA · MegaDepth 0015 5-frame pose | pose AUC@5: 24.4 -> 72.4 | Improved |
| add DeT instead of NN association (ablation) | EDM · IMC Reichstag 5-frame pose | pose AUC@5: 19.0 -> 29.0 | Improved |
| add DeT instead of NN association (ablation) | EDM · MegaDepth 0022 5-frame pose | pose AUC@5: 15.9 -> 19.5 | Improved |
| add DeT instead of NN association (ablation) | JamMA · IMC Reichstag incremental SLAM | pose AUC@5: 59.8 -> 68.7 | Improved |
| add DeT instead of NN association (ablation) | EDM · IMC incremental SLAM (Reichstag, Sacre Coeur) | pose AUC@5: 55.1 -> 55.4; 79.7 -> 79.5 | No clear effect |
| use DeT-JamMA instead of SP+LightGlue | SP+LightGlue · IMC tracking (IMC sequence) | correct tracks: more than five times as many correct tracks | Improved |
| use DeT-JamMA instead of SP+LightGlue | SP+LightGlue · IMC Reichstag 5-frame pose | pose AUC@5: 75.0 -> 84.0 | Improved |
| use DeT-EDM instead of SP+LightGlue | SP+LightGlue · IMC Reichstag 5-frame pose | pose AUC@5: 75.0 -> 29.0 | Hurt |
| use DeT-JamMA in incremental SLAM instead of NN-JamMA (ablation) | NN-JamMA · IMC incremental SLAM | registered 3D points: more than four times as many 3D points | Improved |
| add DeT to JamMA in two-view setting (ablation) | JamMA · MegaDepth two-view pose | pose AUC@5: 64.1 -> 61.4 | Hurt |
| add DeT to EDM in two-view setting (ablation) | EDM · MegaDepth two-view pose | pose AUC@5: 64.5 -> 63.6 | Hurt |
Closest papers by meaning
- Towards Complementary Keypoint Detection via Mixture of Detectors (NeurIPS 2026 · similarity 0.89)
- Detector-Free Structure from Motion (CVPR 2024 · similarity 0.88)
- From Pairs to Sequences: Track-Aware Policy Gradients for Keypoint Detection (CVPR 2026 · similarity 0.87)
- TrackFlow: Multi-Object tracking with Normalizing Flows (ICCV 2023 · similarity 0.87)
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.87)
- CoMatcher: Multi-View Collaborative Feature Matching (CVPR 2025 · 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.
