Research map / Detection and segmentation / Open-set and few-shot detection
Is Tracking Really More Challenging in First Person Egocentric Vision?
Matteo Dunnhofer, Zaira Manigrasso, Christian Micheloni
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 |
|---|---|---|---|
| train STARK from scratch on FPV videos instead of generic data (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | FPV-AUC: 35.5 -> 49.8 | Improved |
| train STARK from scratch on TPV videos instead of generic data (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | TPV-AUC: 42.8 -> 48.0 | Improved |
| train STARK on TPV videos instead of generic data, measured in FPV (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | FPV-AUC: 35.5 -> 36.8 | No clear effect |
| train STARK on both FPV and TPV instead of one view (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | Delta AUC: -7.3 -> -2.1 | Improved |
| fine-tune generically pretrained STARK on FPV instead of training from scratch on FPV (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | FPV-AUC: 49.8 -> 51.9 | Improved |
| fine-tune generically pretrained XMem on FPV instead of training from scratch on FPV (ablation) | XMem · Ego-Exo4D FPV-TPV test (long-term) | FPV-AUC: 56.7 -> 56.3 | No clear effect |
| fine-tune generic STARK on both views (ablation) | STARK · Ego-Exo4D FPV-TPV test (long-term) | Delta AUC: -7.3 -> -0.5 | Improved |
| train XMem on both FPV and TPV (ablation) | XMem · Ego-Exo4D FPV-TPV test (long-term) | Delta AUC: -8.8 -> -0.2 | Improved |
| weight sequence scores by annotation length (ablation) | DAM4SAM · Ego-Exo4D FPV-TPV test (long-term) | Delta AUC: 0.9 -> -4.1 | Mixed |
| evaluate only short visible-target subsequences instead of full long-term tracking | STARK · Ego-Exo4D FPV-TPV test (short-term) | FPV-AUC: 35.5 -> 56.7 | Improved |
| evaluate short-term visible-target segments instead of long-term | XMem · Ego-Exo4D FPV-TPV test (short-term) | Delta AUC: -8.8 -> -3.0 | Improved |
| use SAM2-based foundation trackers instead of small trackers | SAM2/SAMURAI/DAM4SAM · Ego-Exo4D FPV-TPV test (short-term) | viewpoint bias | Improved |
Closest papers by meaning
- EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset (NeurIPS 2023 · similarity 0.89)
- Visual Object Tracking in First Person Vision (IJCV 2022 · similarity 0.89)
- Instance Tracking in 3D Scenes from Egocentric Videos (CVPR 2024 · similarity 0.88)
- Where and What: Long-Term Object Tracking in Egocentric Videos (ECCV 2026 · similarity 0.87)
- Robust Egocentric Referring Video Object Segmentation via Dual-Modal Causal Intervention (NeurIPS 2025 · similarity 0.86)
- EgoGen: An Egocentric Synthetic Data Generator (CVPR 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.
