Research map / Video / Video segmentation and tracking
Low-Shot Video Object Segmentation
Kun Yan, Fangyun Wei, Shuyu Dai, Minghui Wu, Ping Wang, Chang Xu
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 phase-1 pseudo-label training (ablation) | 2-shot STCN · YouTube-VOS 2019 val | G: 80.6 -> 81.6 | Improved |
| add phase-2 training (ablation) | 2-shot STCN + phase-1 · YouTube-VOS 2019 val | G: 81.6 -> 82.7 | Improved |
| use mean-teacher pseudo-labeler instead of plain STCN (ablation) | phase-1 · YouTube-VOS 2019 val | G: 81.2 -> 81.6 | Improved |
| set EMA alpha 0.995 versus 0.990 or 0.999 (ablation) | MT-STCN · YouTube-VOS 2019 val | G: 81.2 -> 81.6 | Improved |
| use bidirectional instead of unidirectional intermediate inference (ablation) | phase-2 pseudo-labels · YouTube-VOS 2019 val | G: 82.1 -> 82.7 | Improved |
| update pseudo-label bank dynamically in phase-2 (ablation) | phase-2 · YouTube-VOS 2019 val | G: 82.2 -> 82.7 | Improved |
| use mean teacher in phase-2 (ablation) | phase-2 · YouTube-VOS 2019 val | G | No clear effect |
| use more labeled shots (4 or 6 instead of 2) (ablation) | STCN w/ Ours · YouTube-VOS 2019 val | G: 82.7 -> 82.7 | No clear effect |
| apply our approach to 2-shot XMem | XMem · YouTube-VOS 2018 val | G: 79.2 -> 84.8 | Improved |
| apply our approach to 2-shot RDE-VOS | RDE-VOS · YouTube-VOS 2019 val | G: 78.4 -> 82.1 | Improved |
| apply our approach to 2-shot STCN | STCN · DAVIS 2017 val | J&F: 81.0 -> 85.1 | Improved |
| two labeled frames instead of one (full data halved) (ablation) | STCN · YouTube-VOS 2019 val | G: 72.9 -> 79.8 | Improved |
Closest papers by meaning
- Two-Shot Video Object Segmentation (CVPR 2023 · similarity 0.99)
- Learning What to Learn for Video Object Segmentation (ECCV 2020 · similarity 0.93)
- Point-VOS: Pointing Up Video Object Segmentation (CVPR 2024 · similarity 0.90)
- Learning Fast and Robust Target Models for Video Object Segmentation (CVPR 2020 · similarity 0.90)
- Learning Video Object Segmentation From Unlabeled Videos (CVPR 2020 · similarity 0.90)
- Make One-Shot Video Object Segmentation Efficient Again (NeurIPS 2020 · similarity 0.90)
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.
