Research map / Video / Video segmentation and tracking
Two-Shot Video Object Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 semi-supervised training with pseudo-labels (first frame labeled) (ablation) | 2-shot STCN · YouTube-VOS 2019 | G: +1.0 | Helped |
| add phase-2 training with unrestricted reference frame and pseudo-label bank (ablation) | 2-shot STCN (after phase-1) · YouTube-VOS 2019 | G: +1.1 | Helped |
| apply two-shot training paradigm (phases 1+2) to STCN | STCN · DAVIS 2017 | J&F: +4.1 | Helped |
| apply two-shot paradigm to RDE-VOS | RDE-VOS · YouTube-VOS 2019 | G: +3.7 | Helped |
| apply two-shot paradigm to XMem | XMem · YouTube-VOS 2019 | G: +5.4 | Helped |
| apply two-shot paradigm to XMem | XMem · DAVIS 2017 | J&F: +3.9 vs naive 2-shot; -0.6 vs full set | Mixed |
| use MT-STCN (mean teacher EMA, alpha 0.995) as phase-1 pseudo-labeler instead of STCN (ablation) | phase-1 pseudo-labeling · YouTube-VOS 2019 | G: +0.4 (81.2 to 81.6) | Helped |
| bidirectional instead of unidirectional intermediate inference for pseudo-labels (ablation) | phase-2 pseudo-label bank · YouTube-VOS 2019 | G: +0.6 | Helped |
| dynamically update pseudo-label bank during phase-2 instead of freezing it (ablation) | phase-2 training · YouTube-VOS 2019 | G: +0.5 | Helped |
| increase number of labeled shots from 2 to 4 or 6 with the two-shot pipeline (ablation) | STCN · YouTube-VOS 2019 | G: 4-shot and 6-shot both 82.7 after phase-2, same as 2-shot | No clear effect |
Closest papers by meaning
- 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)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.88)
- RMem: Restricted Memory Banks Improve Video Object Segmentation (CVPR 2024 · similarity 0.87)
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.