Research map / Video / Video segmentation and tracking
RVOS: End-To-End Recurrent Network for Video Object Segmentation
Check 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 both spatial and temporal recurrence instead of spatial recurrence only (ablation) | RVOS-Mask-ST · YouTube-VOS validation, one-shot | J_unseen: 37.3 -> 44.6 | Helped |
| Add both spatial and temporal recurrence instead of temporal recurrence only (ablation) | RVOS-Mask-ST · YouTube-VOS validation, one-shot | J_unseen: 39.2 -> 44.6 | Helped |
| Fine-tune using inferred previous-frame masks rather than ground-truth masks alone (ablation) | RVOS-Mask-ST+ · YouTube-VOS validation, one-shot | F_seen: 63.7 -> 67.1 | Helped |
| Fine-tune using inferred previous-frame masks rather than ground-truth masks alone (ablation) | RVOS-Mask-ST+ · YouTube-VOS validation, one-shot | J_unseen: 44.6 -> 44.5 | No clear effect |
| Add both spatial and temporal recurrence instead of spatial recurrence only (ablation) | RVOS-ST · YouTube-VOS validation, zero-shot | J_seen: 40.8 -> 44.7 | Helped |
| Add both spatial and temporal recurrence instead of temporal recurrence only (ablation) | RVOS-ST · YouTube-VOS validation, zero-shot | J_seen: 37.1 -> 44.7 | Helped |
| Fine-tune the pretrained one-shot model on DAVIS-2017 training data (ablation) | RVOS-Mask-ST+ · DAVIS-2017 test-dev, one-shot | J: 46.4 -> 48.0 | Helped |
| Fine-tune the pretrained zero-shot model on DAVIS-2017 trainval (ablation) | RVOS-ST · DAVIS-2017, zero-shot | J: 21.7 -> 23.0 | Helped |
| Use RVOS-Mask-ST+ instead of S2S without online learning | One-shot VOS · YouTube-VOS validation | J_seen: 66.7 -> 63.6 | Hurt |
| Use RVOS-Mask-ST+ instead of S2S without online learning | One-shot VOS · YouTube-VOS validation | F_unseen: 50.3 -> 51.0 | Helped |
| Use the one-shot model trained on inferred masks rather than OSMN | One-shot VOS · DAVIS-2017 test-dev | J: 37.7 -> 46.4 | Helped |
| Train a zero-shot segmenter without first-frame masks | RVOS-ST · YouTube-VOS validation, unseen categories | J_unseen: 44.6 -> 21.2 | Hurt |
Closest papers by meaning
- Long-RVOS: A Comprehensive Benchmark for Long-term Referring Video Object Segmentation (CVPR 2026 · similarity 0.91)
- Recurrent Dynamic Embedding for Video Object Segmentation (CVPR 2022 · similarity 0.90)
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.90)
- End-to-End Referring Video Object Segmentation With Multimodal Transformers (CVPR 2022 · similarity 0.89)
- LiVOS: Light Video Object Segmentation with Gated Linear Matching (CVPR 2025 · similarity 0.89)
- Semantic and Sequential Alignment for Referring Video Object Segmentation (CVPR 2025 · 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.