Research map / Video / Video segmentation and tracking
Towards Open-Vocabulary Video Instance Segmentation
Haochen Wang, Cilin Yan, Shuai Wang, Xiaolong Jiang, Xu Tang, Yao Hu, Weidi Xie, Efstratios Gavves
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 |
|---|---|---|---|
| Use end-to-end video queries (OV2Seg) instead of propose-reduce-association baselines | OV2Seg · LV-VIS val novel | AP novel: 5.4 -> 11.9 (vs Detic-XMem, R50) | Improved |
| Use class-independent queries instead of class-dependent queries (ablation) | OV2Seg · LVIS / LV-VIS(O) image-level | fps: 0.10 -> 21.6 fps at comparable AP novel (17.4 vs 17.5) | Improved |
| Add memory update (alpha=0.7) to query association (ablation) | OV2Seg · YouTube-VIS 2019 | mAP: 23.7 -> 27.2 | Improved |
| Vary memory update factor alpha between 0.3 and 0.8 (ablation) | OV2Seg · YouTube-VIS 2019, LV-VIS val | mAP | No clear effect |
| Classify with memory queries instead of averaging per-frame scores (ablation) | OV2Seg · LV-VIS val | AP novel: 11.3 -> 11.9 | Improved |
| Use OV2Seg instead of Detic at image level | OV2Seg · LVIS | AP: 30.2 -> 25.4 AP (R50) | Hurt |
| Use OV2Seg with SwinB backbone instead of R50 | OV2Seg · LV-VIS val | AP: 14.2 -> 21.1 | Improved |
Closest papers by meaning
- Unified Embedding Alignment for Open-Vocabulary Video Instance Segmentation (ECCV 2024 · similarity 0.94)
- OV-VIS: Open-Vocabulary Video Instance Segmentation (IJCV 2024 · similarity 0.94)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.93)
- Segmenting, Fast and Slow: Real-Time Open-Vocabulary Video Instance Segmentation with Dual-Path Processing (ECCV 2026 · similarity 0.92)
- Video Instance Segmentation in an Open-World (IJCV 2024 · similarity 0.92)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
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.
