Research map / Video / Video segmentation and tracking
Unified Embedding Alignment for Open-Vocabulary Video Instance Segmentation
Hao Fang, Peng Wu, Yawei Li, Xinxin Zhang, Xiankai Lu
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 unified embedding alignment (cross-attention between instance queries and CLIP image features) (ablation) | OV2Seg baseline · LV-VIS val | mAP: 14.2 -> 20.7 | Improved |
| add unified embedding alignment (ablation) | OV2Seg baseline · YouTube-VIS 2019 (zero-shot) | mAP: 27.2 -> 34.8 | Improved |
| add unified embedding alignment (ablation) | OV2Seg baseline · LV-VIS val | mAP (novel): 11.9 -> 21.7 | Improved |
| use video-based training instead of image-level training (ablation) | OVFormer · YouTube-VIS 2019 (trained on YTVIS19) | mAP: +9.3 mAP | Improved |
| use semi-online clip-by-clip inference instead of offline whole-video inference (ablation) | OVFormer · LV-VIS val | mAP (novel): 19.9 -> 21.8 | Improved |
| use semi-online inference (ablation) | OVFormer · YouTube-VIS 2019 (trained on YTVIS19) | mAP (novel): 18.0 -> 22.3 | Improved |
| semi-online inference with video-based training (ablation) | OVFormer · LV-VIS val | mAP (base): 22.6 -> 22.1 | Hurt |
| increase inference clip length T (ablation) | OVFormer · LV-VIS val | mAP | Mixed |
| train on YouTube-VIS with novel labels kept (close-set) | OVFormer vs Mask2Former-VIS · YouTube-VIS 2019 | mAP: 46.4 -> 51.4 | Improved |
Closest papers by meaning
- Towards Open-Vocabulary Video Instance Segmentation (ICCV 2023 · similarity 0.94)
- OV-VIS: Open-Vocabulary Video Instance Segmentation (IJCV 2024 · similarity 0.92)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.90)
- Aligning Instance Brownian Bridge with Texts for Open-Vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.90)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.88)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.88)
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.
