Research map / Video / Video segmentation and tracking
MinVIS: A Minimal Video Instance Segmentation Framework without Video-based Training
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 |
|---|---|---|---|
| train a query-based image segmentation model and match query embeddings across frames | MinVIS with Swin-L · OVIS | AP: 28.9 -> 39.4 | Helped |
| train a query-based image segmentation model without video-based training | MinVIS with Swin-L · YouTube-VIS 2021 | AP: 52.6 -> 55.3 | Helped |
| train with only 1% of labeled frames (ablation) | MinVIS with Swin-L · YouTube-VIS 2021 | AP: 55.3 -> 52.9 | Hurt |
| train with only 1% of labeled frames (ablation) | MinVIS with Swin-L · OVIS | AP: 39.4 -> 31.7 | Hurt |
| replace query matching with mask-overlap heuristics (ablation) | MinVIS · YouTube-VIS 2019 | AP: 61.6 -> 58.2 | Hurt |
| replace query matching with mask-overlap heuristics (ablation) | MinVIS · OVIS | AP: 39.4 -> 31.7 | Hurt |
| combine mask-overlap heuristics with query matching (ablation) | MinVIS · YouTube-VIS 2019 | AP: 61.6 -> 61.3 | Hurt |
| combine mask-overlap heuristics with query matching (ablation) | MinVIS · OVIS | AP: 39.4 -> 39.1 | No clear effect |
| add supervised matching using the original frame sampling (ablation) | MinVIS · YouTube-VIS 2019 | AP: 61.6 -> 61.0 | Hurt |
| add supervised matching using the original frame sampling (ablation) | MinVIS · YouTube-VIS 2021 | AP: 55.3 -> 54.4 | Hurt |
| add supervised matching using the original frame sampling (ablation) | MinVIS · OVIS | AP: 39.4 -> 38.7 | Hurt |
| add supervised matching with consecutive-frame sampling (ablation) | MinVIS · OVIS | AP: 39.4 -> 39.6 | Helped |
Closest papers by meaning
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.94)
- A Generalized Framework for Video Instance Segmentation (CVPR 2023 · similarity 0.92)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.90)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.88)
- Decoupled Motion Expression Video Segmentation (CVPR 2025 · similarity 0.87)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 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.