Research map / Video / Video segmentation and tracking
Mask-Free Video Instance 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 |
|---|---|---|---|
| Replace BoxInst supervision with MaskFreeVIS temporal KNN-patch consistency | Mask2Former, ResNet-50 · YouTube-VIS 2019 val | AP: 38.6 -> 42.5 | Helped |
| Replace BoxInst supervision with MaskFreeVIS temporal KNN-patch consistency | VITA, ResNet-50 · OVIS | AP: 12.1 to 15.7 | Helped |
| Replace BoxInst supervision with MaskFreeVIS temporal KNN-patch consistency | Unicorn · BDD100K MOTS val | mMOTSA: 18.9 -> 23.8 | Helped |
| Add temporal KNN-patch loss to box-projection-only training (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 36.6 to 41.6 | Helped |
| Add spatial pairwise loss to box-projection-only training (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 36.6 -> 38.6 | Helped |
| Use one-to-K patch matching instead of one-to-one matching (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 40.8 to 42.5 | Helped |
| Use temporal KNN matching instead of RAFT flow matching (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 40.2 -> 42.5 | Helped |
| Use L2 rather than L1 patch distance (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 41.2 -> 42.5 | Helped |
| Use cyclic rather than sequential tube connections (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 41.9 -> 42.5 | Helped |
| Use dense instead of cyclic tube connections (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 42.5 -> 42.7 | Helped |
| Replace frame-wise sequence matching with spatio-temporal box-mask matching (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 37.6 -> 40.8 | Helped |
| Remove cross-entropy cost from ST-BoxMask matching (ablation) | MaskFreeVIS, ResNet-50 · YouTube-VIS 2019 val | AP: 40.8 -> 42.5 | Helped |
Closest papers by meaning
- Classifying, Segmenting, and Tracking Object Instances in Video with Mask Propagation (CVPR 2020 · similarity 0.91)
- InstanceFormer: An Online Video Instance Segmentation Framework (AAAI 2023 · similarity 0.91)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.90)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
- Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking (CVPR 2025 · similarity 0.89)
- S2D: Sparse-To-Dense Keymask Distillation for Unsupervised Video Instance Segmentation (NeurIPS 2026 · 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.