Research map / Video / Video segmentation and tracking
InstanceFormer: An Online Video Instance Segmentation Framework
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 |
|---|---|---|---|
| Use InstanceFormer rather than TeViT | online video instance segmentation · OVIS validation | AP: 17.4 -> 20.0 | Helped |
| Use InstanceFormer rather than VISOLO | online video instance segmentation · YTVIS-19 validation | AP: 38.6 -> 45.6 | Helped |
| Use InstanceFormer rather than Mask2Former | video instance segmentation · YTVIS-21 validation | AP: 40.6 -> 40.8 | Helped |
| Add reference-point propagation to query propagation (ablation) | InstanceFormer prior propagation · OVIS | AP: 14.5 -> 16.5 | Helped |
| Add class-score propagation to query and reference propagation (ablation) | InstanceFormer prior propagation · OVIS | AP: 16.5 -> 17.1 | Helped |
| Add memory attention to query and reference propagation (ablation) | InstanceFormer · OVIS | AP: +1.7% AP | Helped |
| Add temporal contrastive loss to the model with memory attention (ablation) | InstanceFormer · OVIS | AP: +1.2% AP | Helped |
| Add memory and temporal contrastive loss (ablation) | InstanceFormer · YTVIS-22 long videos | AP_L: 23.5 -> 24.8 | Helped |
| Use 10 rather than 5 memory tokens per frame (ablation) | InstanceFormer · OVIS | AP: 15.1 -> 20.0 | Helped |
| Increase memory tokens per frame from 10 to 15 (ablation) | InstanceFormer · OVIS | AP: 20.0 -> 17.1 | Hurt |
| Increase memory tokens per frame from 10 to 20 (ablation) | InstanceFormer · OVIS | AP: 20.0 -> 14.5 | Hurt |
| Use 4 rather than 2 memory frames (ablation) | InstanceFormer · OVIS | AP: 19.0 -> 20.0 | Helped |
Closest papers by meaning
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.90)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.89)
- VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation (CVPR 2024 · similarity 0.89)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.88)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.88)
- Aligning Instance Brownian Bridge with Texts for Open-Vocabulary Video Instance Segmentation (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.