Research map / Video / Video segmentation and tracking
Efficient Video Instance Segmentation via Tracklet Query and Proposal
Check 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 factorised temporal and spatial self-attention instead of joint attention (ablation) | EfficientVIS · YouTube-VIS 2019 val | AP: 33.7 -> 37.0 | Helped |
| Use temporal dynamic convolution over nearby frames instead of still-image dynamic convolution (ablation) | EfficientVIS · YouTube-VIS 2019 val | AP: 36.0 -> 37.0 | Helped |
| Increase video clip length from T=9 to T=36 (ablation) | EfficientVIS · YouTube-VIS 2019 val | AP: 35.3 -> 37.0 | Helped |
| Train inter-clip tracklet linking with correspondence learning (ablation) | EfficientVIS · YouTube-VIS 2019 val, T=9 | AP: 7.4 -> 35.3 | Helped |
| Use a separate embedding for each time step in each tracklet query (ablation) | EfficientVIS · YouTube-VIS 2019 val | AP: 35.5 -> 37.0 | Helped |
| Use fully end-to-end learned inter-clip linking instead of hand-crafted association | EfficientVIS · YouTube-VIS 2019 val, T=18 | AP: 34.6 -> 36.4 | Helped |
| Use multi-scale training (ablation) | EfficientVIS · YouTube-VIS 2019 val, ResNet-50 | AP: 37.0 -> 37.9 | Helped |
| Downsample the video to 1.5 FPS (ablation) | EfficientVIS · YouTube-VIS 2019 val, T=9 | AP: 35.3 -> 35.3 | No clear effect |
Closest papers by meaning
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 2024 · similarity 0.90)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.89)
- Decoupled Motion Expression Video Segmentation (CVPR 2025 · similarity 0.89)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.88)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.88)
- VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation (CVPR 2024 · 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.