Research map / Video / Video segmentation and tracking
Improving Video Instance Segmentation via Temporal Pyramid Routing
Xiangtai Li, Hao He, Yibo Yang, Henghui Ding, Kuiyuan Yang, Guangliang Cheng, Yunhai Tong, Dacheng Tao
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 gating in DACR (ablation) | TPR · Video instance segmentation (appendix ablation) | mAP: 35.3 -> 36.2 | Improved |
| add inner and outer gates (ablation) | TPR with ResNet-50 · Video instance segmentation (appendix ablation) | mAP: 34.5 -> 35.2 | Improved |
| add inner and outer gates (ablation) | TPR with ResNet-101 · Video instance segmentation (appendix ablation) | mAP: 37.0 -> 39.1 | Improved |
| add inner and outer gates (ablation) | TPR with Swin-Tiny · Video instance segmentation (appendix ablation) | mAP: 38.4 -> 40.0 | Improved |
| use CPR depth [4,2,1,1] instead of [3,1,1,1] (ablation) | CPR · Video instance segmentation (appendix ablation) | mAP: 35.4 -> 35.9 | Improved |
| deepen first-stage routing to 4 (ablation) | CPR · Video instance segmentation (appendix ablation) | mAP: 35.4 -> 35.6 | Improved |
| put depth 4 in a later stage instead of the first (ablation) | CPR · Video instance segmentation (appendix ablation) | mAP: 35.6 -> 34.9 | Hurt |
| increase depth of later stages to [4,2,2,2] (ablation) | CPR · Video instance segmentation (appendix ablation) | mAP: 35.9 -> 35.0 | Hurt |
| add CPR with default depth on top of DACR (ablation) | baseline+DACR · Video instance segmentation (appendix ablation) | mAP: 35.2 -> 35.9 | Improved |
| increase budget loss lambda1 from 1.5 to 5.5 (ablation) | TPR · Video instance segmentation (appendix ablation) | mAP: 36.2 -> 34.1 | Hurt |
| use lambda1=1.5 instead of 1.0 (ablation) | TPR · Video instance segmentation (appendix ablation) | mAP: 36.0 -> 36.2 | Improved |
Closest papers by meaning
- End-to-End Video Instance Segmentation With Transformers (CVPR 2021 · similarity 0.89)
- Video Instance Segmentation With a Propose-Reduce Paradigm (ICCV 2021 · similarity 0.89)
- InstanceFormer: An Online Video Instance Segmentation Framework (AAAI 2023 · similarity 0.89)
- InsPro: Propagating Instance Query and Proposal for Online Video Instance Segmentation (NeurIPS 2022 · similarity 0.89)
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation (AAAI 2025 · similarity 0.87)
- SyncVIS: Synchronized Video Instance Segmentation (NeurIPS 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.
