Research map / Video / Video segmentation and tracking
End-to-End Video Instance Segmentation via Spatial-Temporal Graph Neural Networks
Tao Wang, Ning Xu, Kean Chen, Weiyao Lin
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 |
|---|---|---|---|
| increase GNN update iterations from 1 to 3 (ablation) | VisSTG · YouTube-VIS val | AP: 32.1 -> 36.5 | Improved |
| increase GNN update iterations from 3 to 4 (ablation) | VisSTG · YouTube-VIS val | AP: 36.5 -> 36.0 | Hurt |
| add relation-based GNN feature fusion instead of no fusion (ablation) | VisSTG · YouTube-VIS val | AP: 31.6 -> 33.4 | Improved |
| use relation-based fusion instead of simple concatenation (ablation) | VisSTG · YouTube-VIS val | AP: 32.1 -> 33.4 | Improved |
| use edge-feature tracking instead of embedding association (ablation) | VisSTG · YouTube-VIS val | AP: 33.4 -> 34.8 | Improved |
| use edge-feature tracking instead of offset association (ablation) | VisSTG · YouTube-VIS val | AP: 32.6 -> 34.8 | Improved |
| add mask information propagation branch with edge-feature tracking (ablation) | VisSTG · YouTube-VIS val | AP: 34.8 -> 35.2 | Improved |
| add mask information propagation branch with embedding association and relation fusion (ablation) | VisSTG · YouTube-VIS val | AP: 33.4 -> 33.9 | Improved |
| add mask information propagation branch without feature fusion (ablation) | VisSTG · YouTube-VIS val | AP: 31.6 -> 32.0 | Improved |
| add multi-scale training (ablation) | VisSTG · YouTube-VIS val | AP: 35.2 -> 36.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- STEm-Seg — STEm-Seg: Spatio-temporal Embeddings for Instance Segmentation in Videos (ECCV 2020) · better on 3 of 3 shared results
- FEELVOS — FEELVOS: Fast End-To-End Embedding Learning for Video Object Segmentation (CVPR 2019) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Recurrent Graph Neural Networks for Video Instance Segmentation (IJCV 2022 · similarity 0.93)
- Video Instance Segmentation (ICCV 2019 · similarity 0.90)
- SG-Net: Spatial Granularity Network for One-Stage Video Instance Segmentation (CVPR 2021 · similarity 0.89)
- STEm-Seg: Spatio-temporal Embeddings for Instance Segmentation in Videos (ECCV 2020 · similarity 0.88)
- Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · similarity 0.87)
- VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation (CVPR 2024 · similarity 0.86)
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.
