Research map / Video / Video segmentation and tracking
JointFormer: A Unified Framework With Joint Modeling for Video Object Segmentation
Jiaming Zhang, Yutao Cui, Gangshan Wu, Limin Wang
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 |
|---|---|---|---|
| use joint modeling blocks instead of extract-then-match (ablation) | JointFormer · DAVIS 2017 test | J&F: 76.2 -> 86.4 | Improved |
| make reference tokens extract features without current tokens or other refs (mode d vs a) (ablation) | JointFormer · DAVIS 2017 test | J&F: 83.6 -> 86.4 | Improved |
| stop current tokens feeding references (mode d vs b) (ablation) | JointFormer · DAVIS 2017 test | J&F: 83.8 -> 86.4 | Improved |
| stop references sharing with each other (mode d vs c) (ablation) | JointFormer · DAVIS 2017 test | J&F: 83.9 -> 86.4 | Improved |
| make all blocks joint modeling blocks instead of first 6 (ablation) | JointFormer · DAVIS 2017 val | J&F: 87.3 -> 88.1 | Improved |
| joint modeling in first 6 blocks instead of last 6 (ablation) | JointFormer · DAVIS 2017 val | J&F: 87.1 -> 87.3 | Improved |
| make all blocks joint instead of evenly 2 blocks (ablation) | JointFormer · DAVIS 2017 val | J&F: 86.9 -> 88.1 | Improved |
| add compressed memory updated with 2 blocks (ablation) | JointFormer · DAVIS 2017 val | J&F: 88.1 -> 89.1 | Improved |
| update compressed memory with 3 blocks instead of 2 (ablation) | JointFormer · DAVIS 2017 val | J&F: 89.1 -> 88.1 | Hurt |
| replace compressed memory with XMem GRU (ablation) | JointFormer · DAVIS 2017 val | J&F: 88.1 -> 88.1 | No clear effect |
| add static-image pre-training | JointFormer · DAVIS 2017 val | J&F: 89.1 -> 89.7 | Improved |
| add BL30K pre-training | JointFormer · DAVIS 2017 val | J&F: 89.7 -> 90.1 | Improved |
Later papers that compared with this method
1 later accepted paper reports JointFormer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SAM 2: Segment Anything in Images and Videos (ICLR 2025 · better on 2 of 2 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
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.91)
- Look Before You Match: Instance Understanding Matters in Video Object Segmentation (CVPR 2023 · similarity 0.90)
- Learning Fast and Robust Target Models for Video Object Segmentation (CVPR 2020 · similarity 0.90)
- Video Object Segmentation Using Global and Instance Embedding Learning (CVPR 2021 · similarity 0.90)
- Breaking the "Object" in Video Object Segmentation (CVPR 2023 · similarity 0.89)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · 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.
