Research map / Video / Video segmentation and tracking
Learning Motion-Appearance Co-Attention for Zero-Shot Video Object Segmentation
Shu Yang, Lu Zhang, Jinqing Qi, Huchuan Lu, Shuo Wang, Xiaoxing Zhang
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 Multi-Modality Co-Attention Gate (ablation) | concat baseline · DAVIS-16 val (ablation) | J&F: 77.4 -> 81.0 | Improved |
| add the gate function inside MCG (ablation) | MCG · DAVIS-16 val (ablation) | J&F: 78.7 -> 81.0 | Improved |
| add a single Motion Correction Module (ablation) | baseline + MCG · DAVIS-16 val (ablation) | J&F: 81.0 -> 81.4 | Improved |
| stack MCMs on three decoder blocks (ablation) | baseline + MCG · DAVIS-16 val (ablation) | J&F: 81.4 -> 82.9 | Improved |
| stack MCMs on all five decoder blocks (ablation) | baseline + MCG · DAVIS-16 val (ablation) | J&F: 82.9 -> 83.7 | Improved |
| add fully-connected CRF post-processing (ablation) | AMC-Net · DAVIS-16 val (ablation) | J&F: 83.7 -> 84.4 | Improved |
| replace PWCNet flow with RAFT (ablation) | AMC-Net · DAVIS-16 val (ablation) | J&F: 84.4 -> 84.6 | Improved |
| apply MCG at all encoder stages instead of none (ablation) | AMC-Net · DAVIS-16 val (ablation) | J&F: 81.1 -> 83.7 | Improved |
| add supervision on visual-motion attention at all decoder blocks (ablation) | AMC-Net · DAVIS-16 val (ablation) | J&F: 81.5 -> 83.7 | Improved |
| use better optical flow (FlowNetS to FlowNet2CS to PWCNet) (ablation) | AMC-Net · DAVIS-16 val (ablation) | J&F: 80.5 -> 83.1 -> 84.4 | Improved |
| AMC-Net with RAFT versus 3DC-Seg* | AMC-Net · DAVIS-16 val | J&F: 84.5 -> 84.6 | No clear effect |
| AMC-Net versus EGMN | AMC-Net · Youtube-Objects | J mean: 71.4 -> 71.1 | No clear effect |
Closest papers by meaning
- Motion-Attentive Transition for Zero-Shot Video Object Segmentation (AAAI 2020 · similarity 0.89)
- Zero-Shot Video Object Segmentation with Co-Attention Siamese Networks (TPAMI 2020 · similarity 0.88)
- See More, Know More: Unsupervised Video Object Segmentation With Co-Attention Siamese Networks (CVPR 2019 · similarity 0.87)
- Appearance-based Refinement for Object-Centric Motion Segmentation (ECCV 2024 · similarity 0.86)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.85)
- Dual Prototype Attention for Unsupervised Video Object Segmentation (CVPR 2024 · similarity 0.84)
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.
