Research map / Video / Video segmentation and tracking
Dynamic in Static: Hybrid Visual Correspondence for Self-Supervised Video Object Segmentation
Yazhou Yao, Gensheng Pei, Jianbo Jiao, Wenguan Wang, Liqiang Nie, Jinhui Tang
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 static visual correspondence to MoCo baseline (ablation) | HVC · DAVIS17 val | J&F mean: 67.1 -> 70.6 (+3.5) | Improved |
| add dynamic (pseudo-dynamic) correspondence to MoCo baseline (ablation) | HVC · DAVIS17 val | J&F mean: 67.1 -> 69.9 (+2.8) | Improved |
| add both static and dynamic correspondence (ablation) | HVC · DAVIS17 val | J&F mean: 67.1 -> 73.1 (+6.0) | Improved |
| add both static and dynamic correspondence (ablation) | HVC · DAVIS16 val | J&F mean: 72.5 -> 80.1 (+7.6) | Improved |
| add static correspondence only (ablation) | HVC · DAVIS16 val | J&F mean: 72.5 -> 77.4 (+4.9) | Improved |
| add dynamic correspondence only (ablation) | HVC · DAVIS16 val | J&F mean: 72.5 -> 77.0 (+4.5) | Improved |
| add color jittering to augmentation (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 71.0 | Hurt |
| add Gaussian blur to augmentation (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 71.4 | Hurt |
| add affine/perspective transformation (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 72.9 | No clear effect |
| reduce COCO data by sampling every 5th image (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 72.2 | Hurt |
| reduce COCO data by sampling every 10th image (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 71.2 | Hurt |
| lower dynamic loss weight from 1.0 to 0.1 (ablation) | HVC · DAVIS17 val | J&F mean: 73.1 -> 71.9 | Hurt |
Closest papers by meaning
- Betrayed by Attention: A Simple yet Effective Approach for Self-supervised Video Object Segmentation (ECCV 2024 · similarity 0.90)
- Unified Mask Embedding and Correspondence Learning for Self-Supervised Video Segmentation (CVPR 2023 · similarity 0.89)
- Boosting Video Object Segmentation via Space-Time Correspondence Learning (CVPR 2023 · similarity 0.89)
- HODOR: High-Level Object Descriptors for Object Re-Segmentation in Video Learned From Static Images (CVPR 2022 · similarity 0.89)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.88)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 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.
