Research map / Video / Video segmentation and tracking
Modelling Neighbor Relation in Joint Space-Time Graph for Video Correspondence Learning
Zixu Zhao, Yueming Jin, Pheng-Ann Heng
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 neighbor relation graph G_r (ablation) | similarity graph G_s · DAVIS 2017 | J&F_m: 65.6 -> 67.8 (+2.2%) | Improved |
| add node dropout on the joint graph (ablation) | G_s + G_r · DAVIS 2017 | J&F_m: 67.8 -> 68.7 (+3.1%) | Improved |
| enlarge neighborhood from 3x3 to 5x5 (25 nodes) (ablation) | neighbor relation graph · DAVIS 2017 | J&F_m | Hurt |
| use horizontal 1x3 neighborhood instead of vertical 3x1 (ablation) | neighbor relation graph · DAVIS 2017 | J&F_m | Improved |
| initialize edges with topology and learn them instead of fixed or random init (ablation) | edge E · DAVIS 2017 | J&F_m | Improved |
| use node dropout threshold 0.2 instead of 0 (ablation) | training · DAVIS 2017 | J&F_m | Improved |
| use pixel-discrepancy node dropout instead of random dropout (ablation) | training · DAVIS 2017 | J&F_m | Improved |
| use video temporal views instead of single-frame augmentation for contrastive learning | Ours vs MoCo · DAVIS 2017 | J&F_m: 60.8 -> 68.7 | Improved |
| add neighbor relations and node dropout to VideoWalk | VideoWalk · VIP | mIoU: 38.6 -> 40.2 | Improved |
| add neighbor relations and node dropout to VideoWalk | VideoWalk · DAVIS 2017 | J&F_m: 67.6 -> 68.7 | Improved |
| add neighbor relations and node dropout to VideoWalk | VideoWalk · JHMDB | PCK@0.1: 59.3 -> 61.4 | Improved |
Closest papers by meaning
- Space-Time Correspondence as a Contrastive Random Walk (NeurIPS 2020 · similarity 0.94)
- Exposing the Self-Supervised Space-Time Correspondence Learning via Graph Kernels (AAAI 2023 · similarity 0.93)
- Rethinking Self-Supervised Correspondence Learning: A Video Frame-Level Similarity Perspective (ICCV 2021 · similarity 0.92)
- Joint-task Self-supervised Learning for Temporal Correspondence (NeurIPS 2019 · similarity 0.92)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.88)
- When the Future Becomes the Past: Taming Temporal Correspondence for Self-supervised Video Representation Learning (CVPR 2025 · 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.
