Research map / Video / Video segmentation and tracking
Exposing the Self-Supervised Space-Time Correspondence Learning via Graph Kernels
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 VideoHiGraph instead of CLTC | self-supervised video object segmentation · DAVIS-2017 | J &Fm: 70.3 -> 70.5 | Helped |
| use VideoHiGraph instead of CLTC | self-supervised video object segmentation · YouTube-VOS | Overall: 67.3 -> 67.8 | Helped |
| use VideoHiGraph instead of CRW | self-supervised video part segmentation · VIP | mIoU: 38.6 -> 39.6 | Helped |
| use VideoHiGraph instead of CLTC | self-supervised video part segmentation · VIP | AP: 19.1 -> 20.9 | Helped |
| use VideoHiGraph instead of CLTC | self-supervised pose keypoint tracking · JHMDB | PCK@0.1: 60.5 -> 61.7 | Helped |
| use VideoHiGraph instead of CRW | self-supervised pose keypoint tracking · JHMDB | PCK@0.2: 84.9 -> 85.2 | Helped |
| use VideoHiGraph instead of STEm-Seg | video instance segmentation · YouTube-VIS21 val | AP: 30.6 -> 30.8 | Helped |
| replace the biased graph kernel with the shortest path kernel (ablation) | VideoHiGraph · DAVIS-2017 | J &Fm: 67.7 vs 70.5 | Hurt |
| replace the biased graph kernel with the graphlet kernel (ablation) | VideoHiGraph · DAVIS-2017 | J &Fm: 68.5 vs 70.5 | Hurt |
| replace the biased graph kernel with the Weisfeiler-Lehman subtree kernel (ablation) | VideoHiGraph · DAVIS-2017 | J &Fm: 68.9 vs 70.5 | Hurt |
| add node-level correspondence loss (ablation) | VideoHiGraph · DAVIS-2017 | J &Fm: 68.4 -> 70.5 | Helped |
| add graph-level correspondence loss (ablation) | VideoHiGraph · DAVIS-2017 | J &Fm: +1.5 | Helped |
Closest papers by meaning
- Space-Time Correspondence as a Contrastive Random Walk (NeurIPS 2020 · similarity 0.93)
- Contrastive Learning for Space-Time Correspondence via Self-Cycle Consistency (CVPR 2022 · similarity 0.90)
- Spatial-Then-Temporal Self-Supervised Learning for Video Correspondence (CVPR 2023 · similarity 0.89)
- Learning Correspondence From the Cycle-Consistency of Time (CVPR 2019 · similarity 0.88)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.87)
- Motion Graph Unleashed: A Novel Approach to Video Prediction (NeurIPS 2024 · similarity 0.85)
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.