Research map / Video / Video segmentation and tracking
Beyond Short-Term Snippet: Video Relation Detection With Spatio-Temporal Global Context
Check 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 multi-scale sliding windows to generate tracklet proposals of varying duration | Video relation detection · ImageNet-VidVRD | R@50: 8.07 -> 11.21 | Helped |
| Use multi-scale sliding windows to generate tracklet proposals of varying duration | Video relation detection · VidOR | R@50: 6.89 -> 8.21 | Helped |
| Add spatial and temporal GCNs to proposal-pair features (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.08 -> 17.58 | Helped |
| Add spatial GCN (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.08 -> 16.94 | Helped |
| Add temporal GCN (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.08 -> 16.91 | Helped |
| Use 3 spatial GCN layers rather than 0 layers (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.08 -> 16.94 | Helped |
| Increase spatial GCN depth from 3 to 4 layers (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.94 -> 16.63 | Hurt |
| Use 2 temporal GCN layers rather than 0 layers (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.08 -> 16.91 | Helped |
| Increase temporal GCN depth from 2 to 3 layers (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.91 -> 16.64 | Hurt |
| Add I3D features to detection features (ablation) | Relationship Classification module · ImageNet-VidVRD | R@50: 8.47 -> 9.14 | Helped |
| Add relative motion features to detection and I3D features (ablation) | Relationship Classification module · ImageNet-VidVRD | R@50: 9.14 -> 11.21 | Helped |
| Replace spatial-GCN-only proposal encoding with spatial and temporal GCNs (ablation) | Pair Proposal module · ImageNet-VidVRD | R@50: 16.94 -> 17.58 | Helped |
Closest papers by meaning
- VRDFormer: End-to-End Video Visual Relation Detection With Transformers (CVPR 2022 · similarity 0.90)
- Localize, Assemble, and Predicate: Contextual Object Proposal Embedding for Visual Relation Detection (AAAI 2020 · similarity 0.87)
- Multi-Modal Prompting for Open-Vocabulary Video Visual Relationship Detection (AAAI 2024 · similarity 0.86)
- Zero-shot Visual Relation Detection via Composite Visual Cues from Large Language Models (NeurIPS 2023 · similarity 0.86)
- Few-Shot Referring Relationships in Videos (CVPR 2023 · similarity 0.86)
- OED: Towards One-stage End-to-End Dynamic Scene Graph Generation (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.