Research map / Video / Video segmentation and tracking
Social Fabric: Tubelet Compositions for Video Relation Detection
Shuo Chen, Zenglin Shi, Pascal Mettes, Cees G. M. Snoek
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 visual, language, I3D and mask features to motion features (ablation) | Social Fabric · VidOR val | P@1: 50.97 -> 68.86 | Improved |
| add all features to motion-only features (ablation) | Social Fabric · VidOR val | mAP: 6.14 -> 11.21 | Improved |
| use 64 primitives instead of 1 (ablation) | Social Fabric encoding · VidOR val | mAP: 10.05 -> 11.21 | Improved |
| increase primitives from 64 to 128 (ablation) | Social Fabric encoding · VidOR val | mAP: 11.21 -> 11.01 | Hurt |
| use 64 primitives instead of 8 (ablation) | Social Fabric encoding · VidOR val | mAP: 10.69 -> 11.21 | Improved |
| add stage 1 interaction proposals with Social Fabric (ablation) | two-stage network · VidOR val | P@1: 60.72 -> 64.19 | Improved |
| add stage 1 interaction proposals (ablation) | two-stage network · VidOR val | Recall@50: 8.73 -> 9.62 | Improved |
| add stage 2 predicate prediction with Social Fabric (ablation) | two-stage network · VidOR val | P@1: 64.19 -> 68.86 | Improved |
| use Social Fabric instead of average pooling (ablation) | encoding · VidOR val | P@1: 62.73 -> 68.86 | Improved |
| use transformer encoding instead of average pooling (ablation) | encoding · VidOR val | mAP: 10.05 -> 10.07 | No clear effect |
| use NetRVLAD instead of NetVLAD (ablation) | encoding · VidOR val | mAP: 10.15 -> 10.55 | Improved |
| use Social Fabric instead of NetRVLAD (ablation) | encoding · VidOR val | mAP: 10.55 -> 11.21 | Improved |
Closest papers by meaning
- Video-based Human-Object Interaction Detection from Tubelet Tokens (NeurIPS 2022 · similarity 0.88)
- Social Relation Recognition From Videos via Multi-Scale Spatial-Temporal Reasoning (CVPR 2019 · similarity 0.87)
- Few-Shot Referring Relationships in Videos (CVPR 2023 · similarity 0.87)
- Visual Relationship Detection with Language Priors (ECCV 2016 · similarity 0.87)
- Localize, Assemble, and Predicate: Contextual Object Proposal Embedding for Visual Relation Detection (AAAI 2020 · similarity 0.86)
- SocialGPT: Prompting LLMs for Social Relation Reasoning via Greedy Segment Optimization (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.
