Research map / Vision-language and multimodal / Human and embodied modeling
Visual Social Relationship Recognition
Junnan Li, Yongkang Wong, Qi Zhao, Mohan Kankanhalli
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 contextual-region second glance (Attentive R-CNN) to First-Glance | First-Glance · PISC relationship recognition (domain) | mAP (%): +5.2% | Improved |
| add contextual-region second glance (Attentive R-CNN) to First-Glance | First-Glance · PISC relationship recognition | mAP (%): +6.5% | Improved |
| add bounding-box location features to Pair | Pair · PISC relationship recognition | mAP (%): 54.9 -> 56.9 | Improved |
| add union region to Pair+Loc. | Pair+Loc. · PISC relationship recognition | mAP (%): 56.9 -> 58.7 | Improved |
| use whole image instead of union region as global input | First-Glance · PISC relationship recognition | mAP (%): 58.7 -> 58.3 | Hurt |
| add occupation scores to Dual-Glance | Dual-Glance · PISC relationship recognition | mAP (%): 65.2 -> 65.9 | Improved |
| fuse All Attributes SVM scores into Dual-Glance | Dual-Glance · PISC relationship recognition | mAP (%): 65.2 -> 65.4 | No clear effect |
| replace cross entropy with focal loss (ablation) | Dual-Glance · PISC relationship recognition (consistent set) | mAP (%): 63.9 -> 65.2 | Improved |
| replace focal loss with adaptive focal loss | Dual-Glance · PISC relationship recognition (consistent set) | mAP (%): 65.2 -> 66.4 | Improved |
| train on ambiguous set instead of consistent set with adaptive focal loss | Dual-Glance · PISC relationship recognition | mAP (%): 66.4 -> 68.3 | Improved |
| replace KL divergence with adaptive focal loss | Dual-Glance · PISC relationship recognition (ambiguous set) | mAP (%): 65.8 -> 68.3 | Improved |
| add attention to avg aggregation of regional scores (ablation) | Dual-Glance · PISC relationship recognition | mAP (%): 64.0 -> 68.3 | Improved |
Closest papers by meaning
- Dual-Glance Model for Deciphering Social Relationships (ICCV 2017 · similarity 0.95)
- Social Relation Recognition From Videos via Multi-Scale Spatial-Temporal Reasoning (CVPR 2019 · similarity 0.88)
- A Domain Based Approach to Social Relation Recognition (CVPR 2017 · similarity 0.87)
- SocialGPT: Prompting LLMs for Social Relation Reasoning via Greedy Segment Optimization (NeurIPS 2024 · similarity 0.85)
- SocialGesture: Delving into Multi-person Gesture Understanding (CVPR 2025 · similarity 0.85)
- Behavioral Geometric Supervision Aligns Video Foundation Models with Human Social Perception (NeurIPS 2026 · 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.
