Research map / Detection and segmentation / Open-set and few-shot detection
Dual-Glance Model for Deciphering Social Relationships
Junnan Li, Yongkang Wong, Qi Zhao, Mohan S. 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 bounding box geometry features to Pair-CNN (ablation) | Pair-CNN · PISC 3-relationship | mAP: 65.1 -> 69.6 | Improved |
| add bounding box geometry features to Pair-CNN (ablation) | Pair-CNN · PISC 6-relationship | mAP: 48.2 -> 54.3 | Improved |
| add union-region CNN to Pair-CNN+BBox (first glance) (ablation) | Pair-CNN+BBox · PISC 3-relationship | mAP: 69.6 -> 72.2 | Improved |
| use whole image (global) instead of union region as context (ablation) | Pair-CNN+BBox+Union · PISC 3-relationship | mAP: 72.2 -> 70.5 | Hurt |
| use Places-pretrained scene CNN instead of union region (ablation) | Pair-CNN+BBox+Union · PISC 3-relationship | mAP: 72.2 -> 68.0 | Hurt |
| add contextual region second glance with attention to first glance | Pair-CNN+BBox+Union · PISC 3-relationship | mAP: 72.2 -> 79.7 | Improved |
| add contextual region second glance with attention to first glance | Pair-CNN+BBox+Union · PISC 6-relationship | mAP: 56.9 -> 63.2 | Improved |
| add attention to second glance with lse aggregation (ablation) | Dual-Glance · PISC 3-relationship | mAP: 73.0 -> 79.7 | Improved |
| add attention to second glance with avg aggregation (ablation) | Dual-Glance · PISC 6-relationship | mAP: 55.9 -> 61.8 | Improved |
| use lse instead of max aggregation with attention (ablation) | Dual-Glance with attention · PISC 3-relationship | mAP: 77.6 -> 79.7 | Improved |
| use max instead of lse aggregation without attention (ablation) | Dual-Glance without attention · PISC 3-relationship | mAP: 73.0 -> 74.8 | Improved |
| average scores over both pair orderings at test time (ablation) | Dual-Glance · PISC | accuracy: 0.5%-1% | Improved |
Closest papers by meaning
- Visual Social Relationship Recognition (IJCV 2020 · similarity 0.95)
- A Domain Based Approach to Social Relation Recognition (CVPR 2017 · similarity 0.87)
- Social Relation Recognition From Videos via Multi-Scale Spatial-Temporal Reasoning (CVPR 2019 · similarity 0.86)
- SocialGesture: Delving into Multi-person Gesture Understanding (CVPR 2025 · similarity 0.84)
- PairDETR : Joint Detection and Association of Human Bodies and Faces (CVPR 2024 · similarity 0.84)
- Behavioral Geometric Supervision Aligns Video Foundation Models with Human Social Perception (NeurIPS 2026 · similarity 0.83)
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.
