Dual-Glance Model for Deciphering Social Relationships
About
Since the beginning of early civilizations, social relationships derived from each individual fundamentally form the basis of social structure in our daily life. In the computer vision literature, much progress has been made in scene understanding, such as object detection and scene parsing. Recent research focuses on the relationship between objects based on its functionality and geometrical relations. In this work, we aim to study the problem of social relationship recognition, in still images. We have proposed a dual-glance model for social relationship recognition, where the first glance fixates at the individual pair of interest and the second glance deploys attention mechanism to explore contextual cues. We have also collected a new large scale People in Social Context (PISC) dataset, which comprises of 22,670 images and 76,568 annotated samples from 9 types of social relationship. We provide benchmark results on the PISC dataset, and qualitatively demonstrate the efficacy of the proposed model.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Social Relationship Recognition | PIPA-Relation | Accuracy59.6 | 16 | |
| Relation Recognition | PISC Fine | Friend Recall35.4 | 13 | |
| Relation Recognition | PISC Coarse | Intimate Recall73.1 | 11 | |
| Social Relationship Recognition (Fine-level) | PISC | Friends Acc35.4 | 10 | |
| Social Relation Recognition | PIPA (test) | Accuracy59.6 | 10 | |
| Social relationship recognition (3-relationship) | PISC 1.0 (test) | Accuracy (Intimate)73.1 | 9 | |
| Social relationship recognition (6-relationship) | PISC 1.0 (test) | Friends Accuracy35.4 | 9 | |
| Fine Social Relation Recognition | PISC (test) | Acc (Friends)35.4 | 7 | |
| Social Relationship Recognition (Coarse-level) | PISC | Intimate Score73.1 | 6 | |
| Coarse Social Relation Recognition | PISC (test) | Intimate Acc73.1 | 5 |