Dual-Glance Model for Deciphering Social Relationships
This addresses the problem of understanding social interactions in images for computer vision applications, but it is incremental as it builds on existing object relationship research.
The paper tackles social relationship recognition in still images by proposing a dual-glance model that uses attention to explore contextual cues, and it introduces a new large-scale dataset (PISC) with 22,670 images and 76,568 annotated samples, providing benchmark results.
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.