Graph-Based Social Relation Reasoning

TL;DR

This paper introduces GR2N, a graph-based method for social relation recognition in images that jointly infers relations, improving accuracy and efficiency by modeling logical constraints among social relations.

Contribution

The paper presents a novel graph relational reasoning network that constructs multiple virtual relation graphs to better capture social relation patterns.

Findings

Improves social relation recognition accuracy.

Enhances efficiency over existing methods.

Generates consistent social relation graphs.

Abstract

Human beings are fundamentally sociable -- that we generally organize our social lives in terms of relations with other people. Understanding social relations from an image has great potential for intelligent systems such as social chatbots and personal assistants. In this paper, we propose a simpler, faster, and more accurate method named graph relational reasoning network (GR2N) for social relation recognition. Different from existing methods which process all social relations on an image independently, our method considers the paradigm of jointly inferring the relations by constructing a social relation graph. Furthermore, the proposed GR2N constructs several virtual relation graphs to explicitly grasp the strong logical constraints among different types of social relations. Experimental results illustrate that our method generates a reasonable and consistent social relation graph and improves the performance in both accuracy and efficiency.