Situation Recognition with Graph Neural Networks

TL;DR

This paper introduces a Graph Neural Network-based model for situation recognition in images, effectively capturing dependencies between semantic roles and outperforming previous methods by 3-5% in accuracy.

Contribution

The paper presents a novel GNN model for joint role dependency modeling in situation recognition, improving accuracy over prior approaches.

Findings

GNN model outperforms existing methods by 3-5% in full situation prediction.

Propagating information between roles enhances recognition accuracy.

Qualitative analysis reveals the influence of different roles in verbs.

Abstract

We address the problem of recognizing situations in images. Given an image, the task is to predict the most salient verb (action), and fill its semantic roles such as who is performing the action, what is the source and target of the action, etc. Different verbs have different roles (e.g. attacking has weapon), and each role can take on many possible values (nouns). We propose a model based on Graph Neural Networks that allows us to efficiently capture joint dependencies between roles using neural networks defined on a graph. Experiments with different graph connectivities show that our approach that propagates information between roles significantly outperforms existing work, as well as multiple baselines. We obtain roughly 3-5% improvement over previous work in predicting the full situation. We also provide a thorough qualitative analysis of our model and influence of different roles in the verbs.