Scene Graph Generation by Iterative Message Passing

TL;DR

This paper introduces an end-to-end model for scene graph generation that iteratively refines object and relationship predictions using message passing, improving scene understanding accuracy.

Contribution

It presents a novel iterative message passing approach with RNNs for joint scene graph inference from images, outperforming previous methods.

Findings

Significantly outperforms previous scene graph generation methods on Visual Genome.

Effectively infers support relations in NYU Depth v2 dataset.

Utilizes contextual cues for better object and relationship predictions.

Abstract

Understanding a visual scene goes beyond recognizing individual objects in isolation. Relationships between objects also constitute rich semantic information about the scene. In this work, we explicitly model the objects and their relationships using scene graphs, a visually-grounded graphical structure of an image. We propose a novel end-to-end model that generates such structured scene representation from an input image. The model solves the scene graph inference problem using standard RNNs and learns to iteratively improves its predictions via message passing. Our joint inference model can take advantage of contextual cues to make better predictions on objects and their relationships. The experiments show that our model significantly outperforms previous methods for generating scene graphs using Visual Genome dataset and inferring support relations with NYU Depth v2 dataset.