Hierarchical Relationships: A New Perspective to Enhance Scene Graph Generation

Bowen Jiang,Camillo J. Taylor

DOI: https://doi.org/10.48550/arXiv.2303.06842

2023-11-29

Abstract:This paper presents a finding that leveraging the hierarchical structures among labels for relationships and objects can substantially improve the performance of scene graph generation systems. The focus of this work is to create an informative hierarchical structure that can divide object and relationship categories into disjoint super-categories in a systematic way. Specifically, we introduce a Bayesian prediction head to jointly predict the super-category of relationships between a pair of object instances, as well as the detailed relationship within that super-category simultaneously, facilitating more informative predictions. The resulting model exhibits the capability to produce a more extensive set of predicates beyond the dataset annotations, and to tackle the prevalent issue of low annotation quality. While our paper presents preliminary findings, experiments on the Visual Genome dataset show its strong performance, particularly in predicate classifications and zero-shot settings, that demonstrates the promise of our approach.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to significantly improve system performance by using the hierarchical structure between object and relationship labels in the Scene Graph Generation (SGG) task. Specifically, although existing scene graph generation methods have designed complex architectures to handle objects in visual scenes and their complex relationships, they often overlook the inherent hierarchical information in relationship categories, resulting in an incomplete understanding of the interactions between objects. To solve this problem, the author proposes a new classification scheme. By jointly predicting the super - class probability of relationships and the conditional probability distribution of relationships within each super - class through Bayes' rule, the performance of the scene graph generation system is enhanced.

The main contributions of the paper are as follows:
1. **Introducing Hierarchical Structure**: An informative hierarchical structure is created, which can systematically divide object and relationship categories into disjoint super - classes.
2. **Bayesian Prediction Head**: A Bayesian prediction head is designed, which can simultaneously predict the super - class and detailed relationships of relationships, improving the richness and accuracy of prediction.
3. **Experimental Verification**: The experimental results on the Visual Genome dataset show that this method performs excellently in predicate classification and zero - shot settings, demonstrating the effectiveness and potential of this method.

Through these innovations, the paper aims to solve the performance deficiency problems in existing scene graph generation methods caused by ignoring hierarchical information, thereby providing a more comprehensive and accurate scene understanding framework.