Visual Relationship Detection with Internal and External Linguistic Knowledge Distillation

TL;DR

This paper introduces a method that leverages internal and external linguistic knowledge distillation to improve visual relationship detection, especially for unseen relationships, by regularizing the learning process with linguistic statistics.

Contribution

It proposes a novel approach that distills linguistic knowledge from annotations and external text sources into a visual model to enhance generalization and zero-shot prediction capabilities.

Findings

Significant improvement in zero-shot recall on VRD dataset.

Outperforms state-of-the-art methods in visual relationship detection.

Effective use of linguistic knowledge from Wikipedia and annotations.

Abstract

Understanding visual relationships involves identifying the subject, the object, and a predicate relating them. We leverage the strong correlations between the predicate and the (subj,obj) pair (both semantically and spatially) to predict the predicates conditioned on the subjects and the objects. Modeling the three entities jointly more accurately reflects their relationships, but complicates learning since the semantic space of visual relationships is huge and the training data is limited, especially for the long-tail relationships that have few instances. To overcome this, we use knowledge of linguistic statistics to regularize visual model learning. We obtain linguistic knowledge by mining from both training annotations (internal knowledge) and publicly available text, e.g., Wikipedia (external knowledge), computing the conditional probability distribution of a predicate given a (subj,obj) pair. Then, we distill the knowledge into a deep model to achieve better generalization. Our experimental results on the Visual Relationship Detection (VRD) and Visual Genome datasets suggest that with this linguistic knowledge distillation, our model outperforms the state-of-the-art methods significantly, especially when predicting unseen relationships (e.g., recall improved from 8.45% to 19.17% on VRD zero-shot testing set).