Explanation-based Weakly-supervised Learning of Visual Relations with Graph Networks

TL;DR

This paper proposes a weakly-supervised approach using graph neural networks for visual relationship detection, enabling effective predicate classification and relation explanation with minimal image-level labels.

Contribution

It introduces a novel weakly-supervised method that frames relationship detection as explanation, leveraging graph neural networks and minimal annotations for improved robustness and generalization.

Findings

Achieves results comparable to fully- and weakly-supervised methods

Demonstrates robustness to incomplete annotations

Shows good few-shot generalization

Abstract

Visual relationship detection is fundamental for holistic image understanding. However, the localization and classification of (subject, predicate, object) triplets remain challenging tasks, due to the combinatorial explosion of possible relationships, their long-tailed distribution in natural images, and an expensive annotation process. This paper introduces a novel weakly-supervised method for visual relationship detection that relies on minimal image-level predicate labels. A graph neural network is trained to classify predicates in images from a graph representation of detected objects, implicitly encoding an inductive bias for pairwise relations. We then frame relationship detection as the explanation of such a predicate classifier, i.e. we obtain a complete relation by recovering the subject and object of a predicted predicate. We present results comparable to recent fully- and weakly-supervised methods on three diverse and challenging datasets: HICO-DET for human-object interaction, Visual Relationship Detection for generic object-to-object relations, and UnRel for unusual triplets; demonstrating robustness to non-comprehensive annotations and good few-shot generalization.