UniHOI: Unified Human-Object Interaction Understanding via Unified Token Space

TL;DR

UniHOI introduces a unified framework for human-object interaction detection and generation, leveraging a shared token space and semi-supervised learning to improve accuracy and generalization in diverse tasks.

Contribution

The paper presents a novel unified token space and symmetric attention module that jointly models HOI detection and generation, enabling bidirectional interaction understanding.

Findings

Achieves state-of-the-art performance in HOI detection and generation.

Improves long-tailed HOI detection accuracy by 4.9%.

Boosts open-vocabulary interaction generation metrics by 42.0%.

Abstract

In the field of human-object interaction (HOI), detection and generation are two dual tasks that have traditionally been addressed separately, hindering the development of comprehensive interaction understanding. To address this, we propose UniHOI, which jointly models HOI detection and generation via a unified token space, thereby effectively promoting knowledge sharing and enhancing generalization. Specifically, we introduce a symmetric interaction-aware attention module and a unified semi-supervised learning paradigm, enabling effective bidirectional mapping between images and interaction semantics even under limited annotations. Extensive experiments demonstrate that UniHOI achieves state-of-the-art performance in both HOI detection and generation. Specifically, UniHOI improves accuracy by 4.9% on long-tailed HOI detection and boosts interaction metrics by 42.0% on open-vocabulary generation tasks.