Human Object Interaction Detection using Two-Direction Spatial Enhancement and Exclusive Object Prior

TL;DR

This paper introduces a novel spatial enhancement and regrouping method for human-object interaction detection, reducing false positives and improving robustness in complex scenes with multiple humans and objects.

Contribution

It proposes a two-direction spatial enhancement technique and an object-exclusive regrouping strategy to improve HOI detection accuracy.

Findings

Outperforms existing methods on V-COCO and HICO-DET datasets.

Reduces false positives caused by non-interactive human-object pairs.

Enhances robustness in scenes with multiple humans and objects.

Abstract

Human-Object Interaction (HOI) detection aims to detect visual relations between human and objects in images. One significant problem of HOI detection is that non-interactive human-object pair can be easily mis-grouped and misclassified as an action, especially when humans are close and performing similar actions in the scene. To address the mis-grouping problem, we propose a spatial enhancement approach to enforce fine-level spatial constraints in two directions from human body parts to the object center, and from object parts to the human center. At inference, we propose a human-object regrouping approach by considering the object-exclusive property of an action, where the target object should not be shared by more than one human. By suppressing non-interactive pairs, our approach can decrease the false positives. Experiments on V-COCO and HICO-DET datasets demonstrate our approach is more robust compared to the existing methods under the presence of multiple humans and objects in the scene.