Holistic Interaction Transformer Network for Action Detection

TL;DR

This paper introduces a multi-modal transformer network that integrates RGB and pose data, including hand and object interactions, to improve human action detection accuracy across multiple datasets.

Contribution

The novel HIT network effectively models multi-modal interactions with a new fusion mechanism and temporal cues, advancing action detection methods.

Findings

Outperforms previous methods on J-HMDB, UCF101-24, and MultiSports datasets.

Achieves competitive results on AVA dataset.

Demonstrates the importance of hand and pose information in action recognition.

Abstract

Actions are about how we interact with the environment, including other people, objects, and ourselves. In this paper, we propose a novel multi-modal Holistic Interaction Transformer Network (HIT) that leverages the largely ignored, but critical hand and pose information essential to most human actions. The proposed "HIT" network is a comprehensive bi-modal framework that comprises an RGB stream and a pose stream. Each of them separately models person, object, and hand interactions. Within each sub-network, an Intra-Modality Aggregation module (IMA) is introduced that selectively merges individual interaction units. The resulting features from each modality are then glued using an Attentive Fusion Mechanism (AFM). Finally, we extract cues from the temporal context to better classify the occurring actions using cached memory. Our method significantly outperforms previous approaches on the J-HMDB, UCF101-24, and MultiSports datasets. We also achieve competitive results on AVA. The code will be available at https://github.com/joslefaure/HIT.