DWnet: Deep-Wide Network for 3D Action Recognition

TL;DR

DWnet is a novel deep-wide network that combines deep learning with broad learning systems to efficiently recognize 3D actions from skeletal data, achieving near real-time performance and superior feature extraction.

Contribution

The paper introduces DWnet, integrating PruHCN with BLS for improved action recognition and faster testing compared to traditional deep models.

Findings

Faster testing time approaching real-time performance.

Effective local and global spatial-temporal feature learning.

Superior recognition accuracy on skeletal datasets.

Abstract

We propose in this paper a deep-wide network (DWnet) which combines the deep structure with the broad learning system (BLS) to recognize actions. Compared with the deep structure, the novel model saves lots of testing time and almost achieves real-time testing. Furthermore, the DWnet can capture better features than broad learning system can. In terms of methodology, we use pruned hierarchical co-occurrence network (PruHCN) to learn local and global spatial-temporal features. To obtain sufficient global information, BLS is used to expand features extracted by PruHCN. Experiments on two common skeletal datasets demonstrate the advantage of the proposed model on testing time and the effectiveness of the novel model to recognize the action.