Online Human Action Detection using Joint Classification-Regression Recurrent Neural Networks

TL;DR

This paper introduces a multi-task LSTM-based neural network for real-time human action detection from streaming skeleton data, achieving accurate localization and anticipation of actions without sliding windows.

Contribution

It presents a novel joint classification-regression RNN that captures long-range temporal dynamics and predicts action start and end points more precisely.

Findings

Effective action localization and forecasting demonstrated

High computational efficiency achieved without sliding windows

Promising results on both new and public datasets

Abstract

Human action recognition from well-segmented 3D skeleton data has been intensively studied and has been attracting an increasing attention. Online action detection goes one step further and is more challenging, which identifies the action type and localizes the action positions on the fly from the untrimmed stream data. In this paper, we study the problem of online action detection from streaming skeleton data. We propose a multi-task end-to-end Joint Classification-Regression Recurrent Neural Network to better explore the action type and temporal localization information. By employing a joint classification and regression optimization objective, this network is capable of automatically localizing the start and end points of actions more accurately. Specifically, by leveraging the merits of the deep Long Short-Term Memory (LSTM) subnetwork, the proposed model automatically captures the complex long-range temporal dynamics, which naturally avoids the typical sliding window design and thus ensures high computational efficiency. Furthermore, the subtask of regression optimization provides the ability to forecast the action prior to its occurrence. To evaluate our proposed model, we build a large streaming video dataset with annotations. Experimental results on our dataset and the public G3D dataset both demonstrate very promising performance of our scheme.