A Multi-Stream Convolutional Neural Network Framework for Group Activity Recognition

TL;DR

This paper introduces a multi-stream CNN framework for group activity recognition that leverages multiple modalities, including a novel pose-based modality, achieving state-of-the-art accuracy on Volleyball and Collective Activity datasets.

Contribution

The work proposes a multi-stream CNN architecture with modality-specific streams and a new pose-based modality for improved group activity recognition.

Findings

Achieves 90.50% accuracy on Volleyball dataset with multiple frames.

Achieves 86.61% accuracy on Volleyball dataset with single frame.

Achieves 87.01% accuracy on Collective Activity dataset with multiple frames.

Abstract

In this work, we present a framework based on multi-stream convolutional neural networks (CNNs) for group activity recognition. Streams of CNNs are separately trained on different modalities and their predictions are fused at the end. Each stream has two branches to predict the group activity based on person and scene level representations. A new modality based on the human pose estimation is presented to add extra information to the model. We evaluate our method on the Volleyball and Collective Activity datasets. Experimental results show that the proposed framework is able to achieve state-of-the-art results when multiple or single frames are given as input to the model with 90.50% and 86.61% accuracy on Volleyball dataset, respectively, and 87.01% accuracy of multiple frames group activity on Collective Activity dataset.