2D Pose Estimation based Child Action Recognition

TL;DR

This paper introduces a graph convolutional network utilizing 2D pose estimation for child action recognition, achieving comparable results to RGB-based models on a new, unconstrained video dataset.

Contribution

It is the first to apply 2D pose-based graph convolutional networks to child action recognition, demonstrating effectiveness on a novel benchmark dataset.

Findings

Comparable performance to RGB models on child action recognition

First application of 2D pose estimation in this context

Effective on unconstrained environment videos

Abstract

We present a graph convolutional network with 2D pose estimation for the first time on child action recognition task achieving on par results with an RGB modality based model on a novel benchmark dataset containing unconstrained environment based videos.