2D Pose Estimation based Child Action Recognition
Sanka Mohottala, Sandun Abeygunawardana, Pradeepa Samarasinghe, Dharshana Kasthurirathna, Charith Abhayaratne
arXiv:2212.09027v16 citationsh-index: 13
Originality Synthesis-oriented
AI Analysis
This addresses action recognition for children in unconstrained environments, but it is incremental as it adapts existing methods to a new dataset.
The paper tackled child action recognition by using a graph convolutional network with 2D pose estimation, achieving results comparable to an RGB-based model on a new benchmark dataset of unconstrained videos.
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.