A Training Method For VideoPose3D With Ideology of Action Recognition

TL;DR

This paper introduces a faster, flexible training method for VideoPose3D based on action recognition, enabling efficient pose estimation with less data and improved accuracy for action-oriented tasks.

Contribution

The proposed method leverages action recognition to enhance VideoPose3D training, reducing data requirements and improving performance on action-oriented pose estimation tasks.

Findings

Requires less data for pose estimation tasks.

Outperforms original model by 4.5% on Velocity Error of MPJPE.

Handles both action-oriented and common pose estimation.

Abstract

Action recognition and pose estimation from videos are closely related to understand human motions, but more literature focuses on how to solve pose estimation tasks alone from action recognition. This research shows a faster and more flexible training method for VideoPose3D which is based on action recognition. This model is fed with the same type of action as the type that will be estimated, and different types of actions can be trained separately. Evidence has shown that, for common pose-estimation tasks, this model requires a relatively small amount of data to carry out similar results with the original research, and for action-oriented tasks, it outperforms the original research by 4.5% with a limited receptive field size and training epoch on Velocity Error of MPJPE. This model can handle both action-oriented and common pose-estimation problems.