Action Anticipation By Predicting Future Dynamic Images

TL;DR

This paper introduces a novel approach for human action anticipation by predicting future dynamic images of human motion, significantly improving early action prediction accuracy across multiple benchmarks.

Contribution

The paper proposes a new method that uses dynamic images and tailored loss functions to predict future human motion, outperforming existing methods.

Findings

Outperforms current best methods by 4% on JHMDB-21

Achieves 5.2% improvement on UT-Interaction

Gains 5.1% accuracy on UCF 101-24

Abstract

Human action-anticipation methods predict what is the future action by observing only a few portion of an action in progress. This is critical for applications where computers have to react to human actions as early as possible such as autonomous driving, human-robotic interaction, assistive robotics among others. In this paper, we present a method for human action anticipation by predicting the most plausible future human motion. We represent human motion using Dynamic Images and make use of tailored loss functions to encourage a generative model to produce accurate future motion prediction. Our method outperforms the currently best performing action-anticipation methods by 4% on JHMDB-21, 5.2% on UT-Interaction and 5.1% on UCF 101-24 benchmarks.