Motion Focus Recognition in Fast-Moving Egocentric Video

TL;DR

This paper introduces a real-time, motion focus recognition method for egocentric videos that estimates locomotion intention, enabling motion analysis in fast-moving scenarios for edge devices.

Contribution

The work presents a novel real-time motion focus recognition approach using foundation models and system optimizations for scalable inference in egocentric videos.

Findings

Achieves real-time performance with manageable memory use.

Effective on a collected egocentric action dataset.

Enables motion-centric analysis for edge deployment.

Abstract

From Vision-Language-Action (VLA) systems to robotics, existing egocentric datasets primarily focus on action recognition tasks, while largely overlooking the inherent role of motion analysis in sports and other fast-movement scenarios. To bridge this gap, we propose a real-time motion focus recognition method that estimates the subject's locomotion intention from any egocentric video. We leverage the foundation model for camera pose estimation and introduce system-level optimizations to enable efficient and scalable inference. Evaluated on a collected egocentric action dataset, our method achieves real-time performance with manageable memory consumption through a sliding batch inference strategy. This work makes motion-centric analysis practical for edge deployment and offers a complementary perspective to existing egocentric studies on sports and fast-movement activities.