Egok360: A 360 Egocentric Kinetic Human Activity Video Dataset

TL;DR

EgoK360 is the first comprehensive 360-degree first-person activity dataset designed to advance research in egocentric 360-degree video understanding, including detailed annotations and experimental evaluations.

Contribution

The paper introduces EgoK360, a novel first-person 360-degree activity dataset with detailed annotations, filling a critical gap in egocentric 360-degree video research.

Findings

Two-stream network variants achieve promising results on EgoK360.

EgoK360 enables new research in 360-degree egocentric activity recognition.

Comprehensive analysis provided for different model architectures.

Abstract

Recently, there has been a growing interest in wearable sensors which provides new research perspectives for 360 {\deg} video analysis. However, the lack of 360 {\deg} datasets in literature hinders the research in this field. To bridge this gap, in this paper we propose a novel Egocentric (first-person) 360{\deg} Kinetic human activity video dataset (EgoK360). The EgoK360 dataset contains annotations of human activity with different sub-actions, e.g., activity Ping-Pong with four sub-actions which are pickup-ball, hit, bounce-ball and serve. To the best of our knowledge, EgoK360 is the first dataset in the domain of first-person activity recognition with a 360{\deg} environmental setup, which will facilitate the egocentric 360 {\deg} video understanding. We provide experimental results and comprehensive analysis of variants of the two-stream network for 360 egocentric activity recognition. The EgoK360 dataset can be downloaded from https://egok360.github.io/.