Ego-centric-Datasets

Eco-centric Dataset 是一个以第一人称视角（头戴式摄像头）拍摄的高清自我中心视频片段数据集。该数据集聚焦于日常活动、家务劳动和体力劳动，提供了复杂的手-物交互的“佩戴者视角”视图。与传统第三人称监控录像不同，该数据集捕捉了人类操作工具和与环境互动的细微差别。数据集包含以下字段：video（屏幕录制视频）、file_name（视频的唯一标识符）、chores（家务类型）、room（房间分隔）、device（设备型号）。数据集结构包含一个训练集，共22个样本。适用于动作识别、时序动作定位、物体检测与跟踪、手-物交互（HOI）等任务。潜在应用场景包括智能家居助手、工业培训与安全、增强现实（AR）指导和机器人学习（模仿学习）。数据集采用 CC BY 4.0 许可协议发布。

Eco-centric Dataset is a high-resolution egocentric video clip dataset captured from a first-person perspective (head-mounted camera). It focuses on daily activities, household chores and manual labor, providing "wearer's view" footage of complex hand-object interactions. Unlike traditional third-person surveillance footage, this dataset captures the subtle nuances of humans operating tools and interacting with their surrounding environments. The dataset includes the following fields: video (screen-recorded video), file_name (unique identifier for the video), chores (type of household chore), room (room categorization), device (device model). The dataset structure contains a training set with a total of 22 samples. It is applicable to tasks such as action recognition, temporal action localization, object detection and tracking, hand-object interaction (HOI), etc. Potential application scenarios include smart home assistants, industrial training and safety, augmented reality (AR) guidance, and robotic learning (imitation learning). The dataset is released under the CC BY 4.0 license.