数据堂—2,341人会议场景人体行为识别数据

2,341人会议场景人体行为识别数据，包含黄种人、白种人、黑种人、棕色人种，年龄以中青年为主，采集了多种室内办公场景，涵盖会议室、咖啡厅、图书馆、卧室等，每人采集了11段视频包含左右晃动身体、吃东西、伸懒腰等人体行为

This is a human action recognition dataset for meeting scenarios, involving 2,341 subjects. It covers four ethnic groups: East Asians, Caucasians, Black individuals, and Brown people, with most participants being young and middle-aged adults. The data were collected in various indoor office-related scenarios, including conference rooms, cafes, libraries, and bedrooms. Each subject contributed 11 video clips capturing common human actions such as left-right body swaying, eating, and stretching.

搜集汇总

背景与挑战

背景概述

该数据集专注于会议场景下的人体行为识别，包含2,341人的视频数据，每人提供11个视频片段，涵盖多种室内办公环境如会议室和咖啡馆。数据涉及不同种族和年龄段，以中青年为主，视频格式为.mp4和.mov，行为识别的准确率超过97%。