2,341人会议场景手势识别数据【数据堂】

2,341人会议场景手势识别数据，包含黄种人、白种人、黑种人、棕色人种多任务，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18张图片和2段视频，图片包含单手握拳，单手比心等18个手势动作，视频包含鼓掌等手势动作

A conference-scenario gesture recognition dataset involving 2,341 participants. The dataset encompasses multi-task annotations across four ethnic groups: Mongoloid (yellow race), Caucasoid (white race), Negroid (black race) and Australoid (brown race), with the majority of participants being young and middle-aged adults. Data collection was carried out in multiple indoor office-related scenarios, including meeting rooms, cafes, libraries, bedrooms and other similar indoor environments. For each participant, 18 still images and 2 video clips were collected. The still images cover 18 gesture categories such as single-handed fist clenching and single-handed heart-shaped gestures, while the video clips include gesture actions like hand clapping.

搜集汇总

背景与挑战

背景概述

该数据集为会议场景手势识别数据，包含来自2,341人的图片与视频，涵盖多种人种、性别和年龄，采集于室内办公环境，手势动作包括单手握拳、比心等，数据准确率超过97%。