602人3,010张多人种人体语义分割数据【数据堂】

602人3,010张多人种人体语义分割数据包括不同人体动作、不同年龄段、不同人种、不同采集背景。在标注方面，对耳机、人体、背景、眼镜进行抠图标注。602人3,010张多人种人体语义分割数据可用于人体分割、视频会议行为检测等任务

This dataset contains 3,010 multi-ethnic human semantic segmentation images collected from 602 subjects, covering diverse human actions, age groups, ethnicities and acquisition backgrounds. For annotation, pixel-wise segmentation annotations are conducted for four categories including headphones, human bodies, backgrounds and glasses. This dataset can be applied to tasks such as human body segmentation and video conference behavior detection.

搜集汇总

背景与挑战

背景概述

该数据集包含602人的3,010张图像，覆盖黄、黑、白、棕四种人种，年龄在16-60岁之间，性别比例均衡。图像采集于办公室、咖啡店等多样环境，标注了耳机、人体等元素，适用于人体分割和行为检测任务。