人体姿势估计数据集

1. 简介该数据集专为人体姿势估计任务设计，用于识别和预测图像中人体主要关节的位置，包含标注了关键点的人体照片；采用关键点工具标记 18 个有序关键点，对应身体特定部位（如鼻子、颈部、肩、肘、腕、臀、膝、脚、眼、耳）；每张图像的关键点信息通过 XML 注释文件记录，注释包含每个关键点的 x、y 坐标，且带有 “Presumed_Location” 属性（指示关键点为推定或准确定义）；当前为数据示例，商业用途需通过指定渠道联系购买1. Introduction This dataset is specifically designed for human pose estimation tasks, which aim to identify and predict the positions of major human joints in images, and contains human photographs annotated with keypoints. It uses keypoint annotation tools to mark 18 ordered keypoints corresponding to specific body parts (e.g., nose, neck, shoulders, elbows, wrists, hips, knees, feet, eyes, ears). The keypoint information of each image is recorded via XML annotation files, which contain the x and y coordinates of each keypoint, along with the "Presumed_Location" attribute that indicates whether the keypoint is a presumed or accurately defined location. This is currently a data sample; for commercial use, please contact for purchase through the designated channel.

搜集汇总

背景与挑战

背景概述

该数据集是专为人体姿势估计任务设计的公开数据集，用于训练和开发识别图像中人体主要关节位置的模型。它包含标注了18个有序关键点（如鼻子、肩、肘等）的人体照片，关键点信息通过XML文件记录，包括坐标和推定属性，适用于生物识别、运动分析等计算机视觉应用。数据集采用cc-by-nc-nd-4.0许可，文件类型为zip，样本为图片。