AI人体分析

服务支持检测图像中的所有人体并返回每个人体的矩形框位置，精准定位核心关键点；识别人体的静态属性和行为；识别和统计图像当中的人体个数（静态统计，暂不支持追踪和去重，以人头为主要识别目标统计人数，无需正脸、全身照，适应各类人流密集场景；默认识别整图中的人数，支持指定不规则区域的人数统计，同时可输出渲染图片）；识别图片中的手势类型）；识别人体的轮廓范围，与背景进行分离，适用于拍照背景替换、照片合成、身体特效等场景。This service supports detecting all human bodies in images and returning the bounding box coordinates for each individual, as well as accurately locating their core keypoints; recognizing the static attributes and behaviors of human bodies; identifying and counting the number of human bodies in images (static counting only, tracking and deduplication are not currently supported. The head is taken as the main recognition target for counting, and no frontal face or full-body photo is required, making it suitable for various crowded scenarios. By default, it counts the number of people in the entire image, supports counting the number of people in specified irregular regions, and can also output rendered images); recognizing gesture types in images; recognizing the contour range of human bodies and separating them from the background, which is applicable to scenarios such as photo background replacement, image composition, and body special effects.

搜集汇总

背景与挑战

背景概述

该数据集是一个人工智能人体分析API服务，发布于2019年3月，核心功能包括检测图像中的人体位置、定位14个关键点、识别人体属性和行为、统计人流量、识别手势以及进行人像分割。其特点在于集成多种分析能力，适用于人流密集场景和图像处理应用，如背景替换和照片合成。