人体检测和属性识别

对于输入的一张图片（可正常解码，且长宽比适宜），检测图像中的所有人体并返回每个人体的矩形框位置，识别人体的静态属性和行为，共支持22种属性，包括：性别、年龄阶段、上下身服饰（含类别/颜色）、是否戴帽子、是否戴口罩、是否戴眼镜、背包、是否吸烟、是否使用手机、人体朝向等。
计费说明：并发量为10QPS时，每月收费5000；
并发量为50QPS时，每月收费21000；
并发量为100QPS时，每月收费3

Given an input image that can be properly decoded with an appropriate aspect ratio, this service detects all human bodies in the image and returns the bounding box coordinates for each human body, while recognizing the static attributes and behaviors of the detected humans. A total of 22 attributes are supported, including gender, age group, upper and lower body clothing (including category and color), hat-wearing status, mask-wearing status, glasses-wearing status, whether carrying a backpack, smoking status, mobile phone usage status, and human orientation, etc. Pricing Instructions: For a concurrency of 10 QPS, the monthly charge is 5000; for a concurrency of 50 QPS, the monthly charge is 21000; for a concurrency of 100 QPS, the monthly charge is 3.

搜集汇总

背景与挑战

背景概述

该数据集专注于人体检测和属性识别，支持22种人体静态属性和行为的识别，包括性别、年龄阶段、服饰等。数据集适用于人脸与人体识别领域，发布于2021年2月，每年进行增量更新。