MIAP 行人检测图像数据集

MIAP 全称 More Inclusive Annotations for People，是 Open Images 数据集中一个子集的拓展，共包含 10 万张图像（7 万张来自训练集，3 万张来自测试/验证集）。图像中的所有可见人物均标注边界框和属性，属性包括性别（主要是女性、男性以及未知）和年龄范围（年轻、中年、老年以及未知）。该数据集的图像标注比原数据集的更详细，增加了近 10 万个边界框，其主要用于机器学习公平性研究（ML Research）。

MIAP, whose full name is More Inclusive Annotations for People, is an extension of a subset of the Open Images dataset. It contains a total of 100,000 images, with 70,000 sourced from the training set and 30,000 from the test/validation set. For all visible persons in these images, both bounding boxes and their corresponding attributes are annotated. The annotated attributes include gender (mainly female, male, and unknown) and age range (young, middle-aged, elderly, and unknown). This dataset has more detailed annotations than the original dataset, adding nearly 100,000 additional bounding boxes. It is primarily utilized for machine learning fairness research (ML Research).

搜集汇总

背景与挑战

背景概述

MIAP行人检测图像数据集是Open Images数据集的扩展子集，包含10万张图像，所有可见人物均标注了边界框及性别和年龄属性。该数据集主要用于机器学习公平性研究，其标注比原数据集更详细。