IMDB-WIKI-SbS

IMDB-WIKI-SbS是由Yandex莫斯科团队创建的大规模评估数据集，用于评估成对比较。该数据集基于著名的IMDB-WIKI数据集，包含9,150张图像，形成了250,249对，旨在通过人群来源的标注来平衡年龄和性别分布。数据集的创建过程涉及从原始数据集中选择照片，并通过Erd˝os-Rényi图生成成对比较，确保了数据集的连通性。该数据集主要应用于信息检索和推荐系统评估，旨在通过主观人类偏好来优化这些系统的性能。

IMDB-WIKI-SbS is a large-scale evaluation dataset developed by the Moscow team at Yandex, designed for pairwise comparison assessment. Built upon the well-known IMDB-WIKI dataset, it contains 9,150 images that form 250,249 image pairs, and aims to balance the age and gender distributions via crowd-sourced annotations. The dataset creation process entails selecting photographs from the original dataset and generating pairwise comparisons using Erdős-Rényi graphs, which ensures the connectivity of the dataset. This dataset is primarily utilized for information retrieval and recommendation system evaluation, with the goal of optimizing the performance of such systems through subjective human preferences.

搜集汇总

背景与挑战

背景概述

IMDB-WIKI-SbS是一个用于评估众包成对比较的数据集，包含来自IMDB和Wikipedia图像的人脸年龄比较标注。数据集提供众包标注者的成对比较结果（crowd_labels.csv）以及真实年龄标签（gt.csv），适用于研究众包标注质量和年龄估计任务。