vrg-prague/ilias

ILIAS是一个用于大规模实例级图像检索评估的大型测试数据集。它包含1000个跨不同领域的对象实例，总计5947张图像，包括1232张图像查询、4715张正图像以及1000个文本查询。此外，它还包括来自YFCC100M的1亿个干扰项，用于在大规模设置下评估检索性能。数据集分为图像查询、文本查询、核心数据库、小规模干扰项集和大规模干扰项集几个部分。

ILIAS is a large-scale test dataset for evaluation on Instance-Level Image retrieval At Scale. It contains 1,000 object instances across diverse domains, totaling 5,947 images, including 1,232 image queries, 4,715 positive images, and 1,000 text queries. Additionally, it includes 100M distractors from YFCC100M to evaluate retrieval performance under large-scale settings. The dataset is split into several parts: image queries, text queries, core database, a mini set of distractors, and a full set of distractors.