suchun/deep1b-mirror
收藏官方服务：
资源简介：
Deep1B 数据集是一个亿级规模的近似最近邻（ANN）基准测试数据集，包含10亿个96维浮点数图像描述符（deep image descriptors），用于特征提取和向量搜索任务。该数据集还包括一个公开的1万个查询测试集，旨在支持大规模ANN算法的可重复性评估。数据由Yandex Research原始发布，基于CVPR 2016的研究工作，镜像版本于2026-05-22创建，以确保数据访问的稳定性。
Deep1B dataset is a billion-scale Approximate Nearest Neighbor (ANN) benchmark consisting of 1,000,000,000 96-dimensional float32 deep image descriptors, along with a public test set of 10,000 queries. It is designed for feature extraction and vector search tasks, originally published by Yandex Research for reproducibility in ANN benchmarking. The mirrored version provides a byte-for-byte copy to ensure consistent access for research purposes.
提供机构：
suchun
