Retail-786k Dataset
收藏NIAID Data Ecosystem2026-05-01 收录
下载链接：
https://zenodo.org/record/7970566
下载链接
链接失效反馈官方服务：
资源简介：
This dataset is a large-scale dataset for ``visual entity matching''. Entity Matching (EM) indicates the identification of semantic groupings. As it occurs like the identification of consumer products in images in retail applications. A product is a retail item on sale that is uniquely identified by the internationally standardized Global Trade Item Number (GTIN). Entities are then defined as semantic groupings of products.
The dataset consists of 786,179 images labeled with 3,298 different entities. The images are split into sub-sets of 748,715 training images and 37,464 test images. The images of the dataset are cropped from scanned advertisement leaflets, collected over several years from different European retailers.
Two versions of the dataset are provided: once with the longer edge fixed to 512 and another one fixed to 256.
The comprehensive dataset description can be found in the corresponding paper: https://arxiv.org/abs/2309.17164
创建时间：
2023-10-02
