WiRLD

该数据集包含了从Wikidata中收集的10万个商业品牌的标志图像。每个实体都对应一张标志图片，这些图片是通过一个三阶段的过程从Wikidata中提取出来的。该数据集规模宏大，包含了10万张标志图片，涵盖了超过10万个标志类别。其任务是进行一次性的标志识别。

This dataset contains 100,000 commercial brand logo images collected from Wikidata. Each Wikidata entity corresponds to one logo image, which is extracted via a three-stage pipeline from Wikidata. Boasting a large-scale collection, this dataset includes 100,000 logo images covering over 100,000 logo categories, and its core task is one-shot logo recognition.