IGLUE

该数据集源自Flickr30K和COCO，其中的多语言文本经过了手动重新标注。需要注意的是，这个数据集可能与Multi30K和COCO上的性能表现关联不大。其任务是跨语言跨模态检索。

This dataset is derived from Flickr30K and COCO, whose multilingual texts have been manually re-annotated. It should be noted that this dataset may have little correlation with the performance on Multi30K and COCO. Its task is cross-lingual cross-modal retrieval.