MLM

该数据集是一个用于训练和评估多任务系统在多种模态下的资源，涵盖了跨模态检索和位置估算。它包含了来自236,000个人类居住地的维基数据条目的三种语言的文本、图像和位置数据。该数据集的主要任务是进行多模态下的多任务学习。

This dataset is a resource for training and evaluating multi-task systems across multiple modalities, covering cross-modal retrieval and position estimation. It contains text, images and location data in three languages for Wikidata entries from 236,000 human settlements. The primary task of this dataset is multi-task learning under multimodal scenarios.