COCO Captions
收藏arXiv2025-09-30 收录
下载链接:
https://github.com/tylin/coco-caption
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了与COCO图片相关的标题,这些标题被用于预训练图像-文本检索模型。此外,该数据集是预训练过程中所使用的一部分数据集,与之配合使用的还包括VG、SBU和CC等数据集。该数据集规模宏大,涵盖了400万张图片,涉及多个数据集。其主要任务是针对图像-文本检索进行预训练。
This dataset contains captions paired with COCO images, which are employed for pre-training image-text retrieval models. Furthermore, this dataset is part of the pre-training corpus, used in conjunction with other datasets including VG, SBU, and CC. Boasting a substantial scale, it encompasses 4 million images from multiple datasets. Its core purpose is to facilitate pre-training for image-text retrieval tasks.
提供机构:
Common Objects in Context (COCO)



