CC152K

该数据集名为CC152K，是概念性标题（Conceptual Captions）的一个子集，包含了从互联网上搜集的152,000组图像和文本配对。由于缺乏人工标注，CC152K中大约有20%的样本配对存在不匹配的情况。该数据集的规模为152,000组图像和文本配对，其任务是跨模态检索。The dataset named CC152K is a subset of Conceptual Captions. It contains 152,000 pairs of images and their corresponding texts crawled from the Internet. Due to the lack of manual annotations, approximately 20% of the sample pairs in CC152K are mismatched. With a total of 152,000 image-text pairs, this dataset is designed for cross-modal retrieval tasks.

搜集汇总

背景与挑战

背景概述

CC152K是概念性标题的子集，包含15.2万组图像-文本配对，其中约20%存在不匹配问题，适用于跨模态检索研究。